University of Southern California Dissertations and Theses

The impact of economic shocks on firm behavior: insights from three studies

(USC Thesis Other)

The impact of economic shocks on firm behavior: insights from three studies

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content The Impact of Economic Shocks on Firm Behavior: Insights from Three Studies by Yukun Ding A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY [ECONOMICS] May 2023 Copyright 2023 Yukun Ding Table of Contents List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii Chapter 1 How Do First and Second Moment Trade Policy Shocks Af- fect Firm Export Activities? . . . . . . . . . . . . . . . . . . . . 1 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 1.2.1 Trade Policy Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.2.2 Supply Chain Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 1.3 The Effects of Trade Policy Uncertainty . . . . . . . . . . . . . . . . . . . . 16 1.3.1 Effects of Trade Policy Uncertainty on Extensive Margin . . . . . . . 18 1.3.2 Export Adjustments . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 1.3.3 Sample Selection Issues . . . . . . . . . . . . . . . . . . . . . . . . . . 26 1.4 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 1.4.1 Model Mechanism. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 1.4.2 Trade Policy Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . 33 1.4.3 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 1.4.4 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 1.4.5 Model Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 ii 1.5 Policy Implication and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . 43 1.5.1 Policy Implication . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 1.5.2 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 Chapter 2 Navigating Uncertainty: A Study of Input Choices in the Face of Shocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 2.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 2.2.1 Demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 2.2.2 Production and Trade . . . . . . . . . . . . . . . . . . . . . . . . . . 52 2.2.3 Optimal Sourcing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 2.2.4 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 2.2.5 Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 2.2.6 Intensity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 2.3 Empirical . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 2.3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 2.3.2 Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 2.3.3 Intensity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 Chapter 3 An Anatomy of Exports Around Economic Downturns . . . 74 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 3.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 3.2.1 Economic Contractions . . . . . . . . . . . . . . . . . . . . . . . . . . 77 3.2.2 Trade Margins. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 3.3 Quantifying the Effects of Economic Contractions . . . . . . . . . . . . . . . 82 3.3.1 The Effects on Extensive and Intensive Margins . . . . . . . . . . . . 82 3.3.2 Decomposed Intensive Margin . . . . . . . . . . . . . . . . . . . . . . 86 iii 3.3.3 Decomposed Extensive Margin. . . . . . . . . . . . . . . . . . . . . . 88 3.4 Policy Implication and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . 92 3.4.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 Appendix B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 Appendix C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 Appendix D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 iv List of Tables 1.1 Firm Supply Chain Data Summary Statistics (2015-2019 Firm-Quarter Panel) 16 1.2 Definitions of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 1.3 TheEffectsofExpectationandUncertaintyShockonEntry,Exit,andInvest- ment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 1.4 TheEffectsofExpectationandUncertaintyShockonEntry,Exit,andInvest- ment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 1.5 Summary of Parameters for Tariff Movement . . . . . . . . . . . . . . . . . . 35 1.6 Benchmark Parameter Summary. . . . . . . . . . . . . . . . . . . . . . . . . 39 2.1 Import Value and Tariff Environment . . . . . . . . . . . . . . . . . . . 66 2.2 Import Growth and Trade Policy Uncertainty Firm-Product-Year Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 2.3 Import Growth and Trade Policy Uncertainty Firm-Year Level . . . 68 2.4 Firm productivity and diversification of sourcing . . . . . . . . . . . . 70 2.5 Resilience of firms to WTO accession shock . . . . . . . . . . . . . . . 71 3.1 Effects of Economic Contraction Shocks. . . . . . . . . . . . . . . . . . . . . 83 A1 Term Sets for Uncertainty Indexes . . . . . . . . . . . . . . . . . . . . . . . . 100 A2 Sample of Country-specific TPU Spike Events (2018-2019) . . . . . . . . . . 101 A3 Summary Statistics of Trade Policy Uncertainty Index (2010-2019) . . . . . . 102 A4 U.S Trade Policy Uncertainty and Other Policy Uncertainty . . . . . . . . . 102 B1 Summary Statistic of Industries . . . . . . . . . . . . . . . . . . . . . . . . . 105 C1 Sample Selection Issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 v C2 Firm’s Exposure to U.S Market . . . . . . . . . . . . . . . . . . . . . . . . . 109 vi List of Figures 1.1 Tariff Increase Announcements during Trade War . . . . . . . . . . . . . . . 13 1.2 Country Level Bilateral TPU Index 2010-2019 Monthly . . . . . . . . . . . . 14 1.3 Effects of Expectation and Uncertainty Shock . . . . . . . . . . . . . . . . . 19 1.4 Effects of Uncertainty Shock . . . . . . . . . . . . . . . . . . . . . . . . . . 19 1.5 Uncertainty Shock and Restructure Option . . . . . . . . . . . . . . . . . . . 26 1.6 Distribution of Entry / Exit Rate . . . . . . . . . . . . . . . . . . . . . . . . 37 1.7 Expectation and Uncertainty Shocks . . . . . . . . . . . . . . . . . . . . . . 41 1.8 Expectation Shock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 1.9 Uncertainty Shock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 1.10 GDP Decrease in Expectation and Uncertainty Shock . . . . . . . . . . . . . 43 3.1 The Effects on Extensive and Intensive Margins . . . . . . . . . . . . . . . . 85 3.2 The Effects on Stay Value and Adjust Value . . . . . . . . . . . . . . . . . . 88 3.3 The Effects on Entry Value and Exit Value . . . . . . . . . . . . . . . . . . . 89 3.4 Number of Exported Products . . . . . . . . . . . . . . . . . . . . . . . . . . 90 3.5 The Effects on Number of Entered and Exited Varieties . . . . . . . . . . . . 91 A1 U.S Trade Policy Uncertainty 1995-2019, Monthly . . . . . . . . . . . . . . . 103 A2 Bilateral Trade Policy Uncertainty 1995-2019, Quarterly . . . . . . . . . . . 104 vii Abstract In this thesis, I investigate the impact of various shocks on global trade using both firm and country-level data from over 100 countries, dating back to 1970. I specifically focus on the effects of these shocks on export and import firms, aiming to understand how supply chains are disrupted and the unique adjustments firms make in response to these disruptions. My analysis reveals that uncertainty shocks lead to delays in the entry and exit probabilities for export firms, while import firms with more diversified supply chains demonstrate increased resilience. Furthermore, my results indicate that downstream firms are disproportionately impacted by global supply chain disruptions. In contrast, domestic economic contraction shocks have adverse effects on both extensive and intensive export margins, predominantly driven by a reduction in the intensive margin. viii Chapter 1 How Do First and Second Moment Trade Policy Shocks Affect Firm Export Activities? This paper studies empirically and theoretically the link between trade policy uncertainty (TPU) and firm export activities. I combine data on country-product level tariff change policyannouncementswithcountry-levelnews-text-basedTPUindicestoseparatelyidentify the effects of expectation and uncertainty shocks of trade policy. Firm-level supply chain data from over 45 countries reveal that after controlling for expected future tariffs: (1) an increase in uncertainty decreases the measure of exporters up to 2 percent over 4 quarters; (2) uncertainty shock reduces the firm exit probability by generating an option cost to re- entry; (3) firms more highly exposed to the U.S market are more resilient to uncertainty shock. The empirical results are interpreted through the lenses of a two-country general equilibrium model with nominal rigidities, firms’ export participation decisions, and first- and second-moment shocks to tariffs. I find that a 1 percent increase in both the expected valueandstandarddeviationoftariffratesreducesGDPby0.36percentinthefirst4periods following the shock; of this, about 47 percent (0.17 percentage points) can be attributed to 1 firm responses to higher tariff rate uncertainty. 1.1 Introduction Newtradedeals,tradewars,andrenegotiationsoftradeagreementshaveincreasinglybecome a focus for investors, firms, and politicians over the last decade. These transactions have resulted in a more uncertain outlook on global trade. The extant literature indicates that trade policy uncertainty (TPU) negatively affects firms’ export activities: a higher TPU will delay entry decisions for potential exporting firms and slow their business investment. However, TPU can adjust export activities by changing expected future tariffs or increasing tariff volatilities. It remains an open question to determine whether these effects are due to the increase in expected future tariff rates and other trade costs or whether they are due to higher uncertainty. How do firms adjust their export activities under uncertainty shocks with heterogenous future expectation value? Will uncertainty shock always negatively affect export activities? Higher future tariff expectations will decrease a firm’s expected profit and households’ ex- pectedwages. Non-exportfirmswilldelaythedecisiontostartexporting(entry),andexport firms will terminate their export relationships. However, uncertainties might offset those de- cisions, especially for exporting firms; they might delay the decision to stop exporting (exit) because possible future tariffs are not expected to increase. Uncertainty shocks with differ- ent future tariff expectations might divert export activities in different directions and affect firms’ investment decisions. Even without changes in tariff expectations, uncertainty can still affect macroeconomic outcomes. These considerations underscore the importance of clarifying how TPUs affect international trade with different future tariff expectations. Thispaperattemptstoseparatelyidentifytheconsequencesofshockstoexpectedvalues and shocks to volatility in trade policies for the extensive and intensive margins of export activitiesandinvestmentsbyexportingfirms. First,Iempiricallycharacterizethemechanics 2 of trade adjustment at the firm level during a recent period of large changes in global trade policy. Theanalysiscombinestext-basedTPUindexesatthebilateralcountrylevel, captur- ingchangesinuncertaintyandallowingtheuncertaintyandfuturetariffexpectationtomove in different directions, with detailed country-product level tariff data that measure changes in expected future tariffs. Second, motivated by the empirical evidence, I use a two-country general equilibrium model with nominal rigidities, heterogeneous firms, and firms’ export decisions to quantify the effects of the different channels by which changes in trade policy affect economic activities and provide a structural interpretation of my empirical results. Finally, I also provide evidence that can be used to inform policy decisions. I consider all trade relationships between the United States and each of its supplier countries between 2015 and 2019 and introduce a bilateral news-text-based TPU index to measure the bilateral uncertainty level. I construct a bilateral text-based TPU index 1 be- tweentheU.S.andeachsuppliertomeasureheterogeneouschangesintradepolicyvolatility. First, the index reliably spikes near each uncertainty event and captures public speculation about future events 2 . Second, this index includes all trade-related events that drive up the bilateral uncertainty level between supply and import countries and do not require changes intarifflevels 3 . Third,thisuncertaintyindexallowsdifferentdirectionsforuncertaintiesand expectations 4 . Finally,byconstructingtheTPUindexatthebilateralcountry-pairlevel,the analysis in this paper captures trade policy uncertainty pertaining to U.S tariff variability directly related to the United States and each of its trading partners, avoiding “spillovers” from other countries 5 . On the contrary, more traditional country-level TPU indices used in 1 The uncertainty index based on newspaper coverage frequency is a widely used measure of uncertainty levels introduced by Baker et al. (2016). 2 For example, the September 2018 TPU index only uses news articles from that month but includes forward-looking discussions of future policy changes. 3 For example, Pierce and Schott (2016) used the gap between expected MFN (most favored nation) tariffs and realized tariffs to measure uncertainty. In that method, the uncertainty level changed as long as the tariff expectation changed. 4 Forexample,thebilateralU.S.-VietNamtradepolicyuncertaintyincreasedafter2016. Thisuncertainty was mainly caused by the uncertainty about whether the U.S. will move its supply chain to Viet Nam or not, which is a positive event. 5 Forexample,BrexitaffectsthetraderelationshipbetweentheUnitedKingdomandtheEuropeanUnion (E.U.) countries, increasing the country-level TPU index for the United Kingdom and each E.U. member; 3 related literature reveal all trade-related policy uncertainty for a specific country but do not measure specific bilateral tariff rate volatility. I use U.S. tariff rate increase news-text data 6 for U.S. trade partners during 2018 and 2019 to help define the expectation shock. Firms and products faced the same uncertainty shock because of the “America First” economic policy 7 , but only some of them experienced expectation shocks. With the U.S. government announcing the imposition of extra duties, firms’expectationsforfuturetariffsincreasedsharply. Thismethodcanhelptocomparethe effects on export firm activities by separating them into whether or not they are face-serve expectation shocks and allowing different directions for the uncertainties and expectations. Furthermore, tariff rate changes (significant increases) announced by the Trump administra- tion as a measure of changes in tariff rate expectations were heterogeneous not only across countries but also across products within countries 8 . Moreover, the administration’s 2018 announcementsonlychangedtheexpectedfuturetariffratesbecausetherealizedtariffrates in my data remained unchanged until 2020. To measure the (intensive margin) export and investment responses of firms to changes in tariff expectations and tariff uncertainty, I use firm-level data from Factset 9 which covers each company’s industry and business segment revenue data, including the distribution of revenues across countries and industries. I merge this Factset data with Compustat 10 data by matching each firm’s North American Industry Classification System (NAICS) code to obtain information on firm fundamentals, including capital stock, cash flow, and long- term debt. My final sample includes 120,000 quarterly, bilateral firm-firm relationships for 45 trade partners of the United States. however, Brexit does not directly affect tariff rate uncertainty between the E.U. member countries and the U.S. Therefore, the bilateral U.S.–European union TPU index may not increase around such events. 6 https://www.card.iastate.edu/china/trade-war-data/ 7 The Trump wants to reduce the United States trade deficit by shifting American trade policy from multilateral free trade agreements to bilateral trade deals. 8 For example, the Trump government announced imposing an extra 25% tariff on tubes from China. Then, the expected future tariffs for tubes from China changed, but those from other countries remained the same. However, the expected future tariffs for other products from China remained the same. 9 https://go.factset.com/marketplace/catalog/product/factset-supply-chain-relationships 10 https://wrds-www.wharton.upenn.edu/pages/get-data/compustat-capital-iq-standard-poors/ 4 My results can be summarized as follows: (1) Even in the absence of new tariff rate an- nouncements, increased uncertainty is likely to result in adverse macroeconomic outcomes; (2) an uncertainty shock delays entry and investment, but reduces the probability of exit, increasing the probability that a firm will stay in the market; (3) firms less exposed to the U.S market are more sensitive to uncertainty shocks and restructure their supply chains. My regression models imply that the mass of exporters decreased by 0.4 percent with a one standard deviation tariff rate expectation shock during the first 4 quarters following the shock, but dropped by 2 percent with a one standard deviation uncertainty shock. The adjustment in export participation for firms confronting no tariff rate expectation shock explains more than 45 percentage points of extensive margin movement. The uncertainty shock significantly affects exporter activity after controlling for an increase in the expected future tariff rate; higher uncertainty with a higher expected future tariff rate delays entry and business investment, but reduces the probability of exit and increases that of staying in themarketforexportingfirms. Thelatterresults,Iconjecture, ariseduetoanoptioncostof entry and exit with a higher probability of lower future tariffs, which offsets negative effects of higher future tariff expectations. In fact, firms confronting higher expected future tariff rates decrease entry when TPU rises by one standard deviation by 0.98 percent more than firms confronting no expected future tariff rate shocks which decrease entry by 1.01 percent. Finally, with an increase in trade policy uncertainty, firms can either exit the export market orrestructuretheirexportchaintoothermarkets. Ialsofindthatfirmsrelyingmoreheavily on the U.S market are more resilient to trade policy uncertainty shocks; for firms with more than75percentofrevenuederivingfromtheU.Smarket,themeasureofexportrelationships and the probability of establishing a new export relationship with the U.S does not change much. However, those who rely more on the domestic market shrink the measure of their export relationships and exit export markets at a higher frequency, due to smaller upfront investment. I find the estimated reductions in foreign supply chain relationships due to higher trade 5 policy uncertainty and changes in other, intensive margin export activities are robust to numerous changes in sample and specification; in particular, my empirical findings are not driven by China-U.S. supply chain relationships or specific subsets of firms. Next, I follow Alessandria and Choi (2007) and Caldara et al. (2020) use a two-country general equilibrium model with nominal rigidities and firms’ export decisions to trace the channels by which changes in trade policy uncertainty affect economic activity. Unlike Cal- dara et al. (2020) uses the U.S firm-level data to measure the effects of TPU shocks on importer countries, I use the parameters calibrated from the export countries and focus on shocks on export countries only. There are three experiments in this model. In the bench- mark experiment, this model considers a surprise increase in both expected future tariffs and uncertainty about future tariffs. In the second experiment, I model trade tensions as a downside risk and study the effect of an increase in the probability of higher tariffs in the future. In the third experiment, I model trade tensions as a mean-preserving increase in the variance of tariffs. This model is solved using the third-order perturbation method (Fern´ andez-Villaverde et al. (2015)). In numerical simulations of a carefully calibrated version of this model, both tariff ex- pectation (first-moment shocks) and increased tariff uncertainty (second-moment shocks) reduce export participation and investment and are broadly in line with my empirical find- ings. I find an average 0.36 percent GDP decrease for U.S supplier countries during the period 2016-2019 during the first four quarters following a simultaneous 1 percent increase in the expected tariff rate and a 1 percent increase in uncertainty. Of this 0.36 percent decrease in GDP, my model accounts for over 47 percent via increased uncertainty. Higher expected future tariffs reduce the expected value of exports, lowering firms’ expected profits and household wages and labor supply, and reduce investment and consumption. Anticipat- ing higher marginal costs in the future, wholesale firms increase their markups which serves as a tax on labor, aggravating the decline in hours worked and consumption. In addition, higher expected future tariffs lower the benefit to firms of exporting by shrinking the ex- 6 pected size of, and their comparative advantage in, export markets - hence reducing the expected future gain from participation in the export market. Consequently, exports, the mass of exporters, investment, consumption, and output all decline. In the model used in this paper, higher uncertainty about future tariffs leads to higher variance of future profit- maximizing prices of firms. For non-exporting firms, this higher uncertainty reduces the incentive to invest and build new trade relationships that require upfrontinvestment. However, withnochangeinexpectedfuturetariffrates, delaysinestab- lishing new export relationships only lead to increased waiting costs. Only more vulnerable, smallerfirms“waitandsee”. Ontheotherhand,exportingfirmsweighthetrade-offbetween potential profit lost by exiting the market, with (sunk) re-entry costs, and potentially pric- ing below the period-by-period profit maximizing level by remaining in the export market whilepayingtheperiod-by-periodfixedcostofexportparticipation. Forthesefirms, waiting may result in lower penalties following an increase in tariff uncertainty. In addition, when adjustingpricesiscostly, wholesalingfirmsrespondtohighertariffuncertaintybyincreasing markups to avoid selling at relatively low prices. Higher markups reduce foreign importer demand, and hence domestic hours worked, output, consumption, and investment. Even if higher trade policy uncertainty is associated with unchanged or higher export participation in the short term, it thus nonetheless results in a large decline in investment and output. If firms have complete information regarding the future tariff path, they would know exactly when to either invest and start/continue to export, or to stop paying the per period fixed costofexportingandlosetheirexporterstatus. Inthecaseoftradepolicyuncertainty,when firms don’t know for certain the future tariff path, a potentially valuable waiting period re- sults. The negative real effects of uncertainty shocks are then felt later than under a tariff expectation shock. This paper provides evidence that can be used to inform policy decisions. Even without actualincreasesintradecostsandrestrictionsoninternationalcommerce,policyuncertainty related to trade can worsen economic activity as firms pause hiring and investment, raise 7 prices,andpostponeentryintoexportmarkets. Myresultssuggestthatpolicymakersshould act to avoid the adverse macroeconomic effects of policy uncertainty. More generally, rolling back damaging trade restrictions and reducing uncertainty via clear communication of pol- icy objectives should be a priority. Complementing regional agreements with multilateral reformsandrestoringafullyfunctionalWorldTradeOrganizationdisputesettlementsystem can mitigate discriminatory policies’ potential negative impacts on trade partners and help resolve international tensions. The contributions could be summarized in three aspects. First, this paper contributes to the literature on trade policy uncertainty by offering a new approach to studying the heterogenous effects of higher future tariffs expectation and higher uncertainty levels. There isincreasingevidencethatreductionsinTPUincreasetradeandexportactivities. Brogaard and Detzel (2015) finds that policy uncertainty reduces asset returns, Handley (2014) and Handley and Lim˜ ao (2015) find that uncertainty related to trade policy delays firm entry, and the firm’s entry decision will not be affected by “good news” because of the waiting cost. Baker et al. (2016), Julio and Yook (2016), Gulen and Ion (2016)and Caldara et al. (2020) find negative responses from corporate investment to newspaper-based indexes of policy uncertainty. Pierce and Schott (2016) documents that China’s entry into the WTO contributed to a sharp drop in U.S. manufacturing employment. Steinberg (2019) studies the implications of Brexit for the U.K.’s economy. Those papers find that beliefs about future policy changes affect the response of export decisions to foreign destinations because of sunk costs. Trade agreements play a role in shaping such beliefs. This paper exploits heterogeneity across firms in their exposure to tariff increases which provides a method to investigate the effects of first-moment and second-moment shock. Second, this paper also relates to the research on the global supply chain and moves its focus from the importer’s sourcing decision to the exporter’s export decisions. Antr` as et al. (2017) note that sourcing involves a portfolio of inputs with potential interactions between suppliers. Trade liberalization is a major contributor to the growth in vertical 8 specialization(Hanson(2005),JohnsonandNoguera(2017),withimportantimplicationsfor productivity and welfare (Amiti and Konings (2007), Goldberg et al. (2010), Halpern et al. (2015)). This paper shows that export decisions are also important. Exporter firms will shrink their export participation in a highly uncertain environment, increasing the import price. Althoughhighercompetitortariffswillbenefitexportfirms,theuncertaintiescanspill over across supplier countries. Third, this paper contributes to many recent studies that use text-based methods to quantify economic uncertainty and related concepts. Baker et al. (2016) use their index to show that increases in policy uncertainty are associated with upticks in stock price volatility and reductions in industrial production and employment. Caldara et al. (2020) and Julio and Yook (2016) find that election cycles impact foreign direct investment. Baker et al. (2016) find that the responsiveness of firms to any given policy stimulus may be much lower in periods of high uncertainty. This paper is the first to measure country-level bilateral TPU index across over 50 countries which allows studying the heterogeneity effect of self uncertainties and competitor’s uncertainties. Theremainderofthepaperproceedsasfollows. Section2describesthedataandvariable construction. Section 3 describes the empirical effects of trade policy uncertainty. Section 4 contains the model and results. Section 5 presents the policy implications and conclusions. 1.2 Data Thissectionpresentsthemeasurementoftariffexpectations,tradepolicyuncertainty(TPU), and firm-level activity and supply chain data. The full sample covers 45 major U.S foreign suppliers from 2010 to 2019. Export firms are defined as firms having at least one export relationship with a foreign customer. 9 1.2.1 Trade Policy Uncertainty Tariff Expectation First, I use the government announced tariff increase data to measure the changes in future tariff expectations. The year 2018 witnessed the largest trade war in decades. In January 2018, President Trump imposed tariffs on solar panels and washing machines of 30 to 50 percent. In March 2018, he imposed tariffs on steel (25 percent) and aluminum (10 percent) from most countries, which, according to Morgan Stanley, covered an estimated 4.1 percent oftotalU.Simportvalue, andinJune2018, thesetariffswereextendedtoapplytotheE.U., Canada, and Mexico. In the meantime, the United States imposed a 25 percent additional duty on $50 billion worth of Chinese imports related to China’s “Made in China 2025” 11 in-duty policy. In September 2018, the United States raised tariffs by 10 percent on $200 billion worth of products from China and increased this to 25 percent in early 2019, leading to a trade war. In May 2019, President Trump unilaterally announced his intention to impose a 5 percent tariff on all imports from Mexico beginning on June 10th 2019, with tariffs increasing to 10 percent on July 1st and by another 5 percent each month for three successive months, and added illegal immigration restrictions as a condition for U.S-Mexico tariff negotiations. The move was seen as threatening the ratification of the United States- Mexico-CanadaAgreement(USMCA),andtheNorthAmericantradedealsettoreplacethe North American Free Trade Agreement (NAFTA). With a series of unexpected trade policy announcements, the uncertainty level between theU.Sandindividualsuppliercountriesspiked. However, thechangesintariffexpectations are heterogeneous across countries and products. For example, after the U.S government announced imposing extra 25% duty on tubes from China, the expected tariffs for tubes from China increased 25%. Still, China’s expected tariffs for other products ( not on the announcement list) didn’t change. Figure 1.1 represents graphically announced tariff rate 11 The Chinese government has launched “Made in China 2025,” a state-led industrial policy that seeks to make China dominant in global high-tech manufacturing. 10 increasesbytheindustryfortheyears2017-2019. Expectedtariffrateincreasesarecollected from government announcements and aggregated to the industry level, using Global Trade Analysis Project (GTAP) industry codes 12 . China suffered the largest increases in expected tariffs. However,othercountries,forexample,thoseintheEuropeanUnion,alsoexperienced expected tariff rate increases vis the U.S for some industries. The second panel of figure 1.1 indicates that, even though President Trump announced an increase in tariffs worldwide which elevated the global uncertainty level, most tariffs were unchanged until 2020. Uncertainty The measures of TPU is constructed stem from the work of Baker et al. (2016). Here, specifically, a TPU index value is a weighted average of the number of newspaper articles discussing trade policy uncertainties. I first automated text searches - starting in 1995 - of the electronic archives of seven newspapers: The Boston Globe, the Chicago Tribune, the Los Angeles Times, The New York Times, the Wall Street Journal, and The Washington Post. I obtained monthly or quarterly counts of articles that contain at least one “term” in each of the target (word/term) sets for each U.S. supplier country. Table A1 reports the terms in each set for different measurements of the TPU index. I collect this TPU index data for 7 major U.S supplier countries: Australia, Canada, China, India, Japan, Mexico, and the United Kingdom, and for 3 supplier regions: the European Union, Latin-American (excludingMexico),andSoutheastAsia 13 . Thereareatotalof45 14 countriesinthissample. In the second step, I scale the raw monthly or quarterly TPU counts by the number of total 12 The reason I aggregate the product level tariff data to the industry level tariff data is that, with the absence of detailed information on the products traded between firms, I use the industry sector to express the heterogenous tariff expectation across firms. 13 Top 10 U.S total Import Share in 2015: China 21%, the EU 16%, Canada 13%, Mexico 12%, the Southeast Asia 6%, Japan 5.8%, the Latin American (excluding Mexico) 3.8%, the UK 2.5%, India 2%, Australia 1.8% 14 The European Union includes all 27 countries in the EU. Southeast Asia includes Indonesia, Malaysia, the Philippines, Thailand, and Vietnam. Latin America includes Argentina, Brazil, Chile, Colombia, Peru, and Venezuela. Brunei, Cambodia, Laos, Myanmar, Bolivia, Ecuador, Paraguay, Panama, Haiti and Belize, were dropped because of the limited sample available of firm supply chain data. There are 45 countries in the final sample. 11 articlesforthesamenewspaperandperiod. Inthethirdstep, InormalizedtheTPUindexes by subtracting the mean and dividing by the standard deviation, to obtain a “z-score” of TPU for each country to make the TPU index levels comparable across countries. Fourth, I compute the simple average of the standardized series over newspapers by period. In the finalstep, Inormalizeeachperiod’sindexvaluetoanaverageof1. Asimilarmethodologyis appliedtomeasureeachcountry’scompetitors’tradepolicyuncertainty,whichistheaverage of all competitors’ TPU (ROW TPU). A higher competitors’ TPU indicates a relatively low “own” TPU. Figure 1.2 presents the monthly “own” TPU and competitor’s TPU for 10 major economies and regions from 2010 to 2019 15 . Table A2 presents a summary of a selected sample of country-specific TPU “spike” events. For most economies, the bilateral TPU in- dexincreases neartight presidential elections andspikes nearthe beginningof theUS-China trade war in 2018. China was the most vulnerable country to higher TPU, during the trade war. The TPU also increased significantly in Canada, Mexico, and Europe, driven by the re-negotiation of NAFTA and other tariffs. Table A3 also shows the summary statistics of the TPU measures. All countries have excess kurtosis and are positively skewed. There are a few advantages to using this bilateral TPU measure. First, the news-based TPU index contains news articles that express forward-looking concerns about trade pol- icy uncertainty in the future. Second, the news-based TPU index is better for capturing volatility than expectation. Unlike the official government announcements, news articles do not necessarily contain detailed product level information about future tariff levels but do reflecthighertariffvolatility. Forexample,uncertaintyabouttradecostsbetweentheUnited States and Mexico increased significantly right after the NAFTA re-negotiation announce- ment; however, the consensus during the sample period was that the new trade agreement would remain tariff-free for most goods. Finally, country-level bilateral TPU can capture trade policy uncertainty levels directly related to the United States and a specific trade 15 Quarterly TPU Index is presented in Figure A2 12 Figure 1.1: Tariff Increase Announcements during Trade War The dots represent the GTAP industries. The rest of the world (ROW) includes all U.S supplier countries except China. Overlapping dots were omitted for viewing ease. The left graph is the announced tariff increase level in 2018-2019. The right graph compares tariffs in 2016 (pre-trade war) and 2020 (post-trade war). partner. The measurement lets us avoid uncertainty spillover from competitors. 1.2.2 Supply Chain Data Dataonfirm-levelsupplier-customerrelationshipsiscollectedfromFactSet,afinancialinfor- mation and technology provider for investment professionals 16 . The database tracks supply chain relationships for over 23,000 firms globally. Each relationship is identified by the enti- ties involved with details about the customer, supplier, partner/distributor, or competitor, and the dates of inception, and termination if the relationship is no longer active. The FactSetdataoffersignificantlybettersuppliercoveragethanothercomparabledatasets. For example, for 2012, there are 43,068 supplier-customer relationships identified by FactSet compared to only 6,351 in the Compustat segment files. At the firm level, FactSet includes more firms with supplier information and identifies about four times as many suppliers per firm than Compustat. The Compustat data is limited to major US-listed firms’ customers andunder-reportsthoseofnon-USlistedfirms. Ontheotherhand,58percentoftheFactSet 16 Factset dataset relies on firms self-reporting. 13 Figure 1.2: Country Level Bilateral TPU Index 2010-2019 Monthly The European Union includes all 27 countries in the EU. Southeast Asia includes Indonesia, Malaysia, the Philippines, Thailand, and Vietnam. Latin America includes Argentina, Brazil, Chile, Colombia, Peru, and Venezuela. Brunei, Cambodia, Laos, Myanmar, Bolivia, Ecuador, Paraguay, Panama, Haiti and Belize, were dropped because of the limited sample available of firm supply chain data. There are 45 countries in the final sample. 14 data are not listed in the U.S. These results indicate that FactSet has a broader coverage of supply chain relationships on multiple dimensions compared to Compustat. The data denotes the start and end date of supplier relationship. I count the total days of a linkage’s existence in one quarter and then normalize it by the total number of calendar days in that quarter, resulting in a relationship-specific measure ranging between 0 and 1. This variable simultaneously measures whether a relationship is built and if it is, its duration. I use an inverse hyperbolic sine to measure the relationships, which is similar to the log function except that it accounts for zeroes. I define the firm as an “Exporter” if it acts as a “Supplier” with another firm or if the other firm is a “Customer” to it. Since somefirmsonlyreporttheir“Supplier”relationshipsandsomeonlyreporttheir“Customer” relationships, I combine “Supplier” data and “Customer” data to obtain my full sample. I alsodefinetheentryorexitasequalto1ifthefirmdecidestostartanewexportrelationship or terminate an existing one. This data also contains the business segment revenue data for each company over a given period, including the distribution of their revenues across all countries and industries. Iusetherevenuesharetomeasurethefirm’sexposuretodomestic,foreign,andU.Smarkets. For the measure of investment, I combine Factset data with Compustat data to obtain the firm’s fundamental activity data. I define a firm’s capital stock as net property, plant, and equipment (PPENTQ). I use the ratio of cash and short-term investments to beginning-of- period property, plant, and equipment as a measure of cash flow. I also map each firm’s North American Industry Classification System (NAICS) code to its GTAP industry code and classify firms into 63 industries 17 . Table B1 shows a summary of these classifications. The final dataset contains firm-level supply chain data from 2015 to 2019. All rela- tionships that are not classified as suppliers and customers, as well as those that involve governmentprocurement(i.e., havingany federal, state, or localgovernmentas acustomer), were removed to focus on firm-to-firm business relationships. Table 1.1 shows the summary 17 To match the country-industry level tariffs data, I use GTAP industry codes to express each firm’s industry. 15 Table 1.1: Firm Supply Chain Data Summary Statistics (2015-2019 Firm-Quarter Panel) Export Firms Export to US Firms mean sd mean sd number of trade relationships 7.78 15.54 11.08 19.20 number of export relationships 4.70 11.35 7.68 14.75 number of domestic trade relationships 3.08 7.44 3.40 8.09 number of export countreis 5.35 7.36 8.81 9.04 foreign revenue / total revenue % 36.55 36.27 51.59 34.45 U.S revenue / total revenue % 8.88 16.76 15.25 20.76 Revenue (US$) 245953.40 2704651.98 226298.24 2553800.61 PPENTQ / Total Assets 0.26 0.20 0.23 0.18 Long Term Debt / Total Assets 0.15 0.24 0.15 0.27 Export Firms Export to US Firms #. Observations % of Total Sample #.Observations %of Total Sample Europe 42404 23.33 25388 28.20 China 38980 21.45 17076 18.97 Japan 29408 16.18 12464 13.84 Southeast Asia 16992 9.35 4456 4.95 United Kingdom 15948 8.77 10296 11.44 India 11460 6.31 6400 7.11 Canada 9748 5.36 6628 7.36 Australia 8416 4.63 3840 4.27 Latin American 6708 3.69 2432 2.70 Mexico 1680 0.92 1048 1.16 PPENTQ is net property, plant, and equipment The European Union includes all 27 countries in the EU. Southeast Asia includes, Indonesia, Malaysia, the Philippines, Thailand, and Vietnam. Latin America includes Argentina, Brazil, Chile, Colombia, Peru, and Venezuela. dropped Brunei, Cambodia, Laos, Myanmar, Bolivia, Ecuador, Paraguay, Panama, Haiti and Belize, due to the limited firm supply chain data. This left 45 countries in the final sample. statisticsforthesupplychaindata, includingthemeanandstandarddeviationforallexport firms and firms with at least one export relationship with the U.S from 10 economies on a quarterly basis, using the period of 2015 to 2019. 1.3 The Effects of Trade Policy Uncertainty In this section, I examine the empirical effects of TPU export activities by comparing firms’ responses to both expectation shocks and uncertainty shocks. I first investigate the dynamic effects on the extensive margin of the two different shocks by estimating a local projection (LP) model. Then I document reduced-form relationships between the two different shocks andexportentryorexitactivitytocharacterizethemechanicsofexportadjustment. Finally, to address concerns that China drives our empirical results and the recent US-China trade 16 warspecifically,Iuseasub-samplewhichexcludesChinesesupplierstoshowthattheresults are not driven by just production relationships between American and Chinese firms. To investigate the effects of TPU between the United States and each country, I use the total number of firm export relationships with the U.S to express the measure of exporters and the newly established/terminated export relationships to express the entrance/exits. The measure of exporters, entrance, and exits are measured by the inverse hyperbolic sine of the number of export to the U.S relationships. The estimated coefficient with the inverse hyperbolic sine transformation can be interpreted as a percentage change (Bellemare and Wichman(2020)) 18 . Thismeasurehasbeencommonlyusedinthestatisticsandeconometric literature as an alternative to the log transformation, as it preserves observations with a value of zero. Moreover, I use the log difference of the firm’s capital stock to measure their business investment. I measure capital as net property, plant, and equipment (PPENTQ). This approach provides a stable estimate of quarterly capital growth over the sample. Iusetheexpectedtariffincreaselevel(∆ Exp.Tariff)andtheuncertaintyindex(TPU) to measure the expectation shock and uncertainty shock, respectively. The higher the ∆ Exp.Tariff,themoreseverethelevel’sshockis. Beforethetariffincreaseannouncements, the∆ Exp.Tariff f,t equalszero; aftertheannouncements, itchangestotheannouncedratio until the policy is abolished or changed to a higher ratio. The competitor’s future tariff ex- pectation ROW∆ Exp.Tariff is the average of all competitor’s expectations. Competitors in this context are firms in a foreign country that export products in the same sector to the U.S: Row∆ Exp.Tariff industry i,k,t = P ∆ Exp.Tariff industry i ′ ̸=i,k ′ ̸=k,t P 1 firm industry i ′ ̸=i,k ′ ̸=k,t (1.1) where i indexes a firm, k indexes an export country, and t indexes the year-quarter time. Higher TPU indicates higher volatility. The competitor’s TPU index ROW TPU is basedonasimilartext-basedarticlesearch. Irananautomatictextsearchandfoundarticles that talked about trade policy uncertainty between the U.S and at least one competitor 18 arcsinh(x) =ln(x+ √ x 2 +1) 17 country. Table A1 shows the keyword terms. Finally, Irestrictthesampletothreedimensions. First, Iexaminethe2015Q1-2019Q4 period (20 quarters). As discussed in Section 2, in the years leading up to 2015, there was little change in TPU. Second, I include firms in the trade sectors of agriculture, mining, and manufacturing, leaving out the wholesale and service sectors. Finally, I focus on firms that have had at least one export relationship with the U.S since 2015. Table 1.2 lists the dependent and independent variables used in the analysis, as well as their definitions. Table 1.2: Definitions of Variables Variable Definitions Dependent Variable Mass. Exporter Inverse hyperbolic sine transformation of the number of export to the U.S relationships Mass. Entry Inverse hyperbolic sine transformation of the number of newly established export to the U.S relationships Mass. Exit Inverse hyperbolic sine transformation of the number of terminated export to the U.S relationships Investment log(Capital t )− log(Capital t− 1 ) Shocks TPU i,t Text-based bilateral TPU index between country i and U.S at time t. Row TPU i,t Text-based TPU index between country i’s competitors and U.S at time t. ∆ Exp.Tariff f,t Expected tariff increased ratio for firm f at time t. Row ∆ Exp.Tariff f,t Simple average of expected tariff increased ratio for foreign firms in the same industry with firm f at time t. 1.3.1 Effects of Trade Policy Uncertainty on Extensive Margin I start by estimating the dynamic effects of changes in shocks on the extensive margin of exports. My strategy is to regress export activities at various horizons against the contem- poraneous value of shocks. I use the local projection method. This procedure does not constrain the shape of the impulse response functions and is, therefore, less sensitive to misspecification than estimates of VAR models (Jord` a and Taylor (2016)). The benchmark specification at quarterly frequency is as follows: Y i,k,t+h − Y i,k,t− 1 =α h i +α h k +α h t +α h i,season +β h Shock+Γ X i,t− 1 +ϵ i,k,t+h (1.2) 18 Y i,k,t denotes the variable of interest: the inverse hyperbolic sine of the number of export relationships. α i , α k , and α t denote firm, country and year-quarter time fixed effects, re- spectively. Some firms might adjust their export activities across different season, so I also use country firm-season fixed effects α i,season to control for firm seasonalities. h=0,1,2,3,4 denotes the time horizons considered. Shock is either expectation shock ∆ Exp.Tariff i,t or uncertainty shock TPU k,t . X i,t− 1 denotes firm-level control variable: firm’s total revenue, exposure to U.S market, pre-shock’s tariff level, one lag of the growth rate of Y i,k,t , as well as one lag of the shock. Figure 1.3: Effects of Expectation and Uncertainty Shock The shaded area represents the 90 percent confidence interval for Discoll-Kraay standard errors (z-value = 1.64). Left: the effects of expectation shock. Right: the effects of uncertainty shock. Figure 1.4: Effects of Uncertainty Shock The shaded area represents the 90 percent confidence interval for Discoll-Kraay standard errors (z-value = 1.64). Left: the effects of uncertainty shock with no tariff expectation changes. Right: the effects of uncertainty shock with tariff expectation increased. 19 Figure3.12showstheresultsforlargechangesintheTPUindexandtariffexpectations on the extensive margin of exporters. The changes in the tariff expectations slightly shrink the export relationships by 0.5% after 4 quarters (see Left Panel 1 Figure 3.12). An increase in TPU has more significant negative effects on the mass of export relationships with the U.S, reaching 2 percent after 4 quarters (see Right Panel in Figure 3.12). The local projec- tion recovers the result from uncertainty literature that uncertainty shocks lead to overall contractions of exports. Next, I examine whether the baseline results change in response to changes in expectations. One of the main advantages of the local projection method is its flexibility in dealing with non-linearities and state dependency. The state-dependent speci- fication would take the following form, with S i,t being an indicator variable taking the value of 1 if the tariff expectation changed and 0 otherwise. Y i,k,t+h − Y i,k,t− 1 =S i,t α h i +α h k +α h t +α h i,season +β h Shock+Γ X i,t− 1 +(1− S i,t ) ¯α h i + ¯α h k + ¯α h t + ¯α h i,season + ¯ β h Shock+ ¯Γ X i,t− 1 +ϵ i,k,t+h (1.3) Figure 1.4 shows that, even without changes in expectations of tariffs, the extensive margin drops by 1.5% in 2 periods and 1% in 4 periods. The decrease of exporter margins with no changes in expectations of tariffs explained over 45 percent movement in the mea- sure of export relationships. The uncertainty shock continuously decreases the measure of exporters, no matter whether the tariff expectation changes or not. 1.3.2 Export Adjustments Entry, Exit, and Investment Regardingthedifferentresponsestotariffexpectationsandvolatility,itisimportanttostudy how the number of exporters changes is adjusted by dropping the vulnerable export chain or delaying entry decisions. To test whether exporter firms restructure their export chains 20 through the adjustments in entrance or exit and to investigate the effects of expectation shock and volatility shock, I estimate the following equation: Y i,k,t =α i +α k +α t +α i,season +βTPU k,t +γTPU k,t × ∆ Exp.Tariff i,t +δ ∆ Exp.Tariff i,t +Γ X i,t− 1 +ϵ i,k,t (1.4) The variables are defined as in equation (1.2). Y i,k,t denotes the variable of interest: the inversehyperbolicsineofthenumberofexportrelationships,newlyestablishedrelationships, terminated relationships, and business investments. I also controlled total assets, capital share, and long-term debt when considering the investment. The coefficient of interest is β and γ , representing the relation between TPU k,t and export variables based on whether there was an expectation shock ∆ Exp.Tariff i,t . Table 1.3 presents the results of studying the expectation shock and volatility shock usingafirm-quarterpanelandthespecificationfromtheequation(1.4). Table1.3showsthat, onaverage,TPUisassociatedwithadecreaseintheexportfactors,includingthemeasureof exporters, entrance, exit, and investments. The tariff expectations do not significantly affect exportactivitiesbutincreasetheexitprobability. However,Columns(2),(4),(6),and(8)show a stark difference in statistical relations between firms with shocks in tariff expectations versus those with no changes in tariff levels. Column(4) shows firms with higher future tariff expectations decrease the entry when TPU rises by one standard deviation by 0.985% comparedtothosewithnotariffshocks, whichdecreasedby1.018%. Moreover, Columns(6) and (8) show that without tariff shocks, exporter firms are more likely to delay investments. Additionally, they’re more likely to remain in the market compared to when there is a tariff shock. These findings help to shed light on how shocks from trade policy uncertainty may affect a country’s export activities. My analysis indicates: (1) after controlling for future tariff expectations, higher future tariff uncertainties still delay firms’ entry into a market as 21 well as their investment activities; (2) higher tariff expectations will increase the probability of exit, but uncertainties lead to delays in exit; (3) uncertainties offset the negative effects of higher future tariffs. The export adjustments are mainly through the changes in uncertainties instead of the tariff expectations. Reducing uncertainty leads to more market entry . My data shows that as uncertainty rises, firms on the margin become more cautious. Non- exporter firms might suspend their entrance regardless of whether a tariff increased or not, which suggests that the uncertainty in future trade policy may be enough to induce firms to adjust their export chain. Exportingfirms, ontheotherhand, experiencelesspenaltieswhentheywaitforgood news. Highertariffsdecreasethevalueofexportsandgeneratenegativeprofits, whichdrives exporting firms to exit the market. However, uncertainties increase the probability of lower tariffs and induce an option value of waiting. After paying sunk costs, it is less costly for exporting firms to remain in the market. In response to higher future tariff expectations, uncertaintiesgenerateanoptionvaluetolowertariffsandoffsetthenegativeeffectsonexport activities. For example, the announcement of section 301 tariffs on China 19 increased tariff expectations, but the highly uncertain about that policy reduced the arrival rate. In addition, I conduct a similar analysis on export activity adjustments using the com- petitor’s trade policy uncertainty data. An expectation of higher tariffs on competitors implies a positive tariff shock which should lead to more competitive export prices for the home country. Higher competitors’ TPU drives importers to switch their supply chain to countries with lower uncertainty . Given this, I use expectation and uncertainty shocks in competitor countries to test whether firms expand export activities with positive shocks. Table1.4showsthathigherexpectedcompetitortariffspositivelyaffectexportactivities. Specifically, when companies expect their competitors to experience higher tariffs, there is an increase in the number of exporters in the market, a decrease in the probability that the firms in the market will exit and an increase in overall business investment. However, 19 Section 301 determined a list of tariffs added on China’s products. 22 Table1.3: TheEffectsofExpectationandUncertaintyShockonEntry,Exit,andInvestment Exporter Entry Exit Investment (1) (2) (3) (4) (5) (6) (7) (8) TPU -0.277*** -0.152* -0.869*** -1.018*** -1.157*** -0.756*** -0.0214 -0.111* (-3.20) (-1.76) (-6.92) (-6.77) (-11.11) (-6.05) (-0.38) (-1.75) TPU× ∆ Exp.Tariff -0.0292** 0.0335* -0.0965*** 0.0219** (-2.16) (1.70) (-5.92) (2.42) ∆ Exp.Tariff -0.147 -0.326 1.521*** -0.127 (-0.71) (-1.49) (6.13) (-1.00) r2 0.931 0.931 0.509 0.509 0.518 0.519 0.520 0.521 N 109756 109756 109756 109756 109756 109756 61156 61156 Firm FE √ √ √ √ √ √ √ √ Country FE √ √ √ √ √ √ √ √ Year-Quarter FE √ √ √ √ √ √ √ √ Industry FE √ √ √ √ √ √ √ √ Firm-Season √ √ √ √ √ √ √ √ This table shows the main results using the firm-quarter dataset from 2015Q1 to 2019Q4. Columns (1,2) study the measure of exporters represented as a percentage (inverse hyperbolic sine). Columns (3,4) and (5,6) study the number of established and terminated export relationships with the U.S. as a percentage (inverse hyperbolic sine) Columns (7,8) study the business investment (log difference). All regressions control for a firm’s total revenue, exposure to U.S market, and pre-shock’s tariffs. Controlled total assets, capital share, and long-term debt when considering the investment. All regressions include firm, year-quarter, GTAP industry, and firm-by-season fixed effects. The investment is defined as the log capital difference between two periods, so the observations are different. t statistics in parentheses * p< 0.1, ** p< 0.05, *** p< 0.01 23 Table1.4: TheEffectsofExpectationandUncertaintyShockonEntry,Exit,andInvestment (Competitor’s Trade Policy Uncertainty) Exporter Entry Exit Investment (1) (2) (3) (4) (5) (6) (7) (8) Row TPU 0.0902 0.494 1.041** 0.875* -3.226*** -2.569*** -3.375*** -2.667*** (0.30) (1.51) (2.41) (1.84) (-9.01) (-6.51) (-18.34) (-13.16) Row TPU× Row∆ Exp.Tariff -0.187** 0.105 0.413*** -0.480*** (-2.28) (0.88) (4.17) (-8.62) Row ∆ Exp.Tariff 1.245*** -0.477 -1.787*** 2.212*** (3.13) (-0.83) (-3.73) (8.80) r2 0.931 0.931 0.508 0.508 0.518 0.518 0.525 0.527 N 109756 109756 109756 109756 109756 109756 61156 61156 Firm FE √ √ √ √ √ √ √ √ Country FE √ √ √ √ √ √ √ √ Year-Quarter FE √ √ √ √ √ √ √ √ Industry FE √ √ √ √ √ √ √ √ Firm-Season √ √ √ √ √ √ √ √ This table shows the main results using the firm-quarter dataset from 2015Q1 to 2019Q4. Columns (1,2) study the measure of exporters represented as a percentage (inverse hyperbolic sine). Columns (3,4) and (5,6) study the number of established and terminated export relationships with the U.S. as a percentage (inverse hyperbolic sine) Columns (7,8) study the business investment (log difference). All regressions control for a firm’s total revenue, exposure to U.S market, and pre-shock’s tariffs. Controlled total assets, capital share, and long-term debt when considering the investment. All regressions include firm, year-quarter, GTAP industry, and firm-by-season fixed effects. The investment is defined as the log capital difference between two periods, so the observations are different. t statistics in parentheses * p< 0.1, ** p< 0.05, *** p< 0.01 investment is still risk-averse and decreases as competitor TPU rises. The positive effects from relatively lower tariff expectations will be offset by the high uncertainties, which is consistent with the results in Table 1.3. Restructure In this section, I discuss if firms choose to adjust export activities by restructuring their export chains. Whether or not firms switch to domestic or other foreign markets in response to higher trade policy uncertainty is ambiguous. Firms can choose to expand their export chain to other markets to hedge the loss in the U.S market. However, it might be too costly toestablishnewexportrelationshipsintheshortterm, andthelossintheU.Smarketmight pushfirmstoshrinktheirbusinessallovertheworld. Totestwhetherfirmsrestructuretheir 24 export chain under trade policy uncertainties, I regress the following equation: Y i,k,t =α i +α k +α t +α i,season +βTPU k,t +γ ∆ Exp.Tariff i,t +Γ X i,t− 1 +ϵ i,k,t (1.5) Figure1.5presentstheresultsofmyanalysisofforeignexportdiversificationdecisionscaused by their exposure to the U.S market. The U.S exposure is defined as a company’s revenue share from the U.S market: Share= Revenue from Sale to U.S Total Revenue (1.6) After controlling for tariff expectations, I find that firms will shrink their export activi- ties as trade policy uncertainty rises. However, a firm’s restructuring decisions are heteroge- neousacrosstheirexposuretotheU.Smarket. FirmsthatarelessexposedtotheU.Smarket are more sensitive to uncertainty shocks. They will delay their entry decision and exit the market quickly as uncertainty about future tariffs rises. However, domestic selling activities remain relatively unchanged regardless of whether or not tariffs rise. This is due to the fact that these firms have a small market share in the U.S, and it is too costly for them to wait. On the other hand, firms that are exposed mainly to the U.S market are more resilient to uncertainty shocks. The extensive margin of export to the U.S and the probability of entry do not notably change, and the probability of delaying their exit is increased. The more firms rely on the market, the less risk averse they become, and the higher the opportunity cost isfor re-entry. Exporting firms that rely heavily on the U.S market are much less likely to pause and re-enter. 25 Figure 1.5: Uncertainty Shock and Restructure Option Upper left: the export activities with the U.S. Bottom left: the export activities with the domestic country. Upper right: export activities with non-U.S foreign countries. The Baseline contains all observations. 1.3.3 Sample Selection Issues In this subsection, I provide several alternative specifications for result robustness. I first apply robustness tests to sub-samples of supply chains without the China and the U.S. Then, I exclude agriculture, mining, and other sectors in order to restrict my sample to only include firms that manufacture goods. All robustness tests are shown in Appendix C and are discussed below. First, China is a major trading partner with the United States. Relationships between China and the U.S changed considerably over the past few years. It may be a concern that my results are primarily driven by the U.S. and China. I address this concern by using sub-sample robustness checks. I use sub-samples for supply chains without without linkages to the US and China to attest that the results hold beyond the China-U.S trade. Second, I consider manufacturers only. It is a concern that agriculture and mining are 26 less substitutable, and thus the export chain is more resilient to shocks and might overesti- mate the re-entry costs. To show that firms do react to all uncertainty, I study a sub-sample that includes manufacturing firms only. In Appendix C, I qualitatively find the same results. Therefore, the findings are consis- tent and robust, and not driven by China-U.S. supply chain relationships. 1.4 Model I use the model developed by Alessandria and Choi (2007) and Caldara et al. (2020) to investigate how trade policy uncertainty transmits to the economy. There are two countries, home (H) and foreign (F), populated by a large number of identical,infinitelylivedconsumers. Idenoteforeignvariableswithanasterisk. Eachcountry has monopolistically competitive intermediate goods producers, each producing a distinct differentiated variety. The many intermediate good producers are indexed i∈[0,N t ], where N t is the mass of firms. A good intermediate producer uses capital and labor inputs to produce its variety. All firms sell their product in their own country, but only some firms export their goods abroad. As in Dixit (1989), Baldwin and Krugman (1989), and Das et al. (2007), to export, an establishment incurs a fixed cost that depends on its export status in the previous period. There is a (relatively) high up-front sunk cost F that must be borne to gain entry into the export market. In subsequent periods, establishments incur a lower but nonzero period-by- periodfixedcontinuationcost f <F tocontinueexporting. Ifanestablishmentdoesnotpay this continuation cost, it ceases to export. In future periods, the establishment can begin exporting only by incurring the entry cost F again. The cost of exporting implies that the set of goods available to consumers differs across countries and changes over time. In each country, competitive final goods producers purchase intermediate inputs from each country. The cost of exporting implies that the set of goods available to competitive 27 finalgoodsproducersdiffersacrosscountries. Theentryandexitofexportingestablishments imply that the price of intermediate goods available in a country is changing over time. The final goods are used for both domestic consumption and investment. In this economy, there existsacompletesetofone-periodstate-contingentnominalbondsdenominatedinthehome currency. LetB t denotethehomeconsumer’sholdingofabondpurchasedint,andR t denote the return on asset B t− 1 . T t is a lump-sum transfer from the government. 1.4.1 Model Mechanism Households Households in the home country choose final good consumption ( C t ), differentiated labor supply and wages for their members (l j,t , and W j,t for j∈ HH), and bond holdings (B t ) to maximize their utility E s X t β t− s U(C t ,L t ) (1.7) subject to the budget constraint ¯ P t C t +B t ≤ Z w j,t l j,t +B t− 1 R t +Π HH t +T t (1.8) where ¯ P t andw j,t denotethepricelevelandwagerate; Π HH t isthesumofprofitsofthehome country’s intermediate good producers, and T t is a lump-sum transfer from the government. Optimality requires the following saving conditions: 1=βE (Λ t,t+1 R t+1 ) (1.9) where Λ t,t+1 = Uc(t+1) Uc(t) is the real stochastic discount factor for the household in the home country.Insettingthewage,householdmemberj takesasgivenintermediategoodproducers’ 28 labor demand: l j,t = w j,t W t − ϵ w L t (1.10) where ϵ w governs the elasticity of substitution across differentiated labor inputs. The aggregate consumption is a combination of differentiated final goods varieties ( i) according to the constant-elasticity of substitution (CES) aggregator C t = Z Y t (i) ϵ p− 1 ϵ p di ϵ p ϵ p− 1 (1.11) where ϵ p ≥ 0 determines the elasticity of substitution between varieties. Final Good Producers For each variety i, the home country produces final goods using bundles of intermediates produced in the home country (Q H,t ) and bundles produced and exported by the foreign country (Q F,t ) according to: Y t = ω 1 ϵ f Q ϵ f − 1 ϵ f H,t +(1− ω) 1 ϵ f Q ϵ f − 1 ϵ f F,t ! ϵ f ϵ f − 1 (1.12) where ϵ f ≥ 0 is the elasticity of substitution between domestic and foreign bundles and ω determines the weight of home goods in final good. Final good producers’ profits are Π t =P t Y t − P H,t Q H,t − P F,t (1+τ t )Q F,t (1.13) where P H,t and P F,t are the price indexes of the domestic and foreign intermediates, respec- tively. τ t is the tariff that the home country may impose on imported intermediates. P t is the price index for final good i P t = ω(P Ht ) 1− ϵ f +(1− ω)(P Ft (1+τ t )) 1− ϵ f 1 1− ϵ f (1.14) 29 For any given level of demand Y t , cost minimization yields the demand functions Q H,t =ω P H,t P t − ϵ f Y t (1.15) Q F,t =(1− ω) P F,t (1+τ t ) P t − ϵ f Y t (1.16) Thisexpressionimpliesthathighertariffsinthedomesticcountryraisetherelativecost ofimportedintermediateinputsandhenceshiftdemandawayfromimportedinputstowards domestically-produced intermediate inputs, that is: Q H,t Q F,t = ω 1− ω P H,t P F,t (1+τ t ) − ϵ f (1.17) Ineachcountry, manyintermediategoodproducersarenormalizedtoacontinuumwith unitmassindexedj∈[0,1], whobehaveasmonopolisticcompetitors. Competitivedistribu- tors specialize in the production of (CES) bundles of intermediates purchasing intermediate varieties produced both in the home country and in the foreign country. With the export margin of the model, the measure of foreign varieties used in the production of the compos- ite foreign changes over time. With a typical Dixit-Stiglitz aggregator, there is a benefit to using smaller amounts of a greater number of varieties. To counteract the increasing returns to scale from this love-of-variety effect, I modify the aggregator of the foreign composite by introducing the addtional term N ∗− λ . This term allows to separate the love-of-variety effect from the degree of market power, which is related to the elasticity of substitution between individual varieties (Bernanke (1983)). Q H,t = Z y H,t (j) ϵ − 1 ϵ dj ϵ ϵ − 1 (1.18) Q F,t = N ∗ t − λ Z y F,t (j) ϵ − 1 ϵ dj ϵ ϵ − 1 (1.19) 30 where ϵ > 1 determines the elasticity of substitution between firms. The distributors maxi- mize profits given by: π D H,t =P H,t Q H,t − Z p H,t (j)y H,t (j)dj (1.20) π D F,t =P F,t Q F,t − Z p F,t (j)y F,t (j)dj (1.21) So the demand schedules of competitive distributors in the domestic and foreign markets: y H,t (j)= p H,t (j) P H,t − ϵ Q H,t (1.22) y F,t (j)=N ∗ t − λ ϵ − 1 ϵ p F,t (j) P F,t − ϵ Q F,t (1.23) Intermediate Goods Producers I temporarily drop the firm index and country index in this section. Each firm produces output for the domestic market (y h t ) and, if it decides to export, for the foreign market (y f t ), according to constant returns to scale technology: y h t +m t y f t =A t k α t l 1− α t (1.24) wherem t ∈{0,1}isanindicatorfunctiondenotingwhetherornotfirmjdecidestoexportin thecurrentperiod,andk t andl t arethecapitalandlaborinputs. Capitalusedinproduction is augmented by investment of final goods, I t . The law of motion for capital is given by k t =(1− δ k )k t− 1 +I t (1.25) where δ k is the depreciation rate. The term A t denotes the productivity of each firm and is composed of a country-wide 31 component η t and a firm-specific component z t such that lnA t =η t +z t The firm-wide component z t is independently, identically distributed across countries, firms, and time, z t iid ∼ N(0,δ z ). The optimal price setting requires charging a constant markup over marginal costs, which can be written as MC t = w t l t (1− α )A t k α t l 1− α t (1.26) And the optimal price and within-period profits for each firm j are p t = ϵ ϵ − 1 MC t (1.27) π t = 1 ϵ p t q t =π h t +π f t (1.28) Whenafirmdecidestoexport( m t =1),itincursfixedcostsinunitsoflaborthatdependson its export status in the previous period F(m − 1 ). Specifically, firms pay a sunk cost to enter the export market, denoted by F(0) = F, which is higher than the fixed cost of continuing exporting in each period F(1) = f. If a firm exits the export market, it must repay the sunk cost F to re-enter. The individual state of a firm is summarized by ( z,k,m − 1 ,τ ). An intermediate good producer with individual state (z,k,m − 1 ,τ ), chooses m to solves the following dynamic recursive problem: V(z,k,m − 1 ,τ )=max(π − mF(m − 1 ))+E t Λ t,t+1 V(z,k ′ ,m,τ ′ ) (1.29) the value function of exporting is V 1 (z,k,m − 1 ,τ )= π h +π f − f +E t Λ t,t+1 V(z,k ′ ,1,τ ′ ) (1.30) 32 the value function of non-exporting is V 0 (z,k,m − 1 ,τ )=π h +E t Λ t,t+1 V(z,k ′ ,0,τ ′ ) (1.31) Clearly, thevalueofaproducerdependsonitsexportstatusandismonotonicallyincreasing and continuous in z. Hence, it is possible to solve for the firm-specific productivity at which a firm is indifferent between exporting or not exporting. The marginal exit firm z 1 and marginal entry firm z 0 satisfy V 1 (z 1 ,k,1,τ )=V 0 (z 1 ,k,1,τ ) (1.32) V 1 (z 0 ,k,0,τ )=V 0 (z 0 ,k,0,τ )− F (1.33) The percentage of exporters among exporters and nonexporters can be defiend as n 1 and n 0 , respectively: n 1 =Pr(z≥ z 1 ) (1.34) n 0 =Pr(z≥ z 0 ) (1.35) Finally the law of motion for the export ratio among intermediate good producers N is N t =n 1,t N t− 1 +n 0,t (1− N t− 1 ) (1.36) 1.4.2 Trade Policy Uncertainty In practice, the level of future tariffs is uncertain. Tariffs are a random variable with two sources: tariff shock and volatility shock. I use the underlying framework in Handley (2014) to model tariff uncertainty. I assume that a shock to the path of tariffs will arrive tomorrow and take shocks to trade policy as given and do not explicitly model their source. Under a stochastic tariff process, there is an option value of waiting with a structure similar to 33 Baldwin and Krugman (1989). Since firms will make decisions based on their future profit. I assume that firms can only make expectations over the tariffs and take the price indexes as given. So τ is the only stochastic variable in the profit function. While the current tariff is known,futureprofitflowsaresubjecttothestochasticprocessfortariffs. Thefirm’sdecision to enter or exit the market is an optimal stopping problem. The tariffs follow AR(1) process described by: τ t =ρ τ τ t− 1 +µ t +exp(δ t )ϵ τ t where ϵ τ t ∼ N(0,1) (1.37) δ t =ρ δ σ t− 1 +(1− ρ δ ) ¯δ +ηϵ δ t where ϵ δ t ∼ N(0,1) (1.38) where ϵ τ and ϵ δ follow i.i.d. standard normal process 20 . The first innovation ( ϵ τ t ) affects the tariffs itself and, act like a typical fiscal shock, captures commercial policy actions not explainedbypastvaluesoftariffs. Thesecondinnovation( ϵ δ t )affectsthespreadofvaluesfor tariffs and acts like a volatility shock. The parameters ¯δ and η control the degree of mean volatility and stochastic volatility in the tariff: a high ¯δ implies a high mean volatility of the futuretariffsandahigh η ,ahighdegreeofstochasticvolatility. µ t denotesexpectationshock thatisannouncedinperiodt− 1andmaterializesinperiodt. Theparametersofinterestare the average log standard deviation of an innovation to fiscal shocks ( ¯δ ), the unconditional standard deviation of the fiscal volatility shock ( η ), and the persistence of the two processes (ρ τ and ρ δ ). I estimate the movement of tariffs across 9 economies from 1995-2015 21 following a Bayesian approach 22 . Then I use the median of each parameters to calibrate the model. Table 1.5 shows the results. 20 I use the stochastic volatility approach proposed by Fern´ andez-Villaverde et al. (2015) and Born and Pfeifer (2014) 21 I exclude China, because the China’s WTO Accession increased the TPU around 2000. 22 Follow Fern´ andez-Villaverde et al. (2015), Born and Pfeifer (2014), and Caldara et al. (2020) 34 Table 1.5: Summary of Parameters for Tariff Movement ρ τ ρ δ ¯δ η Australia 0.97 0.94 -5.71 0.46 [0.96,0.98] [0.83,0.99] [-6.39,-4.89] [0.33,0.63] Canada 0.95 0.96 -6.06 0.35 [0.93,0.97] [-6.73,-5.27] [0.87,0.99] [0.23,0.52] Europe 0.94 0.91 -6.88 0.32 [0.91,0.96] [0.77,0.98] [-7.40,-6.22] [0.19,0.47] India 0.95 0.95 -6.97 0.28 [0.93,0.96] [0,84,0,99] [-7.49,-6.19] [0.18,0.40] Japan 0.95 0.94 -8.05 0.13 [0.93,0.97] [0.76,0.97] [-8.44,-7.55] [0.04,0.29] Latin American 0.95 0.92 -6.61 0.32 [0.93,0.97] [0.81,0.96] [-7.00,-6.02] [0.21,0.47] Mexico 0.96 0.91 -6.57 0.28 [0.94,0.97] [0.85,0.94] [-6.88,-6.26] [0.22,0.38] Southeast Asia 0.99 0..96 -6.01 0.13 [0.98, 0.99] [0.94,0.97] [-6.20, -5.81] [0.04,0.29] United Kingdom 0.95 0.94 -7.05 0.22 [0.93,0.97] [0.76,0.97] [-7.44,-6.55] [0.11,0.37] Median 0.95 0.94 -6.61 0.28 The sample runs from 1995Q1 through 2015Q4 The entries in the table denote the median, 5- th and 95-th percentiles of posterior the distribution of the parameters of the stochastic volatility model An average standard deviation of 100× exp(6.61) = 0.13 % point A one-standard deviation innovation to the volatility of tariffs ( ϵ δ ) increases the standard deviation of innovations to tariff shocks to about 100 × exp(6.61 + 0.28) = 0.18 % points. Combine the value function (1.31), (1.30) and marginal cutoff conditions (1.33),( 1.32) the marginal entry firm z 0 and the marginal stay firm z 1 can be expressed as: π (z 1 )− f =− β ∗ U 1 (1.39) 35 π (z 0 )− f− (1− β )F =− β ∗ U 0 (1.40) where U 1 and U 0 are uncertainty components: U 1 =Pr(π 1 ≤ π ′ ≤ π 0 ) 1 1− β [E(π ′ |π 1 ≤ π ′ ≤ π 0 )− π 1 ]+Pr(π ′ ≥ π 0 )F ≥ 0 (1.41) U 0 =Pr(π 1 ≤ π ′ ≤ π 0 ) 1 1− β [E(π ′ |π 1 ≤ π ′ ≤ π 0 )− π 0 ]− Pr(π ′ ≤ π 1 )F ≤ 0 (1.42) Where π 0 and π 1 are the trigger profit value π 1 and π 0 , which makes the firm indifferent between entry and waiting, the definition and calculation are shown in Appendix D. Figure 1.6 shows the entry and exit rates distribution. Without uncertainty shock (δ = 0), only high tariff shock affects the probability of entry, and low tariff shock affects the probability of exit. Higher uncertainty will delay the entry even with an expected lower future tariff. Similarly, the probability of exit will decrease with higher uncertainty. Non-export firms will not enter in advance. Even though another policy shock should induce a new tariff that is lower than the current one, it is only the prospect of a bad shock that affects the decision of whether to enter today. It holds despite the convexity of profits in tariffs. When a firm enters, it weighs the expected value of profits from entering today against the value of waiting for a better shock in the future because good news in the future is offset by the opportunity cost of entry, only bad news matters when the entry investment isirreversible. However, highertariffvolatility( δ )generatesanopportunitytohavenegative profit and delay the probability of entry. On the other hand, export firms will stay in the market as long as their profits are larger than the discounted sunk costs. As shown in equation (1.41) and (1.42), U 1 ≥ 0, and U 0 ≤ 0. Uncertainty over the tariff generates a lower cost cutoff for a given current tariff than a deterministic model. The productivity premium necessary to overcome this hurdle z 1 is smaller than the deterministic productivity cutoff ˆz 1 . Since the uncertainty shocks will 36 Figure 1.6: Distribution of Entry / Exit Rate increasetheprobabilityofhavingapositivefutureprofit,firmsexperiencinganegativeprofit will wait for good news. So the TPU shocks will not increase their exit rate. Uncertainty about future trade policies delays entry and exit at the margin relative to the deterministic model. Uncertainty makes export firms on the margin more risk-loving and non-export firms on the margin more cautious. Export firms have already paid the one-time entry cost; as long as the expected profit is positive, they will stay in the market. Ontheotherhand,theentrycutoffwillincreasetheexpectationoffuturetariffsandinduces a higher entry threshold z 0 . As discussed before, non-exporter firms are more risk-averse to negative shocks, and the opportunity cost of entry offsets good news in the future. 1.4.3 Equilibrium To close the model, I specify a rule for monetary and fiscal policy. The monetary authority follows a Taylor rule that responds to inflation only: R t = 1 β (βR t− 1 ) ρ R (π ϕ π t ) 1− ρ R (1.43) whereρ R istheinertialparameterandϕ π istheweightoninflation. Thegovernmentbalances its budget each period: T t = τ t 1+τ t P Ft Q Ft (1.44) 37 Given the stochastic process for tariffs, taxes, and technology, an equilibrium is defined in the usual form, which can be found in Appendix D. 1.4.4 Calibration I solve and simulate the model by a third-order perturbation method using the pruning algorithm by Andreasen et al. (2018). As explained in Fern´ andez-Villaverde et al. (2015), the third-order approximation of the policy function is necessary to analyze the effects of uncertainty shocks independently of the first-moment shocks. The volatility shock plays an independentroleandentersasanindependentargumentintheapproximatedpolicywithout interacting with any other variable function only in a third-order approximation. Then describe the functional forms and parameter values considered for our benchmark economy. The instantaneous utility function is given as U(C,L)= (C γ (1− L) 1− γ ) 1− ν 1− ν where 1/ν is the inter-temporal elasticity of substitution, and γ is the shared parameter for consumption in the composite commodity. The parameterization is based on values commonly found in the literature. The time discount factor is β = 0.99. The curvature parameter, ν , determines the intertemporal elasticity of substitution equals 2, which is widely used in the international business cycle literature, e.g., Backus et al. (1994) and Stockman and Tesar (1995). The share parameter for consumption in the composite commodity, γ , is set to equal 0.294. The parameters for theproductionofintermediategoodsandforinternationaltradefollowAlessandriaandChoi (2007)andCaldaraetal.(2020). Isettheelasticityofgoodsdemand ϵ p to10. Theelasticity of substitution between countries ϵ f and between firms ϵ equal to 5 and 1.5, respectively. The home bias parameter ω is 0.85. The capital share of traded goods α is 0.36, and the depreciation rate δ k is 0.025. I tie the love of variety to the elasticity of substitution across 38 varieties and set λ =0. The inertia coefficient ρ R is 0.85, and the coefficient on inflation ϕ π is 1.25. The parameter F, f, and δ z jointly determine the amount of trade, characteristics of exporters and nonexporters, and the dynamics of export status.To pin these parameters down, I follow Alessandria and Choi (2007). The fixed entry cost F = 0.20897, the smaller export cost f = 0.05043, and δ z = 0.5. The choice of δ z = 0.5 is made as it leads exporters to be 15.5 percent more productive and to ship 90.2 percent more output (and hire 90.2 percent more workers). The parameters governing the remaining exogenous processes are reported in Table 1.6. Table 1.6: Benchmark Parameter Summary Preferences Discount factor β 0.99 Inter-temporal elasticity of substitution ν 2 Shared parameter for consumption γ 0.308 Production Capital Share α 0.36 Capital depreciation rate δ k 0.025 Idiosyncratic TFP volatility σ z 0.5 Country-wide TFP volatility σ η 0.007 International Trade Fixed sunk entry cost F 0.21 Fixed continuation export cost f 0.05 Elasticity of goods demand ϵ p 10 Elasticity of substitution between countries ϵ f 5 Elasticity of substitution between firms ϵ 1.5 Home bias ω 0.85 Love-of-variety λ 0 Monetary Inertia coefficient ρ R 0.85 Inflation ϕ π 1.25 Trade Policy Uncertainty Persistence of tariff level ρ τ 0.95 Persistence of tariff volatility ρ δ 0.94 Average Tariff Volatility ¯δ -6.61 Stochastic Volatility η 0.28 39 The model shows the effects of a rise in trade tensions as both a first-moment shock (an increase in the expectation of future tariffs) and a second-moment shock(an increase in the uncertainty about future tariffs). In all the experiments, I track the economy’s response to the shocks, starting from the risk-adjusted steady state and assuming that all other shocks are equal to zero. 1.4.5 Model Results Figure 1.7 presents the economy’s response to the rise in trade tensions together with higher future tariff expectations and higher tariff volatility. A rise in trade tensions leads to a sizable decline in export value (intensive margin) and the number of exporters (extensive margin). The probability of entry drops 0.4% and merges to a new steady state lower than before. The probability of exit increased by 0.6%, which leads to a reduction in the mass of exporters by 1%. Blue lines present the impulse responses without fixed costs. With free entry, the probability of entry and exit will not divert to the steady state. However, the investment drops and accounts for a significant portion of the contraction in GDP, largely driven by the expectation of a smaller export market. Figure1.8presentstheeffectsofexpectationsofhigherfuturetariffs. Exports, themass of exporters, and the relative investment of exporters all decline when tariffs are expected to rise. Higher tariff expectations about future tariffs increase the expected cost of imports, reducing expected profits. Lower export values decrease investment and labor demand, implying lower out- put and consumption. Moreover, anticipating higher marginal costs in the future, wholesale firms increase their markups, which pushes down hours worked and consumption. Inaddition, higherexpectedtariffslowerthebenefitofexportingbyshrinking theexpectedsizeoftheexportmarketsandhencetheexpectedfuturecomparableadvantage. Figure 1.9 presents the effects of an increase in uncertainty about future tariffs. Higher uncertainty about future tariffs leads to a higher variance of future desired prices. When entering the market is costly, non-export firms respond to higher tariff uncertainty by de- 40 Figure 1.7: Expectation and Uncertainty Shocks Impulse Responses to Expectation and Uncertainty Shocks. Note: Impulse responses to tariff expectations and to uncertainty shocks in the baseline model. The horizontal axis measures quarters since the shock. Variables are deviations from their steady state. Figure 1.8: Expectation Shock Impulse Responses to Expectation Shock about future tariffs. Note: Impulse responses to tariff expectation shocks in the baseline model. The horizontal axis measures quarters since the shock. Variables are deviations from their steady state. 41 Figure 1.9: Uncertainty Shock Impulse Responses to Higher Uncertainty Shock about future tariffs. Note: Impulse responses to uncertainty shocks in the baseline model. The horizontal axis measures quarters since the shock. Variables are deviations from their steady state. laying entry. On the other hand, exporting firms will delay their exit because of the options cost of re-entry. Higher trade policy uncertainty increases export participation in the short term despite fixed export costs in this calibration. Moreover, higher uncertainty reduces in- vestments and hence consumption and output. In this model, pure uncertainty will generate an option cost for non-export and export firms, increasing entry and reducing exit from the market. Figure 1.10 indicates that an isolated uncertainty shock explains the over 45 percent GDP decrease for U.S supplier countries in 2016-2019. The total GDP will decreased by around 0.36% because of the trade policy uncertainty, and the higher tariff expectation explained 53 percent, and higher tariff volatility explained a 47 percent decrease. However, the effects of uncertainty lagged over expectations. When firms know what tariffs will be in the future, they will revise their export activity immediately. Uncertainty generates a “wait and see” mentality which delays the effects. The negative effects caused by uncertainty 42 Figure 1.10: GDP Decrease in Expectation and Uncertainty Shock GDP Responses to Expectation and Uncertainty Shock about future tariffs. Note: GDP in percentage changes. Light Blue: isolated expectation shock. Dark Blue: isolated volatility shock increased until 12 periods, dropping the GDP to 0.26%. 1.5 Policy Implication and Conclusion 1.5.1 Policy Implication Thispaperprovidesevidencethatcanbeusedtoinformpolicydecisions. First,policymakers should act to avoid the adverse macroeconomic effects of policy uncertainty. Clear commu- nication of policy objectives and multilateral trade agreements should be a priority. Higher uncertainty induces higher future price variance, reduces export activities, and disrupts the global supply chain. Early signs of fragmentation have been visible for a while, with trade policy uncertainty spiking in recent years, countries imposing ever more trade restrictions, and national security concerns resulting in new restrictions being placed on inward foreign direct investment. Collaborative solutions are needed to avoid the adverse effects of greater 43 fragmentation and to ensure that trade continues to act as an engine of growth. The fo- cus should be on rolling back damaging trade restrictions and reducing policy uncertainty through clear communication of policy objectives and processes to address legitimate na- tional security concerns while addressing competitive weaknesses through structural reforms that lift productivity. Second, in the long term, governments should invest in transportation to reduce sunk costs to increase firms’ mobility; for example, more ports with higher capacity improve net transportation and easier access to inventory and labor. Especially for young small firms, the lower adjustment cost will improve their reliance on uncertainty shocks. In an economy with no adjustment costs, the results show that increased uncertainty about tariffs causes the measure of the exporter relationships, investment, and GDP to expand. Overall, with higher and higher global trade policy uncertainty, the government and policy- makers should take action to reduce trade tensions. Active engagement and dialogue between policymakers worldwide, including in multilateral forums, will be vital to avoid the sharpest and most harmful fragmentation scenarios. 1.5.2 Conclusion In this paper, I provided an approach to quantifying the macroeconomic effects of changes in trade policy uncertainty by separating expectation shocks from volatility shocks. Facing growing policy uncertainty, firms have a fundamental trade-off regarding production costs and the resiliency of their export chain. How they manage those risks depends on future tariff expectations. Higher tariffs reduce the export value and delay both entry, exit, and investment; however, along with higher uncertainties, exporter firms appear to wait for the uncertainty to turn into actual events because of the option value of re-entry. Finally, I use a two-country general equilibrium model with heterogeneous firms and endogenous export decisions and find that both higher expected tariffs and increased uncertainty about future tariffs deter investment. The framework emphasizes the role of price rigidity and export 44 fixed costs as essential mechanisms that amplify the effects of trade policy uncertainty. 45 Chapter 2 Navigating Uncertainty: A Study of Input Choices in the Face of Shocks The present paper investigates how uncertainty affects firms’ input choices using both theoretical and empirical evidence. It demonstrates that uncertainty prompts importers to assign a higher value to the expected tariff, leading to an increase in the import cost and, as a result, a reduction in imports. Furthermore, the model predicts that firms that are more geographically diversified in sourcing tend to be more resilient to uncertainty disruptions. In addition, high-productivity firms are more likely to be geographically diversified in input sourcing compared to their low-productivity counterparts. To illustrate these effects, the studyemploysfirm-leveltradedatafromChina’saccessiontotheWorldTradeOrganization in 2002. The analysis reveals that import growth significantly increases in response to a reduction in uncertainty, even after controlling for the current tariff level. Specifically, the study finds that a 1% reduction in uncertainty results in a 2.6% increase in import growth, whereas this increase is only 1.7% in the absence of diversification. 46 2.1 Introduction Trade policy uncertainty has become an increasingly relevant topic for researchers and poli- cymakers as globalization continues to shape the world economy. The impact of uncertainty on firms’ input choices has received significant attention in the literature. In this paper, I explore how uncertainty affects a firm’s input choices and how diversification in global sourcing can make the firm more resilient in the face of uncertainty. Global sourcing al- lows firms to access a wide variety of inputs and exposes them to foreign shocks, such as natural disasters, wars, and epidemics. For example, the 2001 earthquake in Japan caused significant disruptions to Japanese firms, and the 2003 SARS epidemic in China resulted in a drop in trade volume (Boehm et al. (2020), Fernandes and Tang (2021)). More recently, the trade war between the United States and China, triggered by a series of tariff measures introduced by the Trump administration, has had significant implications for international trade (Handley and Lim˜ ao (2018)). While much of the existing literature on trade policy uncertainty has focused on export access, the impact of uncertainty on the complex process of input sourcing has received less attention. To address this gap in the literature, I exploit the 2002 China’s WTO entry as an experiment to examine the impact of uncertainty on firms’ input choices and the role of global sourcing diversification in building resilience. The results show that when there is an increase in trade policy uncertainty, firms’ profits and intermediate input purchases decrease and that high-productivity firms are more geographically diversified in their input sourcing than low-productivity firms. Additionally, this paper shows that sourcing diversification makes firms more resilient to trade policy uncertainty if sourcing decisions exhibit complementarity across trade routes. This study provides important insights into the strategic sourcing decisions of firms operating in a globalized and uncertain economic environment. I followed Antr` as et al. (2017) and built a model in which firms source inputs from different origins. I assume that there is a possibility that trade policy will change in the 47 next period. The model has the following key predictions. First, importers need to pay an import cost, which includes the tariff. An increase in the import cost and the firm does not abandon this route will weakly reduce the profit and input purchases by this firm. Secondly, firms value both current tariffs and expected future tariffs. Trade policy uncertainty leads firms to put more weight on expected tariffs and increase the import cost. Thirdly, each firmischaracterizedbyitscoreproductivity, andaddingnewtraderoutesincursfixedcosts; only high-productivity firms can afford to source via more routes if sourcing decisions are complementary across trade routes. Finally, I find that the pass-through of uncertainty on trade routes to marginal cost and purchase value is proportional to the sourcing intensity. Higher trade policy uncertainty induces lower import volume. This study provides important insights into the strategic sourcing decisions of firms operating in a globalized and uncertain economic environment. The findings suggest that firms can benefit from diversifying their sourcing strategies to increase their resilience in the face of uncertainty. Policymakers can also use these insights to design trade policies that promote global sourcing and reduce uncertainty. ItakeChina’sWTOaccessionasanexample. China’sWTOentryprovidedasubstantial reductionintradepolicyuncertaintyduetotheWTOguaranteeoftheMostFavoredNation (MFN) treatment. China obtained a temporary MFN in 1980. In the 1990s, because of the Tiananmen Square protests, U.S. Congress voted on a bill to revoke MFN status every year, and the House passed it three times. After the MFN status had been revoked, the United States would have reverted to Smoot-Hawley tariff levels, and a trade war may have ensued. China’sWTOaccessionremovedthethreatthatforeigncountriesmightatsomefuturetime increase the tariff. In 2000, the average worst tariff faced by China was above 30 percent; after China entered WTO, the average worst tariff was below 10 percent. The analysis of China’s imports reveals a number of robust links between trade policy uncertainty and firm input choices. I find that trade volume growth was positively related to uncertainty reduction following China’s WTO accession. In 2000, the total import value in China was 48 200 billion. It jumped to 400 billion after China entered the WTO in 2003. Moreover, there were 60 thousand importers in 2000 which increased to 80 thousand in 2002. I find strong evidence that the reduction of trade uncertainty has a positive effect on China’s imports. There was a tariff drop after China entered WTO. China’s average tariff was as high as 40% at the beginning of the 1990s, and down to below 20% on the eve of WTO’s entry. After China entered WTO, the tariff dropped to below 10%. However, I find that the reduction of uncertainty has a significant positive effect on imports after I control the tariff level. This paper contributes to the recent literature on trade policy uncertainty (TPU), pi- oneered by Handley and Lim˜ ao (2015) and Handley and Lim˜ ao (2017), who provided the- oretical and empirical evidence that policy uncertainty can significantly affect firm-level investment and entry decisions in the context of international trade. Specifically, Handley and Lim˜ ao (2017) estimated and quantified the impact of trade policy on China’s export boom to the U.S. following China’s WTO entry. They argued that the reduction of the U.S. threat of a trade war accounted for over one-third of export growth in China. Feng et al. (2017) used China’s WTO accession as an example to demonstrate that the trade policy reduction induces firms with higher productivity or lower prices to enter the export market. However, most literature focuses on exporters and export growth, whereas this study pro- vides insights into the changes for importers. The author investigates that import cost is increasing in uncertainty. By using the data in China, the study shows that a 1% reduction in trade policy uncertainty induces 6.8% import growth at the firm-product level and 2.8% at firm level. The author adds a tariff as a dummy variable in the regressions and finds that, after controlling for the tariff level, the reduction of uncertainty still leads to a 2.4% increase in import growth in firm-product data and a 2.2% increase in firm-level data. Second, this paper enriches the model created by Antr` as et al. (2017) and Eaton and Kortum (2002). Antr` as et al. (2017) developed a multi-country sourcing model in which firms self-select into importing based on their productivity and country-specific variables. Thispaperaddsanuncertaintyshockintothatmodelandarguesthattheimportcostshock 49 pass-through to marginal cost and import value is related to the sourcing intensity. With trade policy uncertainty, for a firm with higher sourcing intensity the purchase value will decrease more. Moreover, Fillat and Garetto (2015), and Esposito (2022) examine demand diversifications for multinationals and exporters, respectively. The focus of this study is on diversification in sourcing and the resilience of uncertainty. Finally, this paper also relates to the literature on trade fluctuations caused by shocks. For example, the financial crisis (Baldwin (2011); Bems et al. (2013); Watson (2019); Chor andManova(2012)). Andotherepidemicsornaturaldisasters(Carvalho(2017); Heo(2021); Carre˜ noetal.(2020)). Thisstudyismoregeneralandfocusesonthechangeinimportcosts. Foreign financial shocks, natural disasters, and epidemic shocks can cause supply or demand shocks, price shocks, transportation shocks, and iceberg shocks. Each kind of shock can be consideredasonekindofincreaseinuncertaintyandanincreaseinimportcost. Finally,this study contributes to the rapidly growing literature on the impact of the current coronavirus pandemic by shedding light from the trade angle on which sectors and firms may be more resilient to eventual recovery. The rest of the paper is organized as follows: Section 2 presents a model to explain how the effect of trade policy uncertainty varies with the sourcing intensity. Section 3 introduces the data and presents the empirical results of the study. Finally, Section 4 provides the conclusions. 2.2 Model This section of the paper presents a multi-country model of global sourcing. The model is designed to provide theoretical predictions on sourcing diversification and resilience to supply chain disruptions. These predictions will guide the empirical analysis in this paper. To incorporate shocks and uncertainties into the model, I follow the framework developed by Antr` as et al. (2017). 50 In this model, firms source inputs from different origins, and there is a possibility that tradepoliciesmaychangeinthenextperiod. Themodelmakesseveralkeypredictionsbased on this setup. The model assumes that firms source inputs from different origins, and there is a possibility of changing trade policies in the future. Importers must pay an import cost, which includes the tariff. An increase in the import cost will reduce the profit and input purchases by the firm, assuming the firm does not abandon this route. Firms value both current and expected future tariffs. Trade policy uncertainty leads firms to put more weight on expected tariffs and increase the import cost. The model also incorporates the concept of core productivity, which characterizes each firm’s ability to produce output using inputs. Adding new trade routes incurs fixed costs, and only high-productivity firms can afford to source via more routes if sourcing decisions are complementary across trade routes. I show that high-productivity firms are more geographically diversified in their input sourcing than low-productivity firms. 2.2.1 Demand IconsideraworldwithJ countrieswhereindividualsvaluetheconsumptionofdifferentiated varieties of goods based on a standard symmetric CES aggregator. The utility function of an individual in country i is given by: U i = Z ω∈Ω i q(ω) ( σ − 1 σ ) dω ( σ σ − 1 ) (2.1) where σ is the demand elasticity, and Ω i is the set of final-good varieties available in country i∈J. The demand for final goods in country i is determined by the following equation: q i (ω)=E i P σ − 1 i p i (ω) − σ (2.2) where, p i (ω) is the price of variety ω, P i is the standard ideal price index associated with the CES aggregator, and E i is the aggregate spending on final goods in country i. 51 To simplify the analysis, I define a market demand as: B i = 1 σ σ σ − 1 1− σ E i P i (2.3) whereB i representsthequantitydemandedofallvarietiesincountry i. Themarketdemand capturestheoveralldemandforfinalgoodsincountry ibasedonthepricelevelanddemand elasticity. This setup provides a framework to analyze the global market and trade patterns by considering how the demand for final goods in each country depends on the prices of the different varieties of goods. 2.2.2 Production and Trade There are N i final-good producers in each country i ∈ J, and labor is the only factor of production. Each final-good producer competes in a monopolistically competitive market with free entry. The final-good producers have a core productivity φ that is drawn from a distribution G i (φ), where φ∈[φ i ,∞]. As in Melitz (2003), this productivity is learned only after paying a fixed entry cost f ei units of labor. To produce the final goods, firms assemble intermediate inputs sourced from interme- diate input producers located in different origins. The intermediate input producers use constant-return-to-scale technologies for production with labor as the only input and sell their outputs competitively. The unit labor requirement is defined as a j (v,φ) for the input producer v ∈ [0,1] located in region j who supplies inputs for a firm with productivity φ. The firm-specific a j (v,φ) is drawn from the following Fr`echet distribution: Pr(a j (v,φ)>a)=e − T j a θ ,T j >0 (2.4) where T j is the average efficiency of intermediate input producers from origin j, and θ determines the variability of productivity draws across inputs. Althoughintermediatesareproducedworldwide,afinal-goodproducerbasedincountry 52 i can only offshore in country j after incurring a fixed cost f ij units of labor. The global sourcing strategy of firm φ in country i is denoted by J i (φ), which is the set of countries wherefirm φincountryihaspaidtheassociatedfixedcost w i f ij . Foreachcountryj∈J i (φ), shipping intermediates to country i incurs a cost τ ij , which includes tariffs and other iceberg costs. As a result, the cost for firm φ in country i to buy intermediates v from country j is given by τ ij a j (v,φ)w j . The price is given by: z i (v,φ;J i (φ))= min j∈J i (φ) τ ij a j (v,φ)w j (2.5) The marginal cost for firm φ in country i of producing one unit of a final good is expressed as: c i (φ)= 1 φ Z 1 0 z i (v,φ;J i (φ)) 1− ρ dv 1 1− ρ (2.6) This model provides a framework to analyze production and trade patterns by consid- ering how firms produce and source inputs from different origins to produce final goods in a global market. 2.2.3 Optimal Sourcing Intheprocessofdeterminingtheoptimalsourcingstrategy,final-goodproducersmustdecide which trade routes to use for sourcing (extensive margin) and how much inputs to source from each route (intensive margin). To this end, Eaton and Kortum (2002) proposes a probability function for sourcing through customs district j, given by: χ ij (φ)= T j (τ ij w j ) − θ Θ i (φ) ,j∈J i (φ) (2.7) Here, Θ i (φ) summarizes the sourcing capability of firm φ from i, given by: 53 Θ i (φ)= X k∈J i (φ) T k (τ ik w k ) − θ (2.8) Based on the Fr`echet assumptions, Eaton and Kortum (2002) shows that the cost func- tion for a firm sourcing from customs district i is given by: c i (φ)= 1 φ (γ Θ i (φ)) − θ (2.9) Using the demand equation and the derived marginal cost function, the firm’s profits can be expressed as a function of the sourcing strategy J i (φ): π i (φ)=φ σ − 1 (γ Θ i (φ)) (σ − 1)/θ B i − w i X j∈J i (φ) f ij (2.10) The optimal sourcing strategy I ij (φ)∈0,1 J j =1 is such that it maximizes firm profits, where the indicator Iij equals 1 if firm φ in country i sources from country j and equals 0 otherwise. However, there is no explicit solution to this problem, and it can deliver 2 J possible strategies. As noted by Antr` as et al. (2017), the optimal sourcing strategy has the property that a firm’s sourcing capability Θ i (φ) is non-decreasing in φ. Moreover, if (σ − 1)/θ ≥ 1, then J i (φ L )⊆ J i (φ H ) for φ L ≤ φ H . This is characterized in Proposition ??. The sourcing strategies given by J i (φ) is characterized by the following problem: max I ij ∈{0,1} J j=1 π i (φ,I i1 ,I i2 ,....,I iJ )=φ σ − 1 (γ J X j=1 I ij T j (τ ij w j ) − θ ) σ − 1 θ B i − w i J X j=1 I ij f ij (2.11) wheretheindicatorI ij equals1iffirm φincountryisourcefromcountryj, otherwiseequals 0. As noted by Antr` as et al. (2017), there is no explicit solution to this problem, this can deliver 2 J possible strategies. However, the solution has the following properties: Proposition 1 The optimal sourcing strategy I ij (φ)∈{0,1} J j=1 is such that: 54 1. a firm’s sourcing capability Θ i (φ) is non-decreasing in φ. 2. if (σ − 1)/θ ≥ 1, then J i (φ L )⊆ J i (φ H ) for φ L ≤ φ H . Part (1) asserts that more productive firms choose a larger sourcing capacity. This means that such firms can either import from more countries or source inputs from the country with higher technology. Part (2) states that when sourcing decisions are complementary, the optimal sourcing strategy J i (φ) is weakly increasing in productivity. This implies that firms with higher productivity levels tend to have a more diverse sourcing strategy and are less concentrated. In other words, they are less likely to rely on a single trade route or a single input source. Overall, the proposition suggests that a firm’s productivity level plays a crucial role in determining its optimal sourcing strategy, and more productive firms tend to have a more diversified and less concentrated approach to sourcing. FollowingAntr` asetal.(2017), thefirm-levelbilateralinputpurchasesfromanycountry j are a fraction (σ − 1)χ ij (φ) of firm profits, they can be expressed as: M ij (φ)=(σ − 1)B i γ (σ − 1)/θ φ σ − 1 (Θ i (φ)) (σ − θ − 1)/θ T j (τ ij w j ) − θ (2.12) When (σ − 1)/θ ≥ 1, that is it is complementary, the M ij (φ) is increasing in terms of Θ i (φ). Proposition 2 Holding constant the market demand level B i , whenever (σ − 1)/θ ≥ 1, 1. an increase in the sourcing potential T j (τ ij w j ) − θ or a reduction in the cost τ ij of any country j, will weakly increases the input purchases by firms in i not only from j but also from all other countries. 2. an increase in the cost τ ij and firm φ in country i does not abandon this route, will weakly reduce the input purchases from all j ∈ J i (φ) and reduce more for purchases from country j. 55 Proposition 2 provides insights into how changes in the sourcing potential and costs of a particular country affect the input purchases of final-good producers. Part(1)ofthepropositionstatesthatanincreaseinthesourcingpotentialoradecrease in the cost of any country will lead to an increase in the input purchases of final-good producers not only from that country but also from all other countries. This means that an improvement in the efficiency or lower costs of any input supplier will make it more attractive for final-good producers to source from that supplier. Additionally, the increase in input purchases from other countries shows that the gains are not just restricted to the country with the improvement, but it also benefits other input suppliers. Part (2) of the proposition highlights that if the cost of a particular sourcing route increases,afinal-goodproducerincountry iwillnotimmediatelyabandontheroute. Instead, it will reduce input purchases from all countries in the sourcing strategy. This implies that inputpurchasesfromallcountriesinthesourcingstrategyarecomplementary,andadecrease inpurchasesfromonecountrywillalsoreducepurchasesfromothercountries. Furthermore, the reduction in input purchases from a particular country will be higher compared to the reduction in input purchases from other countries in the sourcing strategy. 2.2.4 Equilibrium In the context of the general equilibrium of the model, the fixed cost of entry implies that final-good producers can only observe their productivity level after paying this cost. As noted by Antr` as et al. (2017), the free entry condition can be expressed as: Z ∞ ˜ φ i π i (φ)dG(φ)=w i f ei (2.13) Here,G(φ)representsthedistributionofproductivitylevelsacrossfinal-goodproducers, and ˜ φ i denotestheproductivitythresholdrequiredforfirmsincountry itoenterthemarket. The measure of final-good producers in each region can be expressed as: 56 N i = L i σ ( R ∞ ˜ φ i f ij dG i (φ)+f ei ) (2.14) where L i denotes the labor endowment of country i. Finally, the total imports of all firms in country i from country j is given by: M ij (φ)=N i Z ∞ ˜ φ i M ij (φ)dG i (φ) (2.15) This equation captures the fact that the quantity of inputs sourced from country j depends on the productivity levels of firms in country i and their corresponding sourcing strategies. 2.2.5 Uncertainty There are different types of uncertainties that can impact the cost of imports. In this study, the focus is on measuring tariff uncertainty. The approach used follows that of Handley (2014),wherepolicyuncertaintyistreatedastheappliedtariffrate. BeforeChina’saccession to the World Trade Organization, foreign countries had the option to change the tariff rate at any time. This is modeled as uncertainty, and λ is defined as the arrival rate, which characterizes the risk that a foreign country will increase its tariff rate in each period. In each period, firms pay fixed costs to enter the market and learn their productivity φ. They arealsoawareofthecurrentappliedtariffrateandtheleveloffuturetradepolicyuncertainty, which enables them to calculate the present profit. Based on their belief in the trade policy environment,theymakedecisionsonthequantityandroutestoimport. Ifthecurrentapplied tariff rate or the degree of future policy uncertainty changes, the current equilibrium will not hold, and it will trigger firms to change their M ij (φ) decisions. If the risk is significant enough or the level of future uncertainty is high, firms may choose to abandon the route. Giventhetimelineforfirmdecisions, thefirmproblemcanbesolvedbackwards. First, firms pay fixed costs and learn their productivity. Second, conditional on the current tariff rate 57 and policy environment, firms calculate their present value of profits. Third, firms decide on their intensive and extensive sourcing strategies. Inordertoaccountfortariffuncertainty,thecost τ ij isassumedtoincludethetariff. The changeintariffisthenrepresentedbyachangein τ ij , holdingtheicebergcostconstant. The weighted applied tariff faced by a firm in location i is represented by τ i . The assumption is made that a change in tariff will not lead firms to abandon import routes, though the possibility of abandonment and fixed costs associated with it may be considered in future studies. For a firm in country i with productivity φ, the value function can be expressed as: Π i (τ i,t ,φ)=π i (τ i,t ,φ)+β ((1− λ )Π i (τ i,t ,φ)+λE τ Π i (τ i,t+1 ,φ)) (2.16) where β represents the discount rate of future profits, and λ represents the probability of a change in trade policy. The expectation term E τ is taken with respect to the distribution of possible tariffs. Firms, conditional on the current tariff rate and policy environment, calculatetheirpresentvalueofprofitsanddecideontheirsourcingstrategies.Theexpectation term is taken based on the distribution of possible tariffs. After defining the value function as the expected present discounted value of profits, I take expectations on both sides to obtain the simplified expression: Π i (τ i ,φ)= 1 1− β E τ π i (τ i ,φ) (2.17) Substituting this back into the equation 2.16 for the value function, I obtain: Π i (τ i,t ,φ)= 1 1− β 1− β 1− β (1− λ ) π i (τ i,t ,φ)+ βλ 1− β (1− λ ) E τ π i (τ i,t+1 ,φ) = π i (τ i,t .φ) 1− β 1− β 1− β (1− λ ) + βλ 1− β (1− λ ) E τ π i (τ i,t+1 ,φ) π i (τ i,t ,φ) = π i (τ i,t ,φ) 1− β U (2.18) 58 where U is the uncertainty factor: U = 1+µ Eτ π i (τ i,t+1 ,φ) π i (τ i,t ,φ) (1+µ ) (2.19) with µ = βλ 1− β ∈[0,1]. Thus, the simplified expression for the value function becomes: Π i (τ i,t ,φ)= 1 1− β 1 1+µ π i (τ i,t ,φ)+ µ 1+µ E τ π i (τ i,t+1 ,φ) (2.20) The value of an importing firm depends on the uncertainty factor it faces, which can be summarizedbytwoterms: theexpectationtermandthearrivalrate. Asnotedpreviously, if theuncertaintyfactorU islessthan1inequation2.18,thevalueofthefirmwillbelower. The expectation term, denoted E τ π i (τ i,t+1 ,φ), accounts for the expected future tariffs that the firm will face. A larger unconditional distribution of tariffs results in a smaller expectation term. For example, before China’s entry into the WTO, Chinese firms faced the non-normal trade relation tariff (non-NTR tariff), which were much higher than the WTO bound tariffs that they faced afterwards. This led to a decrease in the expectation term. The second term is the level of trade policy uncertainty U and µ , which depend on the arrival rate λ . A larger arrival rate indicates a greater probability that tariffs will rise compared to their current low applied rate. An increase in λ leads to 1 1+µ becoming smaller than µ 1+µ . Firms care more about future expectations and place greater weight on the expectation term. Therefore, an increase in tariffs will result in a greater reduction in the firm’s value function. However, China’s entry into the WTO has reduced the likelihood of changes to tariffs due to the MFN treatment that the WTO guarantees. Incorporating the firm’s profit function (equation 2.10), the only variable that changes isthetariff. Toaccountforthis, the”realtariff”isdefinedas τ r , whichrepresentstheactual value that the firm uses to calculate its profit. The real tariff is a weighted average of the current applied tariff and the expected future tariff. 59 Proposition 3 Define τ r as the real value firm plug into their profit function: 1. τ r is the weighted average of current applied tariff and expected tariff: τ r i,t = 1 1+µ τ i,t + µ 1+µ E τ (τ i,t+1 ) (2.21) 2. Firm’s profit can be expressed as: π i (τ r i,t ,φ)=φ σ − 1 (γ Θ i (τ r i,t ,φ)) (σ − 1)/θ B i − w i X j∈J i (φ) f ij (2.22) 3. Firm’s input purchases vale varies to λ and E τ (τ i,t+1 ): M ij (τ r i,t ,φ)=(σ − 1)B i γ (σ − 1)/θ φ σ − 1 (Θ i (τ r i,t φ)) (σ − θ − 1)/θ T j (τ r ij,t w j ) − θ (2.23) An positive of λ and an increase in E τ (τ ij,t+1 ) of any country j, will increase the ”real tariff”, then decreases the importer’s profit and input purchases by them in i not only from j but also from all other countries. Proof: see Appendix. The uncertainty of unfavorable trade policies increases the ”real tariff” that firms face. Evenifthecurrenttariffisnothigh,thepossibilityofaworsetradepolicyenvironmentleads firms to be more cautious about future imports and potentially import less in the current period. This negative effect is felt even if the unfavorable ”shock” never materializes. This was observed in China before it entered the WTO, where the high uncertainty level resulted in reduced imports. A similar effect was seen during the ”Trade War” between China and the U.S, where heightened uncertainty led importers to source from alternative routes or use more domestic inputs. Future research could explore when importers decide to change their sourcing strategies and how these changes unfold. 60 2.2.6 Intensity Proposition 3 predicts that under uncertainty importers’ profit will decrease and purchase less from other countries. This section, I will investigate what kind of firms can be more resilient when they are facing an uncertainty. Proposition1impliesthathighproductivityfirmstendtobemorediversifiedalongwith the extensive margin since they source from more trade routes. Moreover, their inputs are less concentrated. Proposition 4 If sourcing decisions are complementary across trade routes, that is σ − 1− θ ≥ 1, the concentration of firms’ sourcing strategies as measures by Herfindahl-Hirschman Index HHI i = P χ 2 ij is non-decreasing in φ In order to measure resilience, I utilize the concept of pass-through, which refers to the extent to which adverse shocks impact firm performance. A higher level of pass-through indicates lower resilience. To estimate pass-through, I employ a methodology introduced by Feenstra (1994) which measures the welfare gains from new product varieties. By adapting this approach using the hat algebra method, I am able to determine: Proposition 5 For a small shock τ ij increase to τ ′ ij and the firm does not abandon this import route, I have: 1. The pass-through to the marginal cost is given by: ∂ln( [ c i (φ)) ∂ln(c τ ij ) = χ ij (φ) 1− P j∈N i (φ) χ ′ ij (φ) 2. If σ − θ − 1≥ 0 then: ∂ln( [ c i (φ)) ∂ln(c τ ij ) =χ ij (φ) ∂ 2 ln( [ c i (φ)) ∂ln(c τ ij )∂φ ≤ 0 61 ∂ 2 ln( [ c i (φ)) ∂ln(c τ ij )∂y ij >0 where y ij =T j (τ ij w j ) − θ , N i (φ) is set of new routes used by firms. Proof: see Appendix. High productivity firms are found to be more resilient to adverse shocks, while firms are lessresilienttoshocksonmoreappealingtraderoutes. Theextensiveandintensivemargins, captured respectively by 1 1− P j∈N i (φ) χ ′ ij(φ) and χij (φ), play a crucial role in determining the pass-through of shocks. If the margins are complementary, the pass-through depends solely on the intensive margin, which measures how much the firm imports from a foreign country. Proposition 5 implies that no firm will add new trade routes when facing adverse shocks, and thus, the pass-through is determined by the intensive margin. The findings suggest that firms that solely rely on the trade route hit by the shock experience a pass-through of 1, whereas diversified firms with χ ij (φ) → 0 experience a pass-through of 0. Additionally, the pass-through of adverse shocks is found to be weakly decreasing in firms’ productivity and increasing in firms’ sourcing potential. This is because high productivity firms tend to have greater sourcing diversity and less concentration. Fi- nally, the pass-through is also found to be more significant on popular trade routes that are facing the shocks. However, it is not usually to obvious the marginal cost. What I can test is the import value for each firm. Proposition 6 For a small shock which increase τ ij to τ ′ ij and the firm do not abandon the route, if the choices are complementary: σ − θ − 1 > 0, the value of inputs that firm φ in country i purchase from other countries: ∂ln( ˆ M ik (φ)) ∂ln(ˆ τ ij ) =        − θ − (σ − θ − 1)χ ik (φ),j =k − (σ − θ − 1),otherwise 62 Proof: see Appendix Other than parameter θ , which captures the direct impact of the shock, there is an additionalterm(σ − θ − 1)χ ij (φ)whichispositiveifthesourcingdecisionsarecomplementary, and negative if inputs are substitutable. The shock drives down the marginal demand curve forallinputsroutesifthesourcingdecisionsarecomplementary. Incontrast,iftheinputsare substitutable, the shock will drive up purchases value from other routes. The pass-through also varies the sourcing intensity χ ij (φ). The feedback effect is stronger if the firm has a heavier load in inputs from the route being shocked, which tends to be the case for a less diversified firm. Finally, it also indicates that the shock not only affects its own routes but also affects each route in the firm’s sourcing strategy. 2.3 Empirical 2.3.1 Background In this study, the Chinese National Bureau of Statistics (NBS) data for the years 2000-2006 was used to calculate the Total Factor Productivity (TFP) for firms in China, following the methodology of Loren Brandt, Johannes Van Biesebroeck and Yifan Zhang. The NBS databasecoversallstate-ownedenterprisesandotherfirmswithsalesabove5millionChinese Yuan (approximately US$60,000) and provides information on firms’ financial statements, names, addresses, phone numbers, postcodes, and other relevant details. In addition, the ChineseCustomsdatafortheyears2000-2006wasused,whichcoversallChineseimportand exporttransactionsandprovidesinformationsuchasthevalue,quantity,origin, destination, Chinese customs district for clearance, and information about the Chinese import/export entity. To match the customs data to NBS, firms’ phone numbers, company names, and party ID were used. The tariff data for the years 2000-2006 were obtained from the World Trade Organization (WTO), providing the applied weighted average tariff at the HS2 and 63 HS6 code level, as well as the MFN bound tariff for the years 2002-2006. Prior to China’s accession to the WTO, non-MFN tariff data were obtained from Feenstra, Romalis, and Schott (2002). China’s trade regime underwent significant changes from the 1980s onwards. Prior to the 1970s, Chinese trade took place within the context of the “planned economy,” where the governmentmadeimportplansforthepreviousyear, coveringover90percentofallimports. China began reforming its trade regime in the 1980s and 1990s, obtaining temporary MFN status by the 1980s. However, China’s annual renewal of MFN status was frequently chal- lengedbyanti-Chinesepressuregroups,particularlyintheaftermathofthe1989Tiananmen Square protests’ suppression. The Bush I administration and Congress imposed administra- tive and legal constraints on investment, exports, and other trade relations with China in response. In 1992, the Clinton presidency began with an executive order that linked renewal of China’s MFN status with seven human rights conditions, although Clinton reversed this position a year later. With the United States-China Relations Act of 2000, China was al- lowed to join the WTO in 2001 and was granted most-favored-nation (MFN) status. By 2005, China had reduced its import tariffs to a general level of 9.9%, down from nearly 40% intheearly1990s. IntheyearsimmediatelyfollowingWTOaccession, bothChineseexports and imports rose rapidly, with imports growing faster than exports. 2.3.2 Uncertainty Proposition3predictsthatimportvaluevarieswithtradepolicyuncertainty,whichImeasure using four variables: τ , dτ , gap, and dgap. The first variable τ is the average weighted tariff rate, while dτ is the average applied tariff difference between two years t 0 and t 1 , where a positive value implies a decline in tariffs. The third variable, gap, measures the level of uncertainty by calculating the difference between the worst possible tariff and the current tariff in a given year, with a decrease indicating a decrease in uncertainty. Before China’s accession to the WTO, each country could change tariff levels and charge higher tariffs at 64 any time, resulting in a large gap. After China’s accession to the WTO, countries could not charge tariffs higher than the bound tariff level, leading to a significant decrease in gap. The fourth variable dgap measures the reduction in uncertainty between two years t 0 and t 1 and iscalculatedasdgap t 0 ,t 1 =gap t 0 − gap t 1 . Positivevaluesofdgapindicateareductionintrade policyuncertainty. Astheboundtariffratesremainedsimilarfordecades,thesevariablesdid not change significantly after 2002, with the shock and uncertainty level changes occurring in 2001 and 2002. To examine the relationship between import value and trade policy uncertainty, I run a simple regression: ln(import i,t )=α fix +β 1 τ t +β 2 gap t +ϵ i,t (2.24) The results are presented in Table 2.1. It should be noted that due to technological changes, varying final goods assembly, and market entry or exit, some firm-product-year leveldataisnotavailable. Tomitigateanypotentialbias, aggregatedataisusedtoestimate the relationship between import value and tariff environment. The dependent variable in columns (1) and (2) is the HS2 firm-product-year level import value, while in columns (3) and (4), it is the firm-year level total import value. The trade policy variables, including the average tariff τ , the uncertainty level gap, the tariff difference dτ , and the reduction in uncertainty dgap, are constructed based on the definitions introduced earlier in this section. The analysis also controls for year and industry fixed effects. The results suggest that the average tariff τ has a significant negative effect on import value, whereas the current uncertainty level does not. This could be because firms evaluate trade policy uncertainty in the pre-shock period, while gap represents the current uncertainty level. As the uncertainty level did not change significantly after China entered the WTO in 2002, it did not have a significant negative effect on import value. Building on the results of the previous simple regression, which indicated that higher tariffs led to lower import values, I estimate the baseline regression using the following 65 [b] Table 2.1: Import Value and Tariff Environment Firm-product-year Level Firm-year level (1) (2) (3) (4) tariff -0.00390 ∗∗∗ -0.00463 ∗∗∗ -0.0210 ∗∗∗ -0.0208 ∗∗∗ (7.88) (9.38) (-32.67) (-32.22) gap 0.00723 0.0117 (22.83) (12.82) cons 10.16 ∗∗∗ 10.03 ∗∗∗ 12.22 ∗∗∗ 11.96 ∗∗∗ (109.43) (107.96) (758.00) (446.33) industry effects Yes Yes No No year effects Yes Yes Yes Yes r2 0.157 0.157 0.00229 0.00256 N 3510265 3510265 633091 633091 t statistics in parentheses ∗ p< 0.1, ∗∗ p< 0.05, ∗∗∗ p< 0.01 In column (1) and (2) tariff is HS2 code level weighted average applied tariff obtained from WTO. Column (3) and column (4) tariff is frim level weighted average applied tariff obtained from WTO. Firm’s import value, date, product, and original country are obtained from China Customs Data. Before 2002, the worst tariff is calculated as the average non-MFN tariff obtained from Datta and Kouliavtsev (2009). After Chian enter WTO, it is the MFN bound tariff, which is obtained from WTO. 66 Table2.2: ImportGrowthandTradePolicyUncertaintyFirm-Product-YearLevel [b] (1) (2) (3) (4) dgap 0.0686 ∗ 0.0344 0.0386 ∗ 0.0240 ∗ (-1.17) (-1.45) (-1.35) (-1.09) dtariff 0.227 0.0385 0.0852 ∗∗ (0.94) (1.19) (1.99) tariff 0.0329 ∗∗ 0.00353 ∗∗ (0.81) (-0.10) gap 0.0873 (0.91) cons 0.329 -0.299 0.00884 -0.106 (0.94) (-0.57) (0.03) (-0.35) industry effects Yes Yes Yes Yes r2 0.00420 0.00420 0.00420 0.00420 N 211246 211246 211246 211246 t statistics in parentheses ∗ p< 0.1, ∗∗ p< 0.05, ∗∗∗ p< 0.01 t 0 = 2000, t 1 = 2002. dgap =gap t0 − gap t1 , dtariff =tariff t0 − tariff t1 equation: dln(import i )=α fix +β 1 dgap i +β 2 dτ i +β 3 τ i +ϵ i (2.25) The dependent variable in this regression is the change in the natural logarithm of import value for each firm. The three trade policy variables included are the reduction in uncertainty (dgap), the change in tariff rate ( dτ ), and the average tariff rate ( τ ). The regression also controls for firm fixed effects ( α fix ). The aim of this study is to examine the relationship between trade policy uncertainty and the change in log value of import for a firm i. The main variable of interest is dgap i , which measures the reduction in trade policy uncertainty. Based on our model’s prediction that reductions in trade policy uncertainty lead to an increase in import value, I expect β 1 to be positive. Another important variable is dτ , which captures changes in applied tariffs. Positive values of this measure indicate a decrease in applied tariffs, and I predict β 2 to be 67 Table 2.3: Import Growth and Trade Policy Uncertainty Firm-Year Level [b] (1) (2) (3) (4) dgap 0.0279 ∗∗∗ 0.0172 ∗∗∗ 0.0225 ∗∗∗ 0.0224 ∗∗∗ (29.65) (17.81) (22.72) (22.59) d tariff 0.0490 ∗∗∗ 0.0423 ∗∗∗ 0.0421 ∗∗∗ (39.11) (32.94) (32.79) tariff -0.0330 ∗∗∗ -0.0335 ∗∗∗ (-21.05) (-21.36) gap -0.195 ∗∗∗ (-5.33) cons -0.653 ∗∗∗ -0.694 ∗∗∗ -0.479 ∗∗∗ -0.460 ∗∗∗ (-31.52) (-34.08) (-21.14) (-20.06) industry effects No No No No r2 0.0222 0.0595 0.0701 0.0708 N 38653 38653 38653 38653 t statistics in parentheses ∗ p< 0.1, ∗∗ p< 0.05, ∗∗∗ p< 0.01 t 0 = 2000, t 1 = 2002. dgap =gap t0 − gap t1 , d tariff =tariff t0 − tariff t1 68 positive, as reductions in applied tariffs have a similar effect to decreases in trade policy uncertainty. Finally, I include the average tariff level τ and gap. To ensure the robustness of our results, I test both product-level data and firm-level data. The MFN tariffs were set decades ago and remained stable over recent decades, they were applied to all members in the world and should have been exogenous. The change in trade policy uncertainty is tiny after China’s accession to the WTO, so I compare the year prior to China’s accession to the WTO (2000) and the year after (2002). Table 2.2 and Table 2.3 present the results of our regression analysis. In Table 2.2, I regress the log growth of import values in firm-product level data on our trade policy vari- ables: uncertainty reduction, applied tariff reduction, average tariff, and current uncertainty level. Table 2.3 shows the firm-level results. I find that uncertainty reduction has a positive and significant effect on import growth. The firm-level results are more significant and fit our model better than the firm-product level data, as technological changes, varying final goods assembly, and market entry or exit affect the continuity of firm-product-year level data. Table 2.3 shows that a 1 percent reduction in uncertainty can cause a 2.8% increase in import growth. Additionally, the change in uncertainty and the change in average tariff have a positive and significant effect on import growth, while the current uncertainty level does not, as it did not change significantly after China’s accession to the WTO. I provide regression results for other years in the Appendix. 2.3.3 Intensity Proposition4indicatesthatfirmswithlessintensitysourcingstrategymaybemoreresilience to shocks. To begin, a regression is run to verify the relationship between HHI and produc- tivity. The HHI index of a firm is regressed against its productivity, as measured by Brandt et al. (2017), in the following equation: HHI i,t =α fix +βlnTFP i,t +ϵ i,t (2.26) 69 Here, β is expected to be negative, as per Proposition 4. The results are presented in Table 2.4. Columns (1) and (2) use the full sample of importers and non-importers, while Columns (3) and (4) focus on importers alone. All columns control for year, ownership, industry, and region fixed effects. Across all specifications, β is negative and highly signifi- cant, indicatingthathighproductivityfirmsaremorediversifiedintheirsourcingstrategies, consistent with the model. Table 2.4: Firm productivity and diversification of sourcing [b] All Firms Importers (1) (2) (3) (4) ln TFP -0.00862 ∗∗∗ -0.000834 ∗∗∗ -0.0797 ∗∗∗ -0.0211 ∗∗∗ (0.000183) (0.000118) (0.00121) (0.00157) Time Effects Yes Yes Yes Yes Industry FE Yes Yes Yes Yes Firm FE No Yes No Yes R 2 0.372 0.0860 0.150 0.595 N 1224884 1224884 166133 166133 ∗ p< 0.1, ∗∗ p< 0.05, ∗∗∗ p< 0.01 Firm’s financial statement, name, phone, postcode, import or export value are obtained from Chinese National Bureau of Statistics, 2000-2006 Firm’s import value, date, product, and original country are obtained from China Customs Data, 2000-2006. HHI i = P χ 2 ij . χ ij is the propotion of import value through route j in time t to total import value in time t. TFP it = (q it − q i,t− 1 )− ¯ S it (l it − l i,t− 1 )− (1− ¯ S t )(k it − k i,t− 1 ). My baseline regression to estimates the relationship between firms’ sourcing diversity and uncertainty is: y =α fix +β 1 HHI i,t +β 2 ∗ Shock t +β 3 HHI i,t ∗ Shock t +ϵ i,t (2.27) The dependent variable y is either the log import value (lnImport i,t ) or the log import value growth(dlnImport it ). IusetheHHIindextomeasurethefirm’sintensityandconcentration levelinsourcing. Theshockisdefinedintwoways: (1)shockequals0before2002andequals 70 1 after 2002; (2) shock equals dgap. To capture the heterogeneous pass-through, I add an interaction term between the shock and the sourcing intensity. The main coefficient of interest is β 3 , which is expected to be negative if sourcing decisions are complementary. I control for both time fixed effects and firm fixed effects. Table 2.5: Resilience of firms to WTO accession shock [b] WTO accession TPU shock (1) (2) (3) (4) (5) (6) lnImport lnImport lnImport dlnImport dlnImport dlnImport HHI -5.078 ∗∗∗ -5.078 ∗∗∗ -5.051 ∗∗∗ -0.741 ∗∗∗ -0.741 ∗∗∗ -0.730 ∗∗∗ (-465.10) (-465.10) (-210.45) (-80.39) (-80.39) (-75.19) shock 0.120 ∗∗∗ 0.146 ∗∗∗ (9.33) (6.70) HHIShock -0.0341 ∗∗∗ (-1.26) dgap 0.0265 ∗∗∗ 0.0551 ∗∗∗ (3.40) (5.23) HHITPU -0.0381 ∗∗∗ (-3.50) cons 15.66 ∗∗∗ 15.66 ∗∗∗ 15.64 ∗∗∗ 0.496 ∗∗∗ 0.499 ∗∗∗ 0.491 ∗∗∗ (1234.14) (1234.14) (836.83) (52.65) (52.71) (51.00) time effects Yes Yes Yes Yes Yes Yes r2 0.231 0.231 0.231 0.0164 0.0165 0.0165 N 633091 633091 633091 396779 396779 396779 t statistics in parentheses ∗ p< 0.1, ∗∗ p< 0.05, ∗∗∗ p< 0.01 ∗ shock=1 if before 2002, shock=1 if after 2002. ∗ HHI i = P χ 2 ij . χ ij is the propotion of import value through route j in time t to total import value in time t. Table2.5presentstheregressionresults. Asexpected,thefirm’sconcentrationindicator HHI has a negative and highly significant coefficient in all columns. The main effect of the shocks is positive and highly significant, as shown in columns (2) and (5). Following China’s accession to WTO, import values increased by 12%, and after the reduction in uncertainty, they increased by 2.6%. The effect is larger when testing the WTO accession shock, as this 71 containsnotonlyuncertaintyreductionbutalsoatariffshock. Theinteractiontermbetween the sourcing intensity and the shock is introduced in columns (3) and (6). When the firm is highly concentrated (HHI=1), the positive effect brought by the shock is reduced. For example,theimportvalueincreasesby14.6%ifthefirmdoesnotconcentrateatall(HHI=0), but only increases by 11.19% if the firm is highly concentrated (HHI=1) following the WTO accession shock. The uncertainty reduction leads to an import growth improvement of 5.5% if the firm is less concentrated (HHI=0), and only 1.7% if the firm is highly concentrated (HHI=1). These findings suggest that a firm with a diversified sourcing strategy can benefit more from a positive shock, while a less intense sourcing strategy may be more resilient to negative shocks. However, due to data limitations, I am unable to explore the latter hypothesis in this study. 2.4 Conclusion Inthisstudy,Ihaveanalyzedtheeffectsoftradepolicyuncertaintyonprofitandimportvalue using a multi-country model with heterogeneous firms. The findings suggest that increased uncertainty has a negative impact on both profit and import value. Furthermore, I have shown that firms with more diverse trade routes are better equipped to handle uncertainty. To demonstrate the practical implications of this result, I have applied the model to China’s WTOaccessionandestimatedtheeffectsofdiversifyingsourcingroutes. WhiletheCovid-19 pandemichasbroughtunprecedentedchallengestoglobaltrade,thisstudyprovidesvaluable insights on how uncertainty affects firms and highlights the importance of diversifying trade routes. Future studies may also consider the extensive margin, as uncertainty may lead to changes in trade routes and restructuring of global trade patterns. The Covid-19 pandemic has had a significant impact on the global economy and dis- rupted supply chains worldwide. According to the World Trade Organization (WTO), the pandemic has resulted in a global trade decline of up to 32% in 2020, which is more than 72 double the magnitude of the ”Great Trade Collapse” during the 2008-2009 global financial crisis (WTO, 2020). While the pandemic is still ongoing, there are valuable lessons to be learned from past crises to shed light on the upcoming trade slump and eventual recovery. In this context, my research examines the effects of trade policy uncertainty and identifies which types of firms are less impacted. By using China’s WTO entry data as an example of a positive shock, I demonstrate that uncertainty shock reduces import purchases through intensive margin and that firms could benefit from diversifying their input sourcing in times of uncertainty. Although we do not have complete data on the impact of the Covid-19 pan- demic, my model and example can provide insights into the possible effects of uncertainty shock on global trade. This paper highlights the impact of trade policy uncertainty on import values and the importance of sourcing diversity for firms’ resilience. While the focus is on the intensive margin,theextensivemarginshouldalsobeconsideredinfutureresearch. Whenuncertainty is high, importers may choose to abandon certain trade routes or vary the fixed cost f ei depending on the intensity and substitutability rate. For instance, the trade war between China and the U.S.increased tradepolicy uncertainty, leadingsome firmsto start importing from other East Asian countries. While these firms have not abandoned trade with China, the possibility of doing so in the future could cause a structural change in Asia. 73 Chapter 3 An Anatomy of Exports Around Economic Downturns 1 This paper empirically characterizes the mechanics of export adjustment during eco- nomicdownturnsusingproduct-levelbilateraltradeflowdatafrom1970to2019. Wedevelop a uniform definition of domestic economic construction shocks across 170 countries and use the local projection method to investigate trade adjustments based on different time win- dows. Economiccontractionssignificantlyandnegativelyaffectbothextensiveandintensive margins. The extensive and intensive margins recover three periods after one year of GDP drops; however, the effects are more persistent with cumulative two-year GDP decreases. Furthermore, the adjusted export value is mainly driven by the changes in intensive margin for existing varieties. Nonetheless, our results also indicate that exporters start to re-enter the market in 3 periods after the shock. 1 Co-authored with Samuel Pienknagura 74 3.1 Introduction Economic contractions can significantly impact international trade, causing declines in both the intensive and extensive margins of trade. Policymakers seeking to mitigate the effects of economic contractions on trade can gain valuable insights from understanding how firms adjust their trade patterns during downturns. This paper investigates the effects of eco- nomic contractions on trade dynamics using a panel of bilateral trade data. Specifically, we examine how economic contractions affect the extensive and intensive margins of trade, as well as the entry and exit of exporting products. Our findings could have important policy implications for countries facing economic downturns. By understanding the behavior of firms during economic contractions, policymakers can design targeted policies that support firms in maintaining their existing market share while also encouraging them to develop and introduce new products. Overall, this paper provides a valuable contribution to the literature on the effects of economic contractions on international trade. Toachieveourobjectives,weuseapanelofbilateraltradedatafromtheUnitedNations Comtrade database covering the period 1970-2019. The data are organized by 4-digit SITC, Revision 1, which covers over 184 countries and around 1000 categories. This dataset is the most disaggregated publicly available dataset for bilateral trade flows for a large number of years and country pairs, constructed on a consistent basis, which is essential for our analysis. To ensure the reliability and consistency of the data, we limit our sample to 170 export countries that have continuously reported their export trade flows since 1970. This approach enables us to avoid potential biases that may arise from missing or inconsistent data. The data also allow us to decompose the extensive and intensive margins of trade, providing insights into the dynamics of trade flows following economic contractions. We examine the impact of economic contractions on international trade by analyzing the extensive and intensive margins of exports. The extensive margin is decomposed into delayed entry and accelerated exit, while the intensive margin is decomposed into changes 75 in the export value of existing products (stay value) and the adjusted export value for newly introduced or terminated products (entry/exit value). Our results suggest that economic contractions have a substantial negative impact on both the extensive and intensive margins of trade. The extensive margin decline is primar- ily driven by delayed entry of new product varieties, while the acceleration in exit is less significant. The intensive margin decline is mainly due to a reduction in the export value of existing products (stay value), rather than an increase in the abandonment of varieties. Firms tend to reduce their export value while still maintaining some exports, rather than abandoningtheirexistingproductlines. Inaddition,thedeclineintheadjustedexportvalue is driven by a reduction in the introduction of new products rather than an acceleration in the exit of existing products. These findings have important policy implications, as they suggest that firms may be hesitant to invest in new products during economic contractions, potentially leading to a long-term decline in innovation and competitiveness. Policymakers could consider providing support for innovation and RD activities to encourage firms to invest in new products and maintain their competitiveness. Additionally, policymakers may want to consider providing support for firms to maintain their existing market share during economic downturns, such as through export promotion programs or financial assistance. This paper relates to many literatures. First, it relates to the literature that empiri- cally characterizes the margins of trade adjustment and is consistent with the findings of Bernardetal.(2009)andBernardetal.(2011),whouseU.S.datatodocumentthatthefirm and aggregate product extensive margins are small while the sub-extensive margin is large. Hummels and Klenow (2005) and Baier et al. (2014) introduced and studied the economic integration agreements and the margins of international trade. Distinct from their analysis, we focus on a dramatically larger trade adjustment episode using the Argentine experience and specifically evaluate the implications of these findings for productivity. Second, this paper relates to literatures studying trade imbalances and crises. For 76 example, Baldwin (2011) found that the 2008 financial crisis led to a 10% decline in world trade in 2009, primarily driven by a decrease in demand. Similarly, Chen and Novy (2011) found that the global financial crisis led to a significant reduction in bilateral trade flows between countries, particularly for countries heavily involved in global value chains. Rose (2004) found that countries more open to trade were less vulnerable to crises, as they were able to diversify their export markets and benefit from increased competition. Similarly, Gourinchas and Jeanne (2013) found that countries with more flexible exchange rates were more resilient to crises, as they were able to adjust to changes in trade flows more easily. Our study focuses on the adjustments of exporters with domestic economic contraction shocks. Unlikeglobalorimportercountrydownturnsthatmainlydisruptdemand, economic downturns in export-oriented countries tend to have a more significant impact on the supply side. Export-orientedfirmsmayneedtoadjusttheirexportactivitiesduetoalackoflaboror decreasedproductivity,leadingtodisruptionsinbothtradeextensiveandintensivemargins. 3.2 Data 3.2.1 Economic Contractions The purpose of this study is to examine the impact of economic downturns on trade im- balances in export-oriented countries. Unlike global or importer country downturns that mainly disrupt demand, economic downturns in export-oriented countries tend to have a more significant impact on the supply side. Export-oriented firms may need to adjust their export activities due to a lack of labor or decreased productivity, leading to disruptions in both trade extensive and intensive margins. To identify the GDP growth rate and analyze the impact of economic downturns on trade imbalances, we rely on the World Development Indicators, a comprehensive and inter- nationally comparable compilation of relevant statistics on global development and poverty reduction. The database contains over 1,400 time series indicators for 217 economies and 77 more than 40 country groups, with some indicators providing data spanning more than five decades. Therefore, the database offers a reliable source to assess the influence of economic downturns in export-oriented countries on trade imbalances, which could inform policymak- ersandstakeholdersseekingtodesigneffectiveinterventionstomitigatetheadverseimpacts of such downturns on both the domestic and global economies. Weintroducetwofilterstoidentifyepisodesofeconomicdownturnsinouranalysis. The first filter is based on the one-year GDP drop and is defined as follows: dummy 1 =        1, if g 1 t ≤ median(g 1 t |g 1 t ≤ 0) & GDP t+1 ≤ GDP t− 1 0, otherwise (3.1) where g 1 t represents the one-year GDP growth rate, which is calculated using the formula: g 1 t = GDP t − GDP t− 1 GDP t− 1 × 100 (3.2) where GDP t denotes the gross domestic product in year t. The first filter identifies episodes ofeconomicdownturnsbasedontheone-yearGDPdrop,whichisacommonapproachinthe literature. By comparing one-year GDP drops against the median of all one-year negative GDP growth rates, we can identify economic downturn episodes that are more severe than typicalfluctuationsingrowthrates. Thisallowsustocaptureshort-termdisruptionsintrade extensive and intensive margins, which may have a significant impact on trade imbalances in export-oriented countries. In addition to the previous filter, we propose a second filter that captures sustained economic downturns based on the two-year cumulative GDP drop: dummy 2 =        1, if g 1 t <0 & g 2 t ≤ median(g 2 t |g 2 t ≤ 0) 0, Otherwise (3.3) where g 2 t represents the two-year cumulative GDP growth rate, which we calculate using the 78 formula: g 2 t = GDP t − GDP t− 2 GDP t− 2 × 100 (3.4) Thesecondfilter,basedonthetwo-yearcumulativeGDPdrop,providesamorenuanced assessment of the severity and duration of economic downturns. By identifying economic contractionshocksthatpersistforatleasttwoyears,wecancapturelonger-termdisruptions in trade extensive and intensive margins. This is particularly relevant for export-oriented countries, where trade imbalances can have long-lasting impacts on the domestic and global economies. Together, these two filters provide a comprehensive approach to identifying economic downturns and their impact on trade imbalances in export-oriented countries. By analyzing the effects of economic downturns on both extensive and intensive trade margins, we can gain a more complete understanding of the mechanisms through which economic downturns affect trade imbalances and the policies that can be implemented to mitigate their adverse effects. To identify economic downturn episodes, the study proposes two filters based on GDP growthrates. Thefirstfilteridentifiesepisodesbasedontheone-yearGDPdropcomparedto the median of all one-year negative GDP growth rates. The second filter identifies sustained economic downturns based on the two-year cumulative GDP drop. By analyzing the effects of economic downturns on both extensive and intensive trade margins, the study provides a comprehensive approach to understanding their impact on trade imbalances. 3.2.2 Trade Margins Inthisstudy,weadoptedthedefinitionofextensiveandintensivemarginsoftradeasdefined by Hummels and Klenow (2005), who introduced a method to decompose these margins for a large set of countries’ bilateral trade flows using publicly available disaggregate trade 79 data. Hummels’ method is considered a significant contribution to the literature, providing a tractable and transparent way to decompose trade margins, and has been widely cited in subsequent studies. Let X ij,t denote the value of country i’s exports to the country j in year t. Following HK, the extensive margin of goods exported from i to j in any year t is defined as: EM ij,t = P m∈M ij,t X m Wj,t P m∈M Wj,t X m Wj,t (3.5) where X m Wj,t represents the value of country j’s imports from the world in product m in year t. M Wj,t is the set of all products exported by the world to country j in year t, and M ij,t is the subset of all products exported from country i to the country j in year t. Therefore, EM ij,t is a measure of the fraction of all products that are exported from country i to the country j in year t, weighted by the importance of each product in world exports to the country j in that same year. In addition to the extensive margin, HK also defines the intensive margin of goods exported from country i to the country j in year t, given by: IM ij,t = P m∈M ij,t X m ij,t P m∈M ij,t X m Wj,t (3.6) In this equation, X m ij,t represents the value of exports from country i to the country j in product m in year t. Thus, IM ij,t measures the market share of country i in the country j’s imports from the world within the set of products that country i exports to the country j in year t. An important property of the HK decomposition methodology is that the product of theextensiveandintensivemarginsequalstheratioofexportsfromcountry itothecountry j relative to the country j’s total imports, as shown below: Share ij,t = X ij,t X j,t =EM ijt, IM ij,t (3.7) 80 where X jt denotes country j’s imports from the world. Taking the natural logarithm of Eq. 3.7 and applying some algebra yields: lnX ij,t =lnEM ij,t +lnIM ijt, +lnX j,t (3.8) Thus, the HK decomposition methodology allows us to linearly decompose the logarithm of the value of the trade flow from country i to the country j in any given year t into (logs of) an extensive margin, an intensive margin, and the value of country j’s imports from the world. Our empirical analysis relies on the use of disaggregated bilateral trade flows as the key variable. To construct our dataset, we use the annual bilateral trade flows for the years 1970-2019 from the UN Comtrade trade data set 2 . The data are organized by 4-digit SITC, Revision 1, covering trade flows reported by over 184 countries and around 1000 categories. This dataset is the most disaggregated publicly available dataset for bilateral trade flows for a large number of years and country pairs, constructed on a consistent basis, which is essentialforouranalysis. PreviousstudiessuchasHillberryandMcDaniel(2005),Kehoeand Ruhl (2009), and Baier et al. (2014) have also used this 4-digit SITC data in their analysis. However, one concern is that the level of disaggregation may bias our results towards the extensivemargin. Nevertheless, comparedtohigherlevelsofdisaggregation, thisdatacovers the longest period and includes the most countries, making it an important resource for our analysis. While our study’s decomposition of the extensive and intensive margins is based on Hillberry and McDaniel (2005), it is important to compare our data set to theirs. HK utilized data from the Harmonized System 6-digit classification code in 1995, which resulted in 5017 goods categories - five times the number in our data set. In examining the levels of extensive margins and the correlations between the extensive margin and factors influencing it(suchasGDP),HKfoundtwointerestingresults. Firstly,asdatabecamemoreaggregated, 2 https://comtrade.un.org 81 extensive margins were able to explain less. However, secondly, even at the 4-digit level, a significant 54% of the elasticity of exports could still be explained by the extensive margin. This suggests that using 4-digit categories may not result in severe intensive margin bias. To ensure that we have consistent and reliable data for our analysis, we focus on a sample of 170 export countries that have continuously reported their export trade flows since 1970. By limiting our sample to countries with continuous reporting, we can avoid potential biases that may arise from missing or inconsistent data. 3.3 Quantifying the Effects of Economic Contractions 3.3.1 The Effects on Extensive and Intensive Margins We begin by investigating whether the shocks from economic contractions impact the bilat- eral export value and the number of export varieties. The bilateral export value refers to the total export value for country i in year t, which can be calculated as: Export Value ij,t =log X m∈M ij,t X m ij,t (3.9) Similarly, the number of export varieties is defined as the log of the number of unique products exported from country i to the country j in year t, which can be calculated as: Export Variety ij,t =log X m 1(m∈Mij,t) ! (3.10) Then we use the following equation to quantify the effects of economic contraction shocks: Y ij,t =α i +α j +α t +βShock i,t +Γ X ′ ij,t− 1+ϵij,t (3.11) where Y ij,t represents the log of the bilateral export value(intensive margin) or the number 82 Table 3.1: Effects of Economic Contraction Shocks One Year GDP Decrease Cumulative Two Year GDP Decreas (1) (2) (3) (4) Export Value Export Variety Export Value Export Variety Economic Contraction Shock -0.0365*** -0.0821*** (-2.97) (-15.54) Economic Contraction Shock -0.0930*** -0.0829*** (-6.60) (-13.67) r2 0.845 0.884 0.845 0.884 N 582299 582299 582299 582299 t statistics in parentheses * p< 0.1, ** p< 0.05, *** p< 0.01 ofexportvarieties(extensivemargin). shock i,t denotesthedummyvariablefortheeconomic shock in the export country i in year t. Additionally, a set of control variables are included in X ij,t , which consists of the GDP level and GDP growth rate for both export and import countries, the distance between the two countries, an index of common language, and the lagged total export value for each export country. To account for fixed effects, α i , α j , and α t represent the exporter, importer, and time-fixed effects, respectively. Table 3.1 highlights the significant negative impact that economic contraction shocks have on both the export value and the number of export varieties for both the one-year and cumulative two-year GDP decrease filters. Such shocks could lead to a contraction in the exportactivitiesoffirms, resultinginadeclineintheexportvalueofeachcontinuingvariety and a reduction in the number of exported varieties. This could occur due to the decline in demandfromcustomers, leadingtolowerpricesorreducedquantitiessold. Additionally, ex- portersmaydiscontinuetheproductionofcertaingoodsinresponsetoeconomiccontraction shocks, resulting in fewer varieties of goods being exported. To gain insights into the impact of economic contraction shocks on trade margins, we examine the effects on both extensive and intensive margins of exports over different time horizons. Thisanalysishelpstounderstandthedynamicsoftheeffectsontrademarginsfol- lowingeconomiccontractions, includingwhethertheextensiveorintensivemarginsaremore 83 affected by the shocks, and whether the effects are short-lived or persistent. This can also help policymakers in designing policies to support exporters during economic contractions and promoting trade recovery. For example, if the extensive margin is more affected and takes longer to recover, policymakers may need to focus on measures to help firms maintain and diversify their export varieties during economic downturns. On the other hand, if the intensive margin is more affected and recovers more slowly, policies that stimulate demand in import countries could be more effective in promoting export recovery. The benchmark specification at yearly frequency is as follows: Y ij,t+h − Y ij,t− 1 =α h i +α h j +α h t +β h Shock i,t +δX ij,t +ϵ ij,t+h (3.12) where Y ij,t+h denotes the bilateral export share, extensive and intensive margin at horizon h after the occurrence of the shock. α h i , α h j , and α h t represent the exporter, importer, and time-fixedeffects,respectively,atthehorizon h. Shock i,t isthedummyvariableforeconomic contractionshocksintheexportcountry iinyeart, andβ h capturestheimpactoftheshock on the export value or varieties at horizon h. Additionally, X ij,t includes a set of control variables such as the GDP level and GDP growth rate for both export and import countries, distance between the two countries, an index of common language, and the lagged total export value for each export country. The error term is denoted by ϵ ij,t+h . The results from Figure 3.1 suggest that the decline in bilateral export shares can be primarily attributed to the reduction in the extensive margin following economic contrac- tion shocks. Additionally, the recovery of both the extensive and intensive margins occurs approximately three periods after the shock for the one-year GDP drop scenario. This sug- gests that the effects of economic contractions on trade margins are relatively short-lived, at least in the context of a one-year GDP decline. However, for the cumulative two-year GDP decrease scenario, the decline in export activity is more severe, and the recovery is 84 Figure 3.1: The Effects on Extensive and Intensive Margins 85 slower. While the extensive margin is initially more affected, the impact on this margin becomes moderate after three periods. These findings have important policy implications, as policymakers can utilize them to design appropriate policies that support exporters dur- ing economic downturns and promote trade recovery. This could include targeted financial assistance programs, tax incentives, and trade promotion initiatives aimed at helping firms to expand their export portfolios and reach new markets. 3.3.2 Decomposed Intensive Margin In this section, we aim to examine the changes in the intensive margin of trade following economic contractions. The intensive margin can change due to various factors, including changes in the export value of existing products, the abandonment of existing products, or the introduction of new products. To investigate these changes, we decompose the intensive marginintotwocomponents: theexportvalueforproductsthatwerealreadybeingexported before the shock (stay value), and the export value for newly introduced products (entry value) and terminated products (exit value). We measure the change in the export value of products that continue to be exported (stay value) by comparing the log difference of the total export value of products that were exported in the period prior to the shock (t-1) and still to be exported h periods after the shock (t+h): ∆ h Stay Value ij,t =log X m∈M ij,t− 1 ,m∈M ij,t+h X m ij,t+h − log   X m∈M ij,t− 1 ,m∈M ij,t+h X m ij,t− 1   (3.13) Moreover, the adjusted export value for newly introduced products (entry value) or 86 abandoned products (exit value) is defined as follows: Adjust Value h ij,t =log X m̸∈M ij,t− 1 ,m∈M ij,t+h X m ij,t+h | {z } Entry Valueij,t+h − log   X m∈M ij,t− 1 ,m̸∈M ij,t+h X m ij,t− 1   | {z } Exit Valueij,t− 1 (3.14) where the entry value represents the log of the total export value of newly introduced prod- ucts in period t+h that were not exported by country i to the country j in period t− 1. Similarly, the exit value represents the log of the total export value of products that were exported by country i to the country j in period t− 1 but not in period t+h. By using a similar LP regression as in Eq.3.12, Figure 3.2 illustrates that changes in the export value are mainly due to a decline in the stayed value, which suggests that firms tend to reduce the value of their existing product lines while still maintaining some exports. When faced with domestic economic contractions, firms tend to reduce their export value rather than abandon the varieties they produce. This can be attributed to the fact that abandoning a variety may result in a permanent loss of market share or reputation, whereas reducingexportsallowsfirmstomaintaintheirpresenceinthemarketandpotentiallyrecover lost demand when economic conditions improve. Furthermore, if the fixed costs associated with entering the market are high, firms may be more inclined to reduce the quantity of the variety exported rather than entirely discontinuing its production and export. Furthermore, the changes in the export value are also influenced by a decline in the adjusted value. Figure 3.3 shows that the drops in adjust value are primarily driven by a reduction in the introduction of new products rather than an acceleration in the exit of existing products. This suggests that firms are more hesitant to enter new product markets when facing economic downturns. Introducing new products requires high fixed costs, which may be difficult to recoup if the economic conditions are unfavorable. Moreover, new prod- ucts may face increased competition and may require significant marketing and advertising expenditurestogainmarketshare, whichcanbechallengingduringadownturn. Asaresult, 87 Figure 3.2: The Effects on Stay Value and Adjust Value firms may be more hesitant to enter new markets during periods of economic contraction, leading to a reduction in the introduction of new products and a decline in the adjusted value. 3.3.3 Decomposed Extensive Margin Inthissection,weinvestigatethereasonsforthesignificantdeclineinthenumberofproducts exportedfromcountryitocountryjfollowingeconomiccontractions. Wemeasuretheeffect of the shock on the number of exported product varieties using the equation: Export Varietiesij,t=log X m 1(m∈Mij,t) ! (3.15) Figure 3.4 illustrates that the economic contraction shock in the export country has a substantial negative impact on the number of bilateral exported products. We observe that the negative effects are even more severe for cumulative two-year GDP decreases. In both scenarios, firms either pause introducing new products or abandon vulnerable existing 88 Figure 3.3: The Effects on Entry Value and Exit Value products due to the domestic country’s economic recession. However, the market recovers after three periods, and firms begin to reenter the market. The recovery of firms in the market after an economic contraction shock is a complex process that several factors can influence. One possible explanation is that, following a recession, firms may need time to adapt their business strategies and operations to suit the changing economic conditions better. This adjustment period may involve restructuring, downsizing, or reallocating resources, which can take time to implement. In addition, firms may need to develop new products or services that are better suited to the post-recession market. Another essential factor is the gradual recovery of consumer demand for products and services. As the economy begins to recover, consumers may gradually resume their spending habits, creating new opportunities for firms to sell their goods and expand their operations. This increased demand can make it easier for struggling firms to reenter the market and regain their footing. 89 Figure 3.4: Number of Exported Products To understand whether the decrease in the number of exported products is due to a pause in introducing new products or an acceleration in abandoning vulnerable varieties, we usemeasuresofthenumberofenteredandexitedvarietiestodeterminewhetherthisdecline is driven by delayed entry or accelerated exit. We compare the number of new exporting products entering and exiting before the shock and after h years of the shock. To measure the number of entered varieties, we use the following equation: Entered Varietiesij,t h =log X m 1(m̸∈Mij,t− 1,m∈M ij,t+h ) ! (3.16) Similarly, we define the number of exited varieties using the following equation: Exited Varietiesij,t h =log X m 1(m∈Mij,t− 1,m̸∈M ij,t+h ) ! (3.17) These measures enable us to determine whether the decline in the extensive margin is driven by a decrease in the number of new varieties entering the market or an increase in the number of existing varieties exiting the market. By analyzing the differences in entered and exited varieties before and after the economic contraction shock, we can gain insights into the underlying causes of the decline in the number of exported products. Figure 3.5 indicates that the decrease in entry is more significant than the acceleration in exit, which is consistent with the findings presented in Figure 3.3. Exporting firms tend 90 Figure 3.5: The Effects on Number of Entered and Exited Varieties to delay introducing new products following a GDP drop, as the lower demand and revenue make it less attractive to invest in new product development and export. In the face of economic contraction, exporters prefer to focus on maintaining their existing market share and reducing their costs rather than taking risks with new product introductions. Moreover, firms tend to adopt a ”wait and see” approach with moderate shocks, as they would face higher opportunity costs to re-enter the market. However, in the case of severe shocks, such as a cumulative two-year GDP drop, firms may abandon more vulnerable varieties immediately, as the opportunity cost of waiting for a recovery in demand for those products may be too high. Notably, both delays in entry and accelerations in exit recover after three periods of the shock. As shown in Figure 3.5, increased entry of product varieties is the primary driver of the recovery of the extensive margin. Firms tend to increase their investment and re-enter the market as the economy recovers, as delaying entry for too long can lead to lost market share and missed opportunities. In this context, policymakers can play an important role in 91 supporting firms during economic downturns. For instance, policies that stimulate demand for existing products can encourage firms to maintain their market share and reduce the need for accelerated exit. Policymakers can also provide support for firms to weather the shock, such as access to finance, training programs, or other forms of assistance that help firms maintain their capabilities and resources. These measures can help firms to re-enter the market more easily once the economy recovers, which can promote a faster and more sustainable recovery. 3.4 Policy Implication and Conclusion 3.4.1 Conclusion In this paper, we investigate the effects of economic contractions on trade dynamics using a panel of bilateral trade data. Our results show that economic contractions have a significant impact on trade, leading to a decline in both the intensive and extensive margins. We find that the decline in the intensive margin is largely driven by a reduction in the stayed value, suggesting that firms tend to reduce the quantity of their existing product lines while still maintaining some exports. Additionally, we observe that the decline in the adjusted value is primarily driven by a reduction in the introduction of new products, indicating that firms are more hesitant to enter new product markets when facing economic downturns. Furthermore, our analysis of the extensive margin reveals that economic contractions lead to delayed entry and accelerated exit of exporting products. Firms tend to delay intro- ducing new products compared to pre-shock levels, preferring to focus on maintaining their existing market share and reducing costs rather than taking risks with new product intro- ductions. Ontheotherhand, theyexitthemarketmorefrequentlyastheshockdeepensand persists, with the opportunity cost of waiting for a recovery in demand for those products becoming too high. 92 Our findings have important policy implications for countries experiencing economic contractions. Governments should consider policies that encourage firms to maintain their existing market share and reduce costs during a downturn while also providing support for firms to develop and introduce new products. Additionally, policies that reduce the fixed costsassociatedwithenteringnewmarketsmaybeparticularlyeffectiveinencouragingfirms to enter new product markets during economic contractions. 93 Bibliography Alessandria, G. and Choi, H. (2007). Do sunk costs of exporting matter for net export dynamics? Quarterly Journal of Economics, 122(1). Amiti, M. and Konings, J. (2007). Trade liberalization, intermediate inputs, and productiv- ity: Evidence from Indonesia. American Economic Review, 97(5). Andreasen, M. M., Fern´ andez-Villaverde, J., and Rubio-Ram´ ırez, J. F. (2018). The pruned state-space system for non-linear DSGE models: Theory and empirical applications. Re- view of Economic Studies, 85(1). Antr` as, P., Fort, T. C., and Tintelnot, F. (2017). The margins of global sourcing: Theory and evidence from US firms. American Economic Review, 107(9). Backus, D. K., Kehoe, P. J., and Kydland, F. E. (1994). Dynamics of the trade balance and the terms of trade: the J-curve? American Economic Review, 84(1). Baier, S. L., Bergstrand, J. H., and Feng, M. (2014). Economic integration agreements and the margins of international trade. Journal of International Economics, 93(2). Baker, S. R., Bloom, N., and Davis, S. J. (2016). Measuring economic policy uncertainty. Quarterly Journal of Economics, 131(4). Baldwin,R.(2011). TheGreatTradeCollapse: Causes,ConsequencesandProspects. Centre for Economic Policy Research. 94 Baldwin, R.andKrugman, P.(1989). Persistenttradeeffectsoflargeexchangerateshocks*. Quarterly Journal of Economics, 104(4). Bellemare, M. F. and Wichman, C. J. (2020). Elasticities and the Inverse Hyperbolic Sine Transformation. Oxford Bulletin of Economics and Statistics, 82(1). Bems, R., Johnson, R. C., and Yi, K. M. (2013). The great trade collapse. Benguria, F., Choi, J., Swenson, D. L., and Xu, M. J. (2022). Anxiety or pain? The impact of tariffs and uncertainty on Chinese firms in the trade war. Journal of International Economics, 137. Bernanke, B. S. (1983). Irreversibility, uncertainty, and cyclical investment. Quarterly Journal of Economics, 98(1). Bernard, A. B., Jensen, J. B., Redding, S. J., and Schott, P. K. (2009). The margins of US trade. In American Economic Review, volume 99. Bernard, A. B., Jensen, J. B., Redding, S. J., and Schott, P. K. (2011). The Margins of U.S. Trade (Long Version). SSRN Electronic Journal. Boehm, C. E., Flaaen, A., and Pandalai-Nayar, N. (2020). Multinationals, Offshoring, and the Decline of U.S. Manufacturing. Journal of International Economics, 127. Born,B.andPfeifer,J.(2014). Riskmatters: Therealeffectsofvolatilityshocks: Comment. American Economic Review, 104(12). Brandt, L., Van Biesebroeck, J., Wang, L., and Zhang, Y. (2017). WTO accession and performance of Chinese manufacturing firms. American Economic Review, 107(9). Brogaard, J. and Detzel, A. (2015). The asset-pricing implications of government economic policy uncertainty. Management Science, 61(1). 95 Caldara, D., Iacoviello, M., Molligo, P., Prestipino, A., and Raffo, A. (2020). The economic effects of trade policy uncertainty. Journal of Monetary Economics, 109. Carre˜ no, I., Dolle, T., Medina, L., and Brandenburger, M. (2020). The implications of the Covid-19 pandemic on trade. European Journal of Risk Regulation, 11(2). Carvalho, F. P. (2017). Pesticides, environment, and food safety. Chen,N.andNovy,D.(2011).Gravity,tradeintegration,andheterogeneityacrossindustries. Journal of International Economics, 85(2). Chor, D.andManova, K.(2012). Offthecliffandback? Creditconditionsandinternational trade during the global financial crisis. Journal of International Economics, 87(1). Das, S., Roberts, M. J., and Tybout, J. R. (2007). Market entry costs, producer heterogene- ity, and export dynamics. Econometrica, 75(3). Datta, A. and Kouliavtsev, M. (2009). NAFTA and the realignment of textile and apparel trade: Trade creation or trade diversion? Review of International Economics, 17(1). Dixit,A.(1989). EntryandExitDecisionsunderUncertainty. Journal of Political Economy, 97(3). Eaton, J. and Kortum, S. (2002). Technology, geography, and trade. Econometrica, 70(5). Esposito, F. (2022). Demand risk and diversification through international trade. Journal of International Economics, 135. Feenstra, R. C. (1994). New product varieties and the measurement of international prices. Feng, L., Li, Z., and Swenson, D. L. (2017). Trade policy uncertainty and exports: Evidence from China’s WTO accession. Journal of International Economics, 106. Fernandes, A. and Tang, H. (2021). How Did the 2003 SARS Epidemic Shape Chinese Trade? SSRN Electronic Journal. 96 Fern´ andez-Villaverde, J., Guerr´ on-Quintana, P., Kuester, K., and Rubio-Ram´ ırez, J. (2015). Fiscal volatility shocks and economic activity. In American Economic Review, volume 105. Fillat, J. L. and Garetto, S. (2015). Risk, returns, and multinational production. Quarterly Journal of Economics, 130(4). Goldberg, P. K., Khandelwal, A. K., Pavcnik, N., and Topalova, P. (2010). Imported inter- mediate inputs and domestic product growth: Evidence from India. Quarterly Journal of Economics, 125(4). Gourinchas, P. O. and Jeanne, O. (2013). Capital flows to developing countries: The allo- cation puzzle. Review of Economic Studies, 80(4). Gulen, H. and Ion, M. (2016). Policy uncertainty and corporate investment. Halpern, L., Koren, M., and Szeidl, A. (2015). Imported inputs and productivity. American Economic Review, 105(12). Handley,K.(2014). Exportingundertradepolicyuncertainty: Theoryandevidence. Journal of International Economics, 94(1). Handley, K. and Lim˜ ao, N. (2015). Trade and investment under policy uncertainty: Theory and firm evidence. American Economic Journal: Economic Policy, 7(4). Handley, K. and Lim˜ ao, N. (2017). Policy uncertainty, trade, and welfare: Theory and evidence for China and the United States. American Economic Review, 107(9). Handley, K. and Lim˜ ao, N. (2018). Trade and investment under policy uncertainty: Theory and firm evidence. In Policy Externalities and International Trade Agreements. Hanson, G. H. (2005). Market potential, increasing returns and geographic concentration. Journal of International Economics, 67(1). 97 Heo, U. (2021). Asia in 2020: The CoviD-19 pandemic and the US-China trade war. Hillberry,R.andMcDaniel,C.A.(2005).ADecompositionofNorthAmericanTradeGrowth since NAFTA. SSRN Electronic Journal. Hummels, D. and Klenow, P. J. (2005). The variety and quality of a nation’s exports. Johnson,R.C.andNoguera,G.(2017). Aportraitoftradeinvalue-addedoverfourdecades. Review of Economics and Statistics, 99(5). Jord` a, andTaylor,A.M.(2016).TheTimeforAusterity: EstimatingtheAverageTreatment Effect of Fiscal Policy. Economic Journal, 126(590). Julio, B. and Yook, Y. (2016). Policy uncertainty, irreversibility, and cross-border flows of capital. Journal of International Economics, 103. Kehoe, T. J. and Ruhl, K. J. (2009). Sudden stops, sectoral reallocations, and the real exchange rate. Journal of Development Economics, 89(2). Melitz, M. J. (2003). The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica, 71(6). Pierce, J. R. and Schott, P. K. (2016). The surprisingly swift decline of US manufacturing employment. American Economic Review, 106(7). Rose, A. K. (2004). Do we really know that the WTO increases trade? American Economic Review, 94(1). Steinberg, J. B. (2019). Brexit and the macroeconomic impact of trade policy uncertainty. Journal of International Economics, 117. Stockman, A. C. and Tesar, L. L. (1995). Tastes and Technology in a Two-Country Model of the Business Cycle: Explaining International Comovements. The American Economic Review, 85(1). 98 Watson, A. (2019). Financial Frictions, the Great Trade Collapse and International Trade over the Business Cycle. Open Economies Review, 30(1). 99 Appendix A. Trade Policy Uncertainty Table A1: Term Sets for Uncertainty Indexes Terms U.S TPU TPU i RowTPU i Economics economy/business √ √ √ Trade Policy import tariffs/ import duty/ import barrier/ WTO/ world trade organization / trade treaty/ trade agreement/ trade policy/ trade act/ √ √ √ Doha round/ Uruguay round/ GATT/ General Agreement on Tariffs and Trade/ dumping/ protectionism/trade barrier/ export subsidies Uncertainty uncertain/uncertainty/ not certain/unsure/ √ √ √ not sure/hard to tell/ unpredictable/unknown Country Australia/ Canada/ China/ Europe/ Country i Country j̸=i Europe/ India/ Japan/ Mexico/ UK/ Vietnam √ √ U.S United States/ U.S √ √ √ The Categorical Data follows Baker et al. (2016). Using similar methods, I develop a trade policy uncertainty index for each country-pair by specifying more restrictive criteria for articles containing at least one keyword of terms about the “Economics”, “Uncertainty”, “Trade Policy”, and “Country”. 100 Table A2: Sample of Country-specific TPU Spike Events (2018-2019) Date Cover Product Target Country Tariff Increase 1/23/2018 Solar Panels Outside United States 30% 1/23/2018 Washing Machines Outside United States 20% 3/1/2018 Steel EU, Canada, Mexico, Australia, Argentina, Brazil, and South Korea 25% 3/1/2018 Aluminum EU, Canada, Mexico, Australia, Argentina, Brazil, and South Korea 10% 3/3/2018 Car Manufactures EU 25%% 3/22/2018 Section 301 Final List China 15-25% 12/23/2018 NAFTA Re-Negotiation Negotiating Objectives for NAFTA. Canada and Mexico 30% 3/1/2019 Automobiles and Parts EU 25% 3/30/2019 All Mexican Imports Mexican 5-15% 12/2/2019 Steel and Aluminum Tariff Withdrawn Argentina, Brazil See more details Trump’s Trade War Timeline: An Up-to-Date Guide Latest U.S. Trade Actions/Tariffs and Other Countries Retaliatory Measures – Updated February 9, 2022 Trump Tariffs - Wikipedia Table A2 presents a sample of country-specific TPU spike events. The data is collected based on newspapers and Wikipedia. 101 Table A3: Summary Statistics of Trade Policy Uncertainty Index (2010-2019) mean p50 sd min max skewness kurtosis Australia 0.36 0.23 0.51 0.00 2.53 2.28 8.68 Canada 0.89 0.35 1.43 0.00 7.13 2.60 9.35 China 2.59 1.15 4.30 0.05 26.12 3.18 13.93 Europe 1.85 1.05 2.28 0.24 11.27 2.49 8.35 India 0.52 0.33 0.62 0.00 2.86 2.03 7.10 Japan 1.11 0.62 1.49 0.00 8.62 2.85 12.18 Latin American 0.54 0.40 0.56 0.00 2.81 1.79 7.27 Mexico 1.14 0.52 1.76 0.00 10.05 2.72 10.81 Southeast Asia 0.56 0.42 0.57 0.00 3.04 1.93 7.51 United Kingdom 1.08 0.58 1.58 0.00 8.17 2.62 9.50 Total 1.00 0.50 1.89 0.00 26.12 5.69 50.51 TableA3summarizesTradePolicyUncertaintyindicesfor10concernedeconomiesinthe samplefromJanuary2010toDecember2019. Thefirstcolumnhascountrynames. Theother columns show statistics including mean, standard deviation, median, minimum, maximum, skewness, and kurtosis. All countries exhibit excess kurtosis and positively skewed. Table A4: U.S Trade Policy Uncertainty and Other Policy Uncertainty TPU Row TPU WTPU EPU TPU 1.000 Row TPU 0.415*** 1.000 WTPU 0.509*** 0.732*** 1.000 EPU 0.520*** 0.277*** 0.322*** 1.000 TableA4 documents the relation between country-level bilateral TPU with the U.S, competitor’s TPU, global TPU Benguria et al. (2022), and Economic Policy Uncertainty Baker et al. (2016). 102 Figure A1: U.S Trade Policy Uncertainty 1995-2019, Monthly FigureA1comparesthemeasurementofaverageU.STPUbetweenmineandBakeretal. (2016) from 1995-2019. I re-average both TPU indices to 1 to let them be comparable. The differences between the two TPU indices come from newspaper sources. Scott’s monthly TPU index for the United States relies on 10 leading newspapers: USA Today, Miami Herald,Chicago Tribune,Washington Post,Los Angeles Times,Boston Globe,San Francisco Chronicle, Dallas Morning News, New York Times, and Wall Street Journal. However, because of the accessibility, I did the automatic texting search on 7 newspapers: Boston Globe, Chicago Tribune, Los Angeles Times, New York Times, Wall Street Journal, and Washington Post. 103 Figure A2: Bilateral Trade Policy Uncertainty 1995-2019, Quarterly FigureA2showsthequarterlyTPUindexIusedinregressions. Iusedthesamemethod to generate that TPU index: search newspaper articles, and aggregate them based on quar- terly level. The results are similar to the monthly TPU indices. 104 Appendix B. Sample Characteristics Table B1: Summary Statistic of Industries GTAP % of Firms in Data Description FRS 21.59 Forestry and logging products ELE 11.22 Manufacture of computer, electronic and optical products GAS 5.88 Extraction of natural gas EEQ 5.14 Manufacture of electrical equipment CNS 4.57 Construction of buildings OME 4.51 Manufacture of machinery and equipment n.e.c. TRD 3.84 Wholesale and retail trade and repair of motor vehicles and motorcycles OBS 3.55 Professional, scientific and technical activities and Administrative and support service activities BPH 3.50 Manufacture of pharmaceuticals, medicinal chemical and botanical products FMP 2.55 Manufacture of fabricated metal products, except machinery and equipment CHM 2.14 Manufacture of chemicals and chemical products MVH 2.14 Manufacture of motor vehicles, trailers and semi-trailers I S 1.99 Manufacture of basic iron and steel CMN 1.82 Postal and courier activities OIL 1.77 Extraction of crude petroleum OAP 1.65 Swine / pigs COA 1.56 Mining of coal and lignite WAP 1.55 Manufacture of wearing apparel OTN 1.43 Manufacture of other transport equipment B T 1.09 Beverages RPP 1.07 Manufacture of rubber and plastics products NMM 1.00 Manufacture of other non-metallic mineral products NFM 0.97 Manufacture of basic precious and other non-ferrous metals AFS 0.97 Accommodation WHS 0.88 Warehousing and support activities for transportation FSH 0.87 Fishing and aquaculture HHT 0.83 Human health and social work activities ATP 0.73 Air transport OSD 0.71 Oil seeds and oleaginous fruit 105 GTAP % of Firms in Data Description OXT 0.66 Mining of metal ores OMF 0.65 Manufacture of furniture ROS 0.63 Arts, entertainment and recreation; Other service activities; OTP 0.60 Land transport and transport via pipelines TEX 0.53 Manufacture of textiles OCR 0.49 Stimulant, spice and aromatic crops WTP 0.48 Water transport GDT 0.46 Manufacture of gas; distribution of gaseous fuels through mains P C 0.43 Manufacture of coke and refined petroleum products V F 0.41 Vegetables SGR 0.34 Sugar and molasses OFD 0.34 Prepared and preserved fish, crustaceans, molluscs and other aquatic invertebrates INS 0.31 Insurance, reinsurance and pension funding, except compulsory social security C B 0.29 Sugar crops LUM 0.26 Manufacture of wood and of products of wood and cork, except furniture; WTR 0.24 Collection, purification and distribution of water, water collection, treatment and supply EDU 0.22 Education PPP 0.20 Manufacture of paper and paper products OMT 0.19 Meat of pigs, fresh or chilled ELY 0.19 Production, collection and distribution of electricity VOL 0.17 Animal fats CTL 0.10 Bovine animals, live WOL 0.09 Raw animal materials used in textiles LEA 0.07 Manufacture of leather and related products PCR 0.05 Rice, semi- or wholly milled, or husked RSA 0.03 Real estate activities OFI 0.02 Financial service activities, except insurance and pension funding RMK 0.02 Raw milk CMT 0.02 Meat of cattle, fresh or chilled 106 Appendix C. Robustness Specifications Table C1: Sample Selection Issue Panel A: Self Shock No China Manufacture (1) (2) (3) (4) (5) (6) (7) (8) Exporter Entry Exit Investment Exporter Entry Exit Investment TPU -0.304** -1.887*** -0.884*** -0.254** 0.105 -1.528*** -0.955*** -0.0460 (-2.33) (-11.18) (-6.12) (-2.14) (0.70) (-7.18) (-5.56) (-0.58) TPU× ∆ Exp.Tariff -0.0135 0.0685 -0.0619 -0.0809*** -0.0406** 0.0933*** -0.103*** 0.0207** (-0.44) (1.46) (-1.60) (-2.70) (-2.38) (3.84) (-5.26) (2.24) ∆ Exp.Tariff -0.275 -0.882* 0.542* -0.838*** -0.548** 0.255 1.595*** -0.0326 (-0.72) (-1.71) (1.73) (-2.89) (-2.07) (0.68) (4.90) (-0.25) r2 0.942 0.533 0.541 0.536 0.930 0.502 0.511 0.557 N 141836 141836 141836 43280 47825 47825 47825 24238 Panel B: Row Shock No China Manufacture (1) (2) (3) (4) (5) (6) (7) (8) Exporter Entry Exit Investment Exporter Entry Exit Investment Row TPU 0.682*** 3.514*** -1.263*** -3.329*** 1.409*** -0.510 -3.921*** -1.960*** (3.35) (11.32) (-4.91) (-17.09) (2.68) (-0.68) (-6.47) (-7.21) Row TPU× Row∆ Exp.Tariff -0.163*** -0.0371 -0.0226 -0.506*** -0.235*** 0.0266 -0.169* -0.250*** (-2.59) (-0.39) (-0.28) (-10.77) (-2.70) (0.21) (-1.68) (-5.31) Row ∆ Exp.Tariff -0.0291 -0.0723 0.103 1.057*** 0.104 -0.155 0.410 0.632*** (-0.09) (-0.14) (0.24) (5.17) (0.26) (-0.27) (0.89) (3.64) r2 0.942 0.533 0.541 0.545 0.930 0.500 0.510 0.561 N 141836 141836 141836 43280 47825 47825 47825 24238 t statistics in parentheses * p< 0.1, ** p< 0.05, *** p< 0.01 Table C1 shows the main results using the firm-quarter dataset from 2015Q1 - 2019Q4. Columns (1,2,3,4) study the subsample without China-U.S export relationships. Columns 107 (5,6,7,8) study the subsample of manufacturing firms. 108 Table C2: Firm’s Exposure to U.S Market Panel A: Self Shock TPU Exp.Tariff (1) (2) (3) (4) (5) (6) (7) (8) Exporter Entry Exit Investment Exporter Entry Exit Investment Share <25% -0.199 -0.887*** -1.202*** -0.0717 -0.361 0.228 0.608** -0.595*** (-1.55) (-4.87) (-8.03) (-0.74) (-1.45) (0.65) (2.10) (-3.41) Share 25-50% -0.203 0.220 -0.617* -0.110 -0.0717 -0.0786 0.381 -0.0946 (-0.76) (0.51) (-1.67) (-0.63) (-0.16) (-0.11) (0.60) (-0.34) Share 50-75% -0.303 -0.705** -0.985*** 0.0533 0.451 -1.334** 0.489 0.112 (-1.30) (-2.10) (-3.54) (0.51) (1.22) (-2.49) (1.10) (0.74) Share >75% -0.203 0.220 -0.617* -0.110 -0.0717 -0.0786 0.381 -0.0946 (-0.76) (0.51) (-1.67) (-0.63) (-0.16) (-0.11) (0.60) (-0.34) Panel B: Row Shock Row TPU Exp.Row Tariff (1) (2) (3) (4) (5) (6) (7) (8) Exporter Entry Exit Investment Exporter Entry Exit Investment Share <25% -0.0289 1.218** -3.654*** -2.813*** -0.112 0.319 -0.0912 0.467 (-0.07) (2.05) (-7.49) (-9.68) (-0.13) (0.26) (-0.09) (0.92) Share 25-50% -0.200 -0.966 -4.321*** -2.582*** -1.227 -1.591 0.227 0.0553 (-0.22) (-0.73) (-3.91) (-6.18) (-1.25) (-1.11) (0.19) (0.15) Share 50-75% 0.826 2.606** -2.767*** -3.434*** 0.183 -0.0964 -0.105 0.405 (0.94) (2.06) (-2.63) (-8.80) (0.25) (-0.09) (-0.12) (1.48) Share >75% 0.0861 1.264 -1.389 -5.465*** -0.373 -0.217 0.0292 -0.0363 (0.12) (1.11) (-1.43) (-12.56) (-0.61) (-0.22) (0.03) (-0.10) t statistics in parentheses * p< 0.1, ** p< 0.05, *** p< 0.01 Table C2 shows the main results using the firm-quarter dataset from 2015Q1 - 2019Q4. Columns (1,2,3,4) study the effects of TPU, Columns(5,6,7,8) study the effects of tariff expectations. Theeffectsarenotheterogeneousacrossthefirm’sexposuretotheU.Smarket. The exposure to the U.S market is measured as: Exposure = Revenue From U.S Market Total Revenue 109 Appendix D. Model Details Trade Policy Uncertainty Whenfirmsrealizethe τ ′ tomorrow,theycandecidewhethertoexitorenterthemarket depending on the realized profit level π ′ . The decision rule for each firm is defined by the trigger profit value π 1 and π 0 , which makes the firm indifferent between entry and waiting: π 1 − f =(1− β ) χ 0 ( π − f 1− β − F)− χ 1 π − f 1− β (D1) π 0 − f =(1− β ) χ 0 ( π − f 1− β − F)− χ 1 π − f 1− β = 1− β 1− β h F (D2) where χ 1 = 1 if π ≥ f, and χ 0 = 1 if π ≥ f +(1− β )F. An export firm in state 1 will stay in the market if π ′ ≥ π 1 and continue exporting; if π ′ < π 1 , the firm will exit the market. For a non-export firm in state 0, it will enter the market next period if π ′ ≥ π 0 and will continue waiting if π ′ < π 0 . After the shock release, the firm will stay or enter the market if it is profitable. Firms’ decisions will depend on the change in operating profits and the repayment of sunk costs. So the value function of export and waiting can be defined as: V 1 =(π − f)+Pr(π ≥ π 1 ) β 1− β (E(π ′ |π ′ ≥ π 1 )− f)+χ 1 π − f 1− β +Pr(π <π 1 )χ 0 π − f 1− β − F (D3) V 0 =Pr(π ≥ π 0 ) β 1− β (E(π ′ |π ′ ≥ π 0 )− f− F)+χ 1 π − f 1− β +Pr(π <π 0 )χ 0 π − f 1− β − F (D4) 110 Equation (1.32) and (??) are zero cutoff conditions for the marginal stay firm z 1 and marginal entry firm z 0 : π (z 1 )− f =− β ∗ U 1 (D5) π (z 0 )− f− (1− β )F =− β ∗ U 0 (D6) where U 1 and U 0 are uncertainty components: U 1 =Pr(π 1 ≤ π ′ ≤ π 0 ) 1 1− β [E(π ′ |π 1 ≤ π ′ ≤ π 0 )− π 1 ]+Pr(π ′ ≥ π 0 )F ≥ 0 (D7) U 0 =Pr(π 1 ≤ π ′ ≤ π 0 ) 1 1− β [E(π ′ |π 1 ≤ π ′ ≤ π 0 )− π 0 ]− Pr(π ′ ≤ π 1 )F ≤ 0 (D8) 111 Equilibrium Households in the home country choose final good consumption ( C t ), differentiated labor supply and wages for their members (l)j,t, and W j,t for j ∈ HH), and bond holdings (B t ) to maximize their utility E s X t β t− s U(C t ,L t ) (D9) subject to the budget constraint ¯ P t C t +B t ≤ Z W j,t l j,t +B t− 1 R t +Π HH t +T t (D10) Optimality conditions are: 1=βE (Λ t,t+1 R t+1 ) (D11) (π w − 1)π w t = ϵ w ρ w − U L,t U C,t − (ϵ w − 1) ϵ w w t +E(Λ t,t+1 R t+1 )(π w t+1 − 1)π w t+1 L t+1 L t (D12) Competitivelabormarketuseindividuallaborvarietiessuppliedbyhouseholdmembers to produce an aggregate labor input L t using a CES production technology. The undifferen- tiated input L t is then supplied to intermediate goods producers. The labor market cleaning condition: maxW t L t − Z W j,t l j,t (D13) s.t L t ≤ Z l ϵ w− 1 ϵ w j,t dj ϵ w ϵ w− 1 (D14) Optimality conditions are: l j,t = W j,t W t − ϵ w (D15) 112 and W t = Z W 1− ϵ w j,t dj 1 1− ϵ w (D16) The final consumption market is to solve: max ¯ P t C t − Z P t (i)Y t (i)di (D17) s.t C t ≤ Y t (i) ϵ p− 1 ϵ p di ϵ p ϵ p− 1 (D18) Optimality conditions are: Y t (i)= P t (i) ¯ P t − ϵ p Y t (D19) and ¯ P t = Z P t (i) 1− ϵ p di 1 1− ϵ p (D20) Wholesale firms choose Y t (i), P t (i), Q H,t , and Q F,t to maximize: maxE s X t≥ s β t− s Λ t,s Π W Y,t P t (D21) s.t Π W Y,t =P t Y t − P H,t Q H,t − P F,t (1+τ t )Q F,t (D22) P t = ω(P Ht ) 1− ϵ f +(1− ω)(P Ft (1+τ t )) 1− ϵ f 1 1− ϵ f (D23) Q H,t =ω P H,t P t − ϵ f Y t (D24) Q F,t =(1− ω) P F,t (1+τ t ) P t − ϵ f Y t (D25) MC t = ω(P H,t ) 1− ϵ f +(1− ω)(P F,t ) 1− ϵ f (1+τ t ) 1− ϵ f 1 1− ϵ f (D26) 113 The distributors maximize profits given by: π D H,t =P H,t Q H,t − Z p H,t (j)y H,t (j)dj (D27) π D F,t =P F,t Q F,t − Z p F,t (j)y F,t (j)dj (D28) So the demand schedules of competitive distributors in the domestic and foreign markets: y H,t (j)= p H,t (j) P H,t − ϵ Q H,t (D29) y F,t (j)=N ∗ t − λ ϵ − 1 ϵ p F,t (j) P F,t − ϵ Q F,t (D30) An intermediate good producer with individual state (z,k,m − 1 ,τ ), chooses m to solves the following dynamic recursive problem: V(z t ,k t ,m t− 1 ,τ t )=max(π t − m t F(m t− 1 ))+E t Λ t,t+1 V(z t+1 ,k t+1 ,m t ,τ t+1 ) (D31) s.t π t =p H,t y H,t +m t q t p ∗ H,t y ∗ H,t − W t l t − p k t i t (D32) y H,t +m t y ∗ H,t ≤ A t z t k α t l 1− α t (D33) k t+1 =(1− δ k )k t +i t (D34) Optimality conditions are: p H,t = ϵ ϵ − 1 w t l t (1− α )(A t z t k α t 1− α t ) (D35) p k t =E t λ t,t+1 V k,t+1 (D36) 114

Abstract (if available)

Abstract In this thesis, I investigate the impact of various shocks on global trade using both firm and country-level data from over 100 countries dating back to 1970. I specifically focus on the effects of these shocks on export and import firms, aiming to understand how supply chains are disrupted and the unique adjustments firms make in response to these disruptions. My analysis reveals that uncertainty shocks lead to delays in the entry and exit probabilities for export firms, while import firms with more diversified supply chains demonstrate increased resilience. Furthermore, my results indicate that downstream firms are disproportionately impacted by global supply chain disruptions. In contrast, domestic economic contraction shocks have adverse effects on both extensive and intensive export margins, predominantly driven by a reduction in the intensive margin.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Essays on the dual urban-rural system and economic development in China

PDF

Empirical essays on trade liberalization and export diversification

PDF

Essays on business cycle volatility and global trade

PDF

Three essays on heterogeneous responses to macroeconomic shocks

PDF

Essays in international economics

PDF

Three essays in international macroeconomics and finance

PDF

Essays on firm investment, innovation and productivity

PDF

Essays in development economics

PDF

The impact of trade liberlaization on firm performance in developing countries -- new evidence from Pakistani manufacturing sector

PDF

The impact of minimum wage on labor market dynamics in Germany

PDF

Understanding teacher burnout through a lens of hope in high-poverty schools

PDF

The importance of mentoring in leadership self-efficacy for women in supply chain: a qualitative study

PDF

Intergenerational transfers & human capital investments in children in the era of aging

PDF

Three essays on the microeconometric analysis of the labor market

PDF

Essays on innovation, human capital, and COVID-19 related policies

PDF

The murky risk of trade protectionism in an interconnected and uncertain global economy: a state, industrial, and regional analysis

PDF

Essays in climate change adaptation: role of market power in incentivizing adaptation behavior

PDF

Characterization of health outcomes in patients with hemophilia A and B: Findings from psychometric and health economic analyses

PDF

Essays on development economics and adolescent behavior

PDF

From the African continent to Colombia: a Case study of African Leadership Academy and developing a new generation of purpose driven Afro-Colombian agents of change

Asset Metadata

Creator Ding, Yukun (author)

Core Title The impact of economic shocks on firm behavior: insights from three studies

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Degree Conferral Date 2023-05

Publication Date 03/31/2023

Defense Date 03/07/2023

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag export,global value chain,international economics,OAI-PMH Harvest,tariff,trade,trade agreements,trade partners

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Betts, Caroline (committee chair), Bairoliya, Neha (committee member), Chaney, Thomas (committee member), Morlacco, Monica (committee member)

Creator Email dyukun@yahoo.com,yukundin@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC112859667

Unique identifier UC112859667

Identifier etd-DingYukun-11545.pdf (filename)

Legacy Identifier etd-DingYukun-11545

Document Type Dissertation

Format theses (aat)

Rights Ding, Yukun

Internet Media Type application/pdf

Type texts

Source 20230331-usctheses-batch-1014 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu