University of Southern California Dissertations and Theses

Essays on the microeconomic effects of taxation policies

(USC Thesis Other)

Essays on the microeconomic effects of taxation policies

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content ESSAYS ON THE MICROECONOMIC EFFECTS
OF TAXATION POLICIES
by
Rachel A. Lee
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 2022
Copyright 2022 Rachel A. Lee
Dedication
For Mom, Dad, and Danni without whom none of this would have been possible.
And for Maddie without whom... this still would have been possible, but not nearly as fun.
ii
Acknowledgments
I’mverygratefulfortheacademicandfinancialsupportoftheUSCeconomicsdepartment
and for my main advisor, Geert Ridder, who knew little of me before investing in me first as a
research assistant and later as a PhD student. His guidance helped tremendously in pushing
throughthedifficultstagesofacademicresearchtothecompletionofmydissertationandwill
never be forgotten. I’d also like to thank Cheng Hsiao and Emily Nix for their participation
in my defense as members of my dissertation committee as well as Paulina Oliva, Monica
Morlacco, and Pai Ling Yin for serving on my qualifying exam committee.
The administrators of the economics department who help orchestrate everything from
printer assistance, to moral support, to petitions and forms required to progress through this
program have been supporting me since day one. Morgan Ponder and Young Miller were
amongst the first to welcome me to Los Angeles and USC while Alex Karnazes and Annie
Le have become instrumental in getting me through my later years in the program. Thank
you for always being there to organize the chaos I bring upon myself.
While the knowledge I’ve gained in this program will accompany me through life, I’m
grateful for friends to bring with me as well. Bryson Yee, Usman Ghaus, and Grigory
Franguridi never failed to help pull me along when I felt I might be lagging behind. Eunjee
Kwon and Jisu Cao reminded me how precious friendships between girls can be and have
served as the best examples to follow in achieving both academic and personal success. Ali
Abboud trained me to reach my physical peak while also becoming one of the most reliable
and caring friends. Juan Espinosa, Monira Al Rakhis, and Amy Mahler became central to
iii
my sanity, always offering a safe enviornment to discuss not only the struggles of the PhD
program, but everything else unrelated to economics.
Together, Andreas Aristidou and I faced difficulties, heartbreak, laughter, excitement,
and anxiety but all seemed manageable with a best friend to lean on or celebrate with at
the end of the day. Thank you for always being there for me and Maddie. Without you, I
would have surely lost my mind somewhere along the way.
Beyond the economics department, I met Stefan Bogdanovich who changed my life in
ways that I didn’t realize needed changing. Thanks for being there to pull me out of the
darkness when everything seemed to fall apart. For understanding when I prioritized my
work above all else. And for loving and caring for me and Maddie when we needed it the
most.
I’m grateful to have hit the biological jackpot with Jesika & Jesse Mak, Simone Lee, and
Hugh Chin who are my best friends by blood. Also to have found my best friends by choice,
Kara Rubin and Evan Hacker, who have stuck with me through this madness and continue
to find ways to support me even from across the country. You all make coming home better
than being anywhere else for the holidays. Unless we’re meeting in Vegas.
My grandparents’ sacrifices taught me that there are no limits to what I can achieve if I
fight hard enough. They devoted their lives to ensuring that their children and grandchildren
have every opportunity to succeed. Thank you for your unconditional love. I’m extremely
proud to be the granddaughter of Ken & Valerie Lee and Hector & Daisy Chin. I hope to
always make you proud as well.
Finally, I want to thank my family for standing by me in every moment of my life that
has led up to this point. Mom, Dad, and Danni never hesitate to answer my calls whether
I’m crying from stress or excitement. I could not have accomplished any of this without you
and I wouldn’t want to celebrate with anyone else. The immense struggles of this program
have taught me how critical a loving and supportive family is when life gets difficult and I
will never take that for granted. I love you and I’m eternally grateful for you.
iv
Table of Contents
Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
1 Examining the Effects of Cigarette Excise Taxes: Evidence from a Chicago
Supermarket Chain 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4.1 Tax Incidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4.2 Causal Effect of the Cook County Tax Increase . . . . . . . . . . . . 11
1.4.3 Consumer Response . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.4.4 Comparison of Results . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.5 Final Thoughts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2 The Effects of SALT Deduction on Homeownership 23
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 Previous Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
v
2.3 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3.1 Tax Cut and Jobs Act of 2017 . . . . . . . . . . . . . . . . . . . . . . 27
2.4 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.6 Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.6.1 Identification Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.6.2 Effect of SALT Deduction Limit on Home Ownership . . . . . . . . . 43
2.6.3 Who is Primarily Affected? . . . . . . . . . . . . . . . . . . . . . . . 46
2.7 Final Thoughts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Bibliography 50
A Appendix to Chapter 1 53
A.1 Cigarette Summary Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . 53
A.2 Difference-in-Differences: Falsification Test . . . . . . . . . . . . . . . . . . . 54
A.3 Synthetic Difference-in-Differences Plots . . . . . . . . . . . . . . . . . . . . 55
A.4 Traditional Difference in Differences . . . . . . . . . . . . . . . . . . . . . . . 59
A.5 Synthetic Difference in Differences . . . . . . . . . . . . . . . . . . . . . . . . 60
A.6 Market Shares of Cigarette Product Types . . . . . . . . . . . . . . . . . . . 66
A.7 Cigarette Quantity Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . 71
vi
List of Tables
1.1 Cigarette Excise Tax Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2 Tax Incidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Cook County Tax Increase on Quantities Sold . . . . . . . . . . . . . . . . . 14
1.4 SDID Tax Increase Effects on Quantities Sold . . . . . . . . . . . . . . . . . 15
1.5 Cigarette Quantity Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.6 Comparing Tax Elasticities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.1 SALT Impact Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.2 Descriptive Statistics by State Political Affiliations . . . . . . . . . . . . . . 40
2.3 Political Leanings on Ownership (Same House) . . . . . . . . . . . . . . . . . 42
2.4 Political Leanings on Ownership (Moved Within State) . . . . . . . . . . . . 42
2.5 Political Leanings on SALT Impact . . . . . . . . . . . . . . . . . . . . . . . 44
2.6 Effect of SALT Deduction Limit on Homeownership . . . . . . . . . . . . . . 45
2.7 Effect of SALT Deduction Limit on Homeownership . . . . . . . . . . . . . . 47
A.1 Summary Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
A.2 Falsification: Grocery Sales DID . . . . . . . . . . . . . . . . . . . . . . . . . 54
A.3 Tax Increase on Quantities Sold . . . . . . . . . . . . . . . . . . . . . . . . . 59
A.4 SDID Tax Increase Effects on Quantities Sold . . . . . . . . . . . . . . . . . 60
A.5 Cigarette Quantity Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . 71
A.6 Relative Tax Quantiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
vii
List of Figures
1.1 Stores by Tax Zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2 Cigarette Product Sales Surrounding the Cook County Tax Increase . . . . . 11
1.3 Stores by Closest Lower-Tax Zone . . . . . . . . . . . . . . . . . . . . . . . . 16
2.1 Percent of SALT Deductions by State, 2017 . . . . . . . . . . . . . . . . . . 30
2.2 Percent of SALT Deductions by State, 2018 . . . . . . . . . . . . . . . . . . 31
2.3 Percent of SALT Deductions by Income, 2017 . . . . . . . . . . . . . . . . . 32
2.4 Percent of Returns by Income, 2017 . . . . . . . . . . . . . . . . . . . . . . . 33
2.5 Budget Constraint: No SALT Deduction . . . . . . . . . . . . . . . . . . . . 34
2.6 Budget Constraint: No SALT Impact . . . . . . . . . . . . . . . . . . . . . . 35
2.7 Budget Constraint: SALT Impact . . . . . . . . . . . . . . . . . . . . . . . . 36
2.8 SALT Impact & Higher State Income Tax Rate . . . . . . . . . . . . . . . . 37
A.1 SDID Discount Pack Sales Residuals . . . . . . . . . . . . . . . . . . . . . . 55
A.2 SDID Discount Carton Sales Residuals . . . . . . . . . . . . . . . . . . . . . 56
A.3 SDID Premium Pack Sales Residuals . . . . . . . . . . . . . . . . . . . . . . 57
A.4 SDID Premium Carton Sales Residuals . . . . . . . . . . . . . . . . . . . . . 58
A.5 SDID Overall Store Cigarette Product Sales . . . . . . . . . . . . . . . . . . 61
A.6 SDID Discount Pack Sales . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
A.7 SDID Discount Carton Sales . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
A.8 SDID Premium Pack Sales . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
viii
A.9 SDID Premium Carton Sales . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
A.10 Market Shares of First Income Quartile . . . . . . . . . . . . . . . . . . . . . 67
A.11 Market Shares of Second Income Quartile . . . . . . . . . . . . . . . . . . . . 68
A.12 Market Shares of Third Income Quartile . . . . . . . . . . . . . . . . . . . . 69
A.13 Market Shares of Fourth Income Quartile . . . . . . . . . . . . . . . . . . . . 70
ix
Abstract
This thesis comprises of two research papers which empirically explore the consumption
and behavioral responses to two different, real-world taxation policies.
The first chapter measures the causal effect of an increase to the cigarette excise tax on
the quantities of cigarettes sold using supply-side scanner data from a Chicago supermarket
chain. It also estimates excise tax elasticities.
The second chapter examines the effect of the state and local tax (SALT) deduction limit
on individuals’ homeownership decisions using a unique instrumental variable to measure a
taxpayer’s SALT impact in a linear probability model for homeownership.
x
Chapter 1
Examining the Effects of Cigarette
Excise Taxes: Evidence from a Chicago
Supermarket Chain
1.1 Introduction
The United States has a long history of tobacco taxation, from the early days of British
colonization to current day where the federal excise tax stands at $1.01 per pack on top of
the median state excise tax of $1.78 per pack. Many local governments have also tacked
on additional excise taxes driving cigarettes prices even higher. According to Citizens for
Tobacco Rights, American consumers paid over $42.3 billion in federal, state, and local
cigarette excise taxes in 2019.
An excise tax is a per unit tax that generally has disproportionate effects on vertically
differentiated products, therefore altering relative prices within the market. Cigarette excise
taxes have been studied for years and continue to be a topic of discussion due to the economic
and health impacts they have on society. By understanding consumer responses to such
taxes, economists gain further insight on how people may react to other sin taxes, like those
1
on sugary drinks or marijuana. Policy makers also have the ability to make more informed
decisions that align with their objectives, whether it be raising funds for public projects via
tax revenue or decreasing consumption levels for the sake of health outcomes.
This paper employs weekly scanner data spanning 1989 to 1997 from the Chicago grocery
chain, Dominick’s Finer Foods, to study consumer responses to changes in cigarette excise
taxes. I begin by examining the causal effect of a tax increase on the quantities of cigarette
products sold, then utilize store-level location and sales data to analyze the relationship
betweenexcisetaxesandlocaltaxborders. Mymainfindingsindicatethatdiscountcigarette
pack sales decreased by 13 percent in response to the Cook County excise tax rate.
The first portion of my analysis estimates the effects of the excise tax rate on the sales
prices of cigarette products by exploiting the four excise tax increases which occur during the
datasample. Ifindthatconsumerstendtobearthemajorityofthetaxburden, payingabout
76 cents of every dollar increase to the tax rate. This aligns with the results of Espinosa
and Evans (2013), Harding et al. (2012), Sullivan and Dutkowsky (2012) who find varying
degrees of cigarette excise tax pass-through rates but all agree that taxes raise consumer
prices.
Motivated by the tax incidence born by consumers, I next perform causal analysis on an
event study surrounding the 1996 Cook County excise tax increase to examine its effect on
sales quantities. Unlike the other tax increases that occur during the data sample, the Cook
County tax hike only affects stores located within its county borders. I calculate difference-
in-differences estimators to find that premium pack sales experience 14.9 percent decrease
in products sold after the tax increase. Additionally, I find evidence of stockpiling behavior
in the sales of discount products. The DID results rely on the key identifying assumption
that the quantity of cigarette product sales would have evolved similarly across the greater
Chicago area covered by the sample in absence of the 1996 Cook County tax increase.
To relax the assumption of parallel pre-trends I use the synthetic difference-in-differences
methodology of Arkhangelsky et al. (2021) to estimate the local average treatment effects of
2
the tax increase. Unlike the standard difference-in-differences approach, this method places
more weight on observations which are similar to the treated times and units. The results
suggest discount pack sales decrease by 13 percent after the Cook County tax increase.
However, I find no signficiant effect of the tax hike on any of the other product types. The
synthetic difference-in-differences results are preferred over the standard DID results due to
increased precision.
In the final portion of this analysis, I estimate tax elasticities and use store distances to
bordering tax zones and relative tax rates to examine the possibility of local border-crossing
and flight to quality. I find that sales of all product types except premium cartons decrease
with an increase in the excise tax amount. The tax elasticities of pack sales are greater
than those of carton sales, which is likely attributable to the types of consumers of packs
and cartons. People who purchase a single pack of cigarettes are more likely to be casual
smokers relative to those buying cartons, which bundle 10 packs of cigarettes. I find no
significant evidence of border-crossing behavior in my analysis.
The results of this paper suggest that sales quantities change in response to cigarette
excise taxes, however, the effects are heterogenous across different product types. That said,
cigarette sales typically decline with increases to the excise tax amounts since much of these
taxes pass-through to the consumer. Local governments should perform cost-benefit analyses
of excise taxes and potentially consider ad-valorium taxes as an alternative if they want to
even the effects across discount and premium products.
The remainder of the paper is organized as follows. Section 1.2 reviews the previous
literature. Section 1.3 discusses the data used in the analysis and provides some summarizing
statistics. Section 1.4 explains the empirical strategy and results of the analysis, highlighting
the main variables of interest. Lastly, section 1.5 concludes the discussion.
3
1.2 Literature
This paper contributes to two strands of literature. The first is that which examines
excise tax incidence. Espinosa and Evans (2013) use scanner data to analyze the effects of
state tax increases on the weighted average prices of different types of cigarettes. They find
a one-to-one relationship between increases in the tax rate and in prices paid by consumers.
Harding et al. (2012) uses Nielson Homescan data to examine heterogeneous geographic and
socioeconomic effects of taxes on cigarette prices. Their results contradict the literature
as they find that cigarette taxes are not fully passed to the consumer, with evidence of
excise tax incidence varying with geographic and socioeconomic characteristics. Unlike the
data used for this paper, their dataset contains information on household-level purchases
so instead of measuring a store’s distance to the border and aggregate demographics for its
neighborhood, they are able to precisely locate and characterize each consumer and their
purchase decisions. Sullivan and Dutkowsky (2012) use panel data to determine relative
effects of city, county, and state taxation on prices and find that a one dollar increase in
tax rates actually increases prices by more than a dollar. However, similarly to Harding
et al. (2012), they find that urban areas located near states with lower cigarette taxes tend
to have lower prices relative to areas located near states with the same or higher tax rates.
Chiou and Muehlegger (2014) use the same dataset as the one in this study to examine the
pass-through of taxes to purchase prices at the UPC-level with time and store fixed effects,
along with a first differences specification to look at changes in prices.
This study also contributes to the strand of literature regarding consumer responses
to excise taxation. One branch of the behavioral response literature explores the theory
introduced by Barzel (1976), which models substitution towards higher-quality goods in the
presence of excise taxation. Espinosa and Evans (2013) investigate whether tax changes
encourage this "flight to quality" by estimating market shares for discount and premium
brands. They find that while there are substitutions from carton to pack sales, there isn’t
evidence of substitution towards name brand products. These results contrast with those
4
found in the current study, which detects decreases in pack sales. Nargis et al. (2014) use
survey data to estimate a logit model that determines the probability of a consumer choosing
discount brand cigarettes in response to relative price changes to premium brands. While
they find evidence supporting the theory of Barzel (1976) for Canadian consumers, they find
no significant effect for Americans. Chiou and Muehlegger (2014) also find that in the long
run, tax increases are associated with substitutions toward high-tier cigarettes. This study
adds to the discussion by using causal methods to find that an increase in the excise tax rate
leads to decreases in the consumption of packs of cigarettes but find no significant increases
in carton sales.
In addition to the "flight to quality" theory, the literature regarding border-crossing for
tax avoidance is particularly relevant to the current study. DeCicca et al. (2013b) extend
the standard optimal Pigouvian tax model to incorporate the possibility of tax avoidance by
border-crossing. After estimating their structural model, they find that when accounting for
tax avoidance, the optimal tax is at least 20 percent smaller than the standard Pigouvian
tax. Other papers such as Stehr (2005) explore tax avoidance and evasion through both
legal channels such as Internet purchases and border-crossing, and illegal channels such as
the smuggling of cigarettes. Their results suggest that state legislations with preferences
for smoking controls pair high taxation with policies to reduce avoidance. Merriman (2010)
collects data from a random sample of littered cigarette packs around Chicago to support the
tax avoidance theory. He finds that three-quarters of the packs don’t contain Chicago tax
stamps, which has rates much higher than neighboring jurisdictions, and that the closer the
packs are to lower-tax state borders, the less likely they are to contain the local tax stamp.
Similarly, Chernick and Merriman (2013) collect data from a random sample of littered
cigarette packs in New York City before and after a tax increase and find that tax avoidance
increased and consumption decreased. Unsurprisingly, they find that avoidance rates are
higher in poorer areas. Chiou and Muehlegger (2008) develop a discrete choice model to
examine tax avoidance and state border crossing in the market for cigarettes. They use their
5
model to estimate individuals’ decisions to travel across borders to purchase cigarettes at
lower tax rates.
The current study contributes to the consumer response literature by examining excise
tax elasticities and the impacts of cigarette excise taxes on border crossing within a state,
paying close attention to the possibility of consumers traveling shorter distances to avoid
local taxation. Most of the previous literature on tax avoidance via border crossing focuses
on interstate differences and fails to account for smaller areas and local taxation. This is
particularly problematic because the likelihood of people traveling shorter distances to save
on non-durable goods should be higher due to the lower transportation costs and the rate
at which people replenish their stocks. I find minimal evidence of excise tax differentials’
effects on cigarette sales quantities, regardless of store distance to border. This suggests that
local policy makers may use cigarette excise taxes to generate revenue regardless of tax rates
in neighboring jurisdictions so long as doing so does not decrease sales by so much that it
offsets the additional tax revenue.
1.3 Data
Toexaminetheimpactofcigaretteexcisetaxesonconsumerbehavior, Iestimateavariety
of empirical models using scanner data from Dominick’s Finer Foods, which is provided by
Chicago Booth’s Kilts Center for Marketing. The dataset includes information on weekly
sales for 84 of the Dominick’s Finer Food grocery stores spread throughout the Chicago area
from September 1989 through May 1997 with prices and the number of units sold at the
UPC-level for each store-week. It also includes each store’s total grocery sales in every week
along with store demographic profiles from the 1990 US Census, describing their assumed
customer bases.
For this project, I focus solely on cigarette sales and reconcile each of the UPC codes
with their manufacturers and brand names. I categorize the products based brand name
6
and packaging. Following the definitions of Cornelius et al. (2014), I seperate brands into
premium or discount categories based on marketing representations made by the cigarette
manufacturers. Additionally, I consider whether the product is packaged individually as a
pack or bundled together as a carton of 10 packs.
Table 1.1: Cigarette Excise Tax Rates
Year Federal Illinois Cook County Chicago Evanston
1989 0.16 0.30 0.10 0.16 0.10
1991 0.20 0.30 0.10 0.16 0.10
1993 0.24 0.44 0.10 0.16 0.10
1996 0.24 0.44 0.18 0.16 0.10
Tax rates are for packs of 20 cigarettes. Packs with 25 cigarettes are charged 1.25x the rates shown
and cartons are charged 10x the rates shown.
Stores are seperated into 4 different taxzones based on their geographical locations. The
Chicago and Evanston stores are subject to taxes levied by the federal, state, Cook County,
and municipality governments. Cook County stores located outside of the Chicago and
Evanston city boundaries are taxed by all but the municipality governments. The remainder
only pay federal and state tax rates. A breakdown of each zone’s tax components is shown
below:
Chicago = Federal + Illinois + Cook County + City of Chicago
Evanston = Federal + Illinois + Cook County + City of Evanston
Cook County = Federal + Illinois + Cook County
Illinois = Federal + Illinois
Cigarette excise tax rates and changes to the tax rates at the federal- and state-levels
were collected from Orzchowski and Walker (2013), while county- and municipality-level
taxes were found in historical accounts of ordinances on local government websites. Tax
rates and years of changes for packs of 20 cigarettes are displayed in table 1.1. The total
excise tax on a product depends on the supplying store’s location, the week in which the
7
purchase was made, and the number of cigarettes bundled together.
An integral part of this analysis relies on store distances to bordering taxzones. While
previous studies have focused mainly on distances to state borders, I consider a store’s
distance from the closest lower tax zone, whether it be at the local- or state-level, to examine
how that may impact quantities sold. This helps shed insight on how consumers might move
between local zones to purchase their cigarettes at lower tax rates. However, since the data is
restricted to grocery stores, it is worth noting that the sample may not be as representative
as one that includes tobacco shops and convenience stores, especially if consumers prefer to
buy their cigarettes from specialty stores and only purchase from grocery stores on occasion.
1.4 Empirical Strategy
Inthissection, IexaminetheCookCountytaxincreaseof1996andestimateit’seffectson
quantities of cigarette products sold in Dominicks Finer Food stores within the Cook County
borders. The difference-in-differences approach relies on the key identifying assumption that
absent this tax increase, all store cigarette sales within the sample would have evolved
similarly across time. I also relax this assumption by adopting the synthetic difference-in-
differences methodology of Arkhangelsky et al. (2021), which estimates the local average
treatment effects of the Cook County tax increase.
Additionally, I explore consumer response to increases in excise taxation by estimating
the demand for cigarettes. The inclusion of a store’s relative tax rate and its distance to
the closest lower tax zone border sheds light on the relationship between cigarette sales
quantities and local tax borders and rates.
1.4.1 Tax Incidence
When imposing an excise tax on a product, it is critical for policy makers to consider
who actually bears the burden of the tax. From the basic principles of microeconomics, we
8
Figure 1.1: Stores by Tax Zone
know that the relatively inelastic group will pay the majority of the tax levied on a good.
This is reflected in the market price set in the presence of an excise tax. The change in
equilibrium price from that without the tax, or before the tax increase, reflects the portion
paid by consumers. Since smokers are often addicted to cigarettes, it is reasonable to imagine
that demand is relatively inelastic in the short run and that they will continue to consume
even after increases to the purchase price due to tax rate hikes. Assuming this is true,
the equilibrium price should increase with the tax rate. To test this theory, I run a price
regression to examine the impact of tax rates on the cigarette prices paid by consumers.
9
I estimate the effect of the taxrate on the sales price of each cigarette product with the
following regression
P
kt
=
1
Tax
kt
+
kt
(1.1)
where k indicates the product UPC and t is the week. Tax
kt
is the total tax paid on a
product k in week t. The regression is first-differenced so fixed effects are dropped from the
equation.
Table 1.2: Tax Incidence
All Products Disc. Packs Disc. Cartons Prem. Packs Prem. Cartons
Tax 0.76*** 1.21*** 0.85*** 1.07*** 0.65***
(0.02) (0.03) (0.03) (0.01) (0.01)
Standard errors are clustered at the UPC-store level.
The excise tax incidence is measured by the tax rate’s effect on the price paid by con-
sumers. Since the actual price paid is the equilibrium price, a one-to-one relationship be-
tween the first-differenced price and tax rate would mean that consumers are bearing the
full burden of the tax. In equation 1.1,
1
represents how much of the tax change is paid by
consumers. Table 1.2 contains the results of equation (1.1) and shows that for every $1.00
increase to the tax rate, about 76 cents is passed on to the consumer. This result supports
my hypothesis that consumers pay a large portion of the tax imposed on cigarettes.
When equation 1.1 is estimated for each of the four product types seperately, I find that
cigarette pack prices are much more heavily affected by changes to the tax rate than carton
prices. In fact, for every $1.00 increase to the excise tax rate, the price of discount packs go
up by $1.21 and the price of premium packs go up by $1.07. These overcompensating pack
prices follow previous findings, Sullivan and Dutkowsky (2012), and are somewhat offset by
the effect of tax changes on carton prices. Discount carton prices increase by 85 cents for
every dollar increase in their tax rate while premium cartons increase by 65 cents. DeCicca
et al. (2013b) reports similar results of carton buyers paying less of the tax than pack buyers.
10
1.4.2 Causal Effect of the Cook County Tax Increase
Motivated by the results of tax rates’ pass through to cigarette prices, I proceed to study
the causal effect of the Cook County cigarette excise tax increase on the sales quantities
of cigarette products. The key identifying assumption is that cigarette sales quantities
would have evolved similarly throughout stores in the greater Chicago area absent the 1996
Cook County excise tax increase. To visually examine the plausibility of this assumption, I
seperately plot weekly cigarette sales over time for Dominick’s Finer Food stores within and
outside of Cook County, shown in figure 1.2.
Figure 1.2: Cigarette Product Sales Surrounding the Cook County Tax Increase
Based on figure 1.2, the sales trends for Cook County relative to the Illinois-only stores
for all four product types are generally parallel to one another before the Cook County tax
11
increase which occurs in week 347. This suggests comparability between the treated Cook
County store outcomes and the Illinois-only stores by means of a difference-in-differences
analysis. After the increase is implemented, Cook County store cigarette sales seem to
decrease at a faster rate than the Illinois-only store sales, implying the Cook County tax
increase may have a possibile negative effect on quantities of cigarette products sold.
To further support the common trends assumption, I perform a falsification test using
total grocery sales as the outcome of interest. The results are reported in appendix A.2.
Since Dominick’s Finer Foods is a supermarket chain, I assume that grocery items are the
primary reason for consumers to shop at these stores and that they don’t abandon their usual
supermarket because of changes in cigarette product prices. I empirically test whether the
cigarette excise tax increase has an effect on sales of grocery items unrelated to the tax and
find no significant effect. This implies store traffic remains unaffected by the increase in the
Cook County cigarette excise tax so if the quantity of cigarette products sold decreases, then
we can argue that the link is causal. The falsification result also supports the assumption
that network effects do not exist or have negligible effects between the stores in the sample.
In other words, if consumers cross borders for cigarettes, they most likely purchase from
stores other than Dominicks Finer Foods such as tobacco stores.
I use the difference-in-differences specification to measure the causal effect of the Cook
County tax increase on cigarette sales quantities. To control for each store’s weekly foot
traffic, I use grocery sales as a proxy and assume that this is exogenous to cigarette sales.
SinceDominick’sFinerFoodisasupermarketchain,it’sreasonabletoassumethatconsumers
go to the stores primarily for grocery items and if people typically stick to a general budget
then grocery sales should reflect store traffic. Additionally, since sales surrounding the Cook
Countytaxincreaseseemtofluctuate, Iincludeastockpilingindicatorvariables, withonefor
the 4 weeks prior to the tax increase and another for the 4 weeks following the tax increase.
I estimate the following regression seperately for each of the four cigarette product types
12
log(Q
jt
) =
1
Cook
j
+
2
Post
t
+
DID
(Cook
j
Post) (1.2)
+
3
log(G
jt
) +
4
B
t
+
5
A
t
+
jt
where Q
jt
is the quantity of cigarette products sold by store j in week t.
DID
is the
difference in differences estimator measuring the causal effect of the tax increase on sales.
The indicator variable, Cook
j
, takes the value of 1 if store j is within Cook County and
Post
t
takes the value of 1 if t is after week 347, when the tax increase has occurred. G
jt
is the total grocery sales for store j during week t and the stockpiling dummy variables, B
t
andA
t
, take a value of 1 in the four weeks prior to the tax increase and the four weeks after
the increase, respectively. Standard errors are clustered at the store level.
The results for equation 1.2 are reported in table 1.3. The increase in the cigarette excise
tax rate has a negative and significant effect premium pack sales but no effect on the other
product sales. This suggests that the Cook County tax hike caused a 14.9 percent decrease
in the premium pack sales for the Dominick’s Finer Food stores located within the county
borders.
In the weeks before the tax increase, discount pack sales increased by almost 32 percent
and discount carton sales increased by 20 percent possibly due to the anticipation of higher
prices. Intheweeksaftertheincrease,discountcartonsalesdeclinedby9percent. Combined,
these results imply consumers of discount products exhibit stockpiling behavior surrounding
the Cook County tax increase. Premium product sales, on the other hand, saw no significant
change prior to the tax increase but both premium packs and cartons experience higher sales
after the tax change. This is consistent with the flight to quality theory, which predicts
substitution towards higher-quality products in response to excise taxation.
In addition to the standard difference-in-differences approach, I also calculate the syn-
thetic difference-in-differences estimator based on the methodology of Arkhangelsky et al.
13
Table 1.3: Cook County Tax Increase on Quantities Sold

did
Grocery Before After
Discount Packs -0.192 0.483* 0.317*** -0.025
(0.135) (0.233) (0.072) (0.062)
Discount Cartons -0.031 0.230* 0.205*** -0.091**
(0.065) (0.110) (0.041) (0.032)
Premium Packs -0.149* 0.987*** 0.013 0.090**
(0.076) (0.219) (0.031) (0.030)
Premium Cartons 0.025 0.826*** 0.018 0.123**
(0.080) (0.212) (0.026) (0.040)
Robust standard errors clustered at store-level in parentheses.
.significant at 10%; *significant at 5%; **significant at 1%; ***significant at 0.1%
(2021). Instead of relying solely on the traditional DID estimator or synthetic control es-
timation, this approach combines the two methods. In doing so, it allows me to relax the
parallel trends assumption as I would with a synthetic controls approach, but then apply
the difference-in-differences estimation technique to the re-weighted panel.
To control for foot traffic and stockpiling behavior as in the standard differences-in-
differences estimator described earlier, I regress these independent variables on quantities
sold and use the residuals to obtain the synthetic difference-in-differences estimators.
log(Q
jt
) =
1
log(G
jt
) +
2
B
t
+
3
A
t
+
j
+
jt
(^
sdid
; ^ ; ^ ;
^
) = argmin
;;;
N
X
j=1
T
X
t=1
(
jt

j

t
W
jt
)
2
^ !
sdid
j
^

sdid
t
(1.3)
The Arkhangelsky et al. (2021) estimator shown in 1.3 is very similar to that which is
used in two-ways fixed effects estimation, except this approach adds store and time weights,
^ !
sdid
j
and
^

sdid
t
. These weights place greater emphasis on stores and times which are similar
to the treated stores and time, effectively localizing the estimated treatment effect. The
re-weighting of the panel takes the synthetic approach, which typically includes unit weights
to relax the common trends assumption, one step further by adding time weights which
14
emphasize pre-treatment periods similar to post-treatment periods. This should lead to
greater precision since the estimated effects will not absorb the differences of time periods
that may be incomparable to treated periods.
Table 1.4 reports the synthetic difference-in-differences results. These estimates are more
conservativethanthoseobtainedfromthestandarddifference-in-differencespecificationsand
the unconditional specifications in appendices A.4 and A.5. When controlling for store traffic
and stockpiling behaviors, discount pack sales decrease by 13 percent following the Cook
County tax increase. This contrasts with the standard difference-in-differences estimates,
which only find a significant effect on premium pack sales. The synthetic difference-in-
differences estimates have greater precision relative to those from the standard difference-in-
differences, due to localization and the removal of systematic heterogeneity in outcomes.
Table 1.4: SDID Tax Increase Effects on Quantities Sold
Treatment Effect S.E. 95% CI
Discount Pack -0.13 0.065 (-0.26, -0.00)
Discount Carton 0.01 0.037 (-0.07, 0.08)
Premium Pack -0.09 0.048 (-0.18, 0.01)
Premium Carton 0.07 0.054 (-0.03, 0.18)
Jackknife standard errors in parenthesis.
Residuals used as dependent variable
1.4.3 Consumer Response
Policy makers should anticipate how taxes will impact the amount of cigarettes sold in
their jurisdictions, as that will directly affect the magnitude of tax revenues generated to
fund public projects. If the objective involves curbing smoking habits, then sales volumes
provide some information on changes in consumption. However, it is important to note
that quantities sold in an area do not directly translate to levels of consumption. Previous
literature has found evidence of illegal smuggling and border crossing to avoid taxation,
Stehr (2005), Merriman (2010), DeCicca et al. (2013a), Chernick and Merriman (2013) so
15
sales data likely underestimates true consumption levels. This also raises the question of
whether an increase in the tax rate is worth potentially losing economic activity to bordering
jurisdictions that have lower tax rates.
Figure 1.3: Stores by Closest Lower-Tax Zone
In this section I consider neighboring jurisdictions and their tax rates, motivated by
previous studies of tax avoidance via border crossing along with the graphical evidence seen
in the weekly store sales. Figure 1.3 maps the stores included in the sample according to
their closest lower-tax zone. If consumers evaluate the taxes imposed on cigarettes as well
as travel costs to purchase the products, stores nearest to borders may have lower quantities
of cigarette sales since shoppers induce smaller travel costs to purchase from stores within
16
the lower-tax zones.
Toexamineconsumerresponsetoexcisetaxesoncigarettes, Iestimatethesalesquantities
for each of the four types of cigarette products. This specification recognizes the possibility
of movements between tax zones to avoid higher rates through the inclusion of the distance
to lower tax border variable and its interaction with the relative tax rate.
log(Q
jt
) =
1
log(T
jt
) +
2
R
ijt
+
3
log(D
j
)R
jt
(1.4)
+
K
X
k=1

k
C
j
k +
s
+
j
+
jt
Q
jt
is the total quantity of product sold at storej during weekt. T
jt
is the excise tax amount
in store j during week t, R
jt
is the relative tax rate, defined as the percent increase that a
consumer would pay in taxes in the home store compared to what they would pay if they
bought the product in the closest lower tax zone at time t. D
j
is the distance of store j to
the closest lower-tax border. C
j
is a vector of control variables describing store j, and
s
and
j
are season and store fixed effects.
Rel:Tax
jt
=
Tax
jt
ClosestLowerTax
jt
ClosestLowerTax
jt
I interact the relative tax rate with log(D
j
) which is the log distance of store j to the
closest lower tax zone border measured in meters. The coefficient on this term,
3
, measures
how a store’s distance to the closest lower tax zone border might affect sales quantities for
different relative excise tax rates. Seasonal dummies are included as cigarette consumption
may adapt to changes in the weather, especially if there are restrictions against smoking
indoors. Forexample,Chicago’sCleanIndoorAirOrdinancehasbeenineffectsince1988and
prohibits tobacco smoking in virtually all enclosed public areas and places of employment.
This means if an individual wishes to smoke in public they must do so outdoors, which may
reduce cigarette consumption during harsher weather seasons.
17
Table 1.5: Cigarette Quantity Estimates
Disc. Packs Disc. Cartons Prem. Packs Prem. Cartons
Tax -1.189** -0.561* -0.868* -0.099
(0.371) (0.249) (0.420) (0.228)
Rel. Tax -0.250 0.490 -0.841 -0.730
(0.783) (0.857) (1.621) (1.219)
Distance: -0.104 -0.175 -0.225 0.127
Rel. Tax (0.221) (0.233) (0.397) (0.340)
Store FE Yes Yes Yes Yes
Controls Yes Yes Yes Yes
Robust standard errors clustered at store-level reported in parentheses
.significant at 10%; *significant at 5%; **significant at 1%; ***significant at 0.1%
Table 1.5 reports the results of equation 1.4 for each of the four types. I discuss the
intuition behind these results in the remainder of this section.
Tax Elasticity
Thecoefficientonlog(T )representstheexcisetaxelasticityofdemandforcigarettessince
it measures the percentage change in quantity demanded caused by a percent change in the
excise tax amount. All product types except premium cartons have negative and significant
coefficients, implying the normalcy of cigarette consumption despite their addictive qualities.
Packs have more elastic demands than cartons for both discount and premium brands,
with discount packs having a coefficient of 1.189, which means that a 1 percent change in the
price results in more than a 1 percent change in quantity demanded. For the same change in
the tax amount, premium pack sales decrease by 0.868 percent. Changes in the tax amount
have no significant effect on premium cartons but discount carton sales decrease by 0.561
percent.
The differences in tax elasticities between packs and cartons are most likely be explained
by a comparison of the types of consumers for these products. Heavier smokers who purchase
cartons of cigarettes are likely less inclined to change their consumption or purchase habits
than more casual smokers buying packs of cigarettes. Another potential difference between
18
consumersofdiscountandpremiumproductsistheirincomelevels. InappendixA.6Iexplore
differences in product market shares across the stores’ neighborhood income levels and find
that premium carton market shares tend to be higher for stores in the upper quantiles of
income.
One possible explanation as to why discount products tend to have greater tax elasticities
than premium products goes back to the flight to quality theory of Barzel (1976). Flight to
quality attributes the differences in responses to excise taxes between quality tiers to changes
in their relative prices since a per-unit excise tax disproportionately affects cheaper prod-
ucts, which experience greater percentage increases in their price relative to more expensive
products. By estimating tax elasticities, I contribute to the empirical debate surrounding
the accuracy of flight to quality Summer and Ward (1981), Sobel and Garrett (1997), Nesbit
(2007), Nargis et al. (2014).
Border-Crossing
Since this study utilizes supply-side scanner data from a supermarket chain across the
Chicago area, I have no way of knowing the actual movements of consumers or where they
live and work. Instead, I assume that most people do their grocery shopping in their own
communities and if they choose to cross borders for the sake of cigarette tax avoidance,
they likely purchase from gas stations, convenience stores, or tobacco shops. Under this
assumption, the distance of a Dominick’s Finer Foods from the closest lower-tax border can
act as a proxy for it’s customer base’s transportation cost required to buy cigarettes in the
lower-tax zone. By interacting the distance variable with the relative tax rate, I’m able to
see whether effects on consumption levels change when the tax differential grows.
The coefficients on the interaction terms between distance and relative tax rate have no
significant effect on sales quantities for any of the product types. In other words, despite
previous literature findings of border-crossing activity, in the current study I find no evidence
of such behavior under my stated assumptions. In appendix A.7, I estimate the quantity
19
specificationswithoutfixedeffectsandinstead includestore-specificcharacteristicsas control
variables to determine whether distance to border has a significant relationship with sales
quantities and find no effect for all except discount cartons, where sales decrease in stores
closer to the border.
1.4.4 Comparison of Results
In section 1.4.2 I use difference-in-differences and synthetic difference-in-differences to
estimate the percent change in quantity sold after the Cook County tax increase for a subset
of the data surrounding the event and in section 1.4.3 I estimate a quantity regression
with tax elasticities for the entire dataset. To compare results across methodologies, I can
calculate the tax elasticities using the results in section 1.4.2.
Tax elasticity measures the percent change in quantity resulting from a percent change
in the tax amount. In 1996, Cook County increased their excise tax amount from $0.10
to $0.18 as seen in table 1.1. This increased the total cigarette excise tax amount for the
Cook County stores from $0.78 to $0.86, or about 10.3% per pack. Since the DID and SDID
estimators represent the percent change in quantities resulting from the tax increase, I can
define the tax elasticities.

tax
=j

10:3%
j
Table 1.6 shows the comparable tax elasticities calculated from the synthetic difference-in-
differences results in section 1.4.2 and those reported in table 1.5.
Table 1.6: Comparing Tax Elasticities
Disc. Packs Disc. Cartons Prem. Packs Prem. Cartons

reg
tax
1.189** 0.561* 0.868* 0.099

sdid
tax
1.262* 0.097 0.874 0.680
The tax elasticities across methods are relatively consistent for the regression and SDID
estimates, however the latter only has a significant elasticity for discount packs. These
20
findingssuggestthattheresponsetotheCookCountytaxincreasedidnotdeviatedrastically
from the usual behaviors seen throughout the 8 year time-span covered by the data.
1.5 Final Thoughts
Cigarette excise taxes are some of the most prevalent sin taxes across the globe. In
recent years, some progressive American cities have attempted to impose similar taxes on
sugary drinks and many have debated the potential benefits and pitfalls to imposing these
taxes. Chicago is home to some of the highest excise tax rates in the nation, so the study
of such taxes in this area comes naturally and understanding demand responses can help
policy makers better serve their communities by properly accounting for resulting behaviors
that can affect the outcomes of their laws. In this paper, I use scanner data from the
Chicago grocery chain, Dominick’s Finer Foods, to examine demand responses to cigarette
excise taxes. As a preliminary exploration, I first estimate the effects of changes in the tax
amounts on that in equilibrium prices to determine tax incidence and find that on average,
consumers pay about $0.76 for a $1.00 increase to the tax rate. These findings align closely
with those reported in existing literature.
In the causal analysis, I examine the effects of the Cook County excise tax increase that
takes effect in 1996. Assuming there are no network effects between the Dominick’s Finer
Food stores in the sample and that cigarette product sales would have evolved similarly
across time absent the tax hike, I calculate difference-in-difference estimators. The results
indicate a 14.9 percent decrease in premium pack sales after the tax increase. Additionally,
I relax the assumption of common trends with one of the first empirical applications of the
Arkhangelsky et al. (2021) synthetic difference-in-differences estimator. This yields results
suggesting a 13 percent decrease in discount pack sales, but no significant change in any of
the other product sales.
To further explore consumer responses, I estimate the quantities of cigarettes sold over
21
timeusingexcisetaxamounts, storedistancetoborder, andrelativetaxratesalongwithcon-
trol variables. I find that tax elasticities are negative and significant for all except premium
cartons, indicating increases in the tax amount leads to decreased product sales.
Despite the findings of previous literature supporting border-crossing behaviors for tax
avoidance, I find no empirical evidence of such movements in this analysis. However, since
the data used in this paper is restricted to the scanner data of Dominick’s Finer Foods, the
lack of evidence may stem from grocery store cigarette sales being less affected by differences
in relative tax rates than the sales of other retail outlets. For a more accurate and current
analysis, using a dataset that contains a wider representation of store types in recent years
would allow for more precise estimates.
22
Chapter 2
The Effects of SALT Deduction on
Homeownership
2.1 Introduction
In 2017, the state and local income tax deduction reduced the taxable income of over 34
million Americans by almost $369 billion making it one of the the government’s most costly
deductions in the tax code. In 2018, these numbers dropped to 12.7 million Americans
reducing their taxabe incomes by $201.5 billion. So what changed? The enactment of the
Tax Cuts and Jobs Act (TCJA) at the end of 2017, which made sweeping changes to the
federal tax code. The law almost doubled standard deductions, decreased the tax rates for
five of the seven personal income brackets, and limited the state and local tax deduction
to $10,000 per single- or married joint-filers. It also reduced the allowance of the mortgage
interest deduction to include no more than $750,000 worth of debt instead of the previous
cap of $1,000,000.
As policy makers negotiate the terms of the Build Back Better plan and we approach the
2025 expiration of many of the TCJA provisions, a large portion of the debate surrounds the
state and local tax (SALT) deduction limit. Proponents of the SALT deduction limit argue
23
that the high costs of the deduction only benefit the wealthy and limiting or abolishing
the it would increase federal tax revenue. Critics claim that the SALT deduction limit
disproportionately affects high-tax and high cost-of-living states, placing an unreasonable
tax burden on the residents of these states.
Arguments surrounding disproportionate geographical taxation is not a new debate. Al-
bouy (2009) studied the unequal burden of federal taxation long before the TCJA passed
and ran simulations to find that taxes lower long-run employment levels by 13 percent, land
prices by 21 percent, and home values by 5 percent in high-wage areas. His model suggests
that tax deductions, such as SALT, help to improve locational inefficiencies. Theoretically-
motivated studies such as Colas and Hutchinson (2021), Fajgelbaum et al. (2019) use spatial
equilibrium models to show that the U.S. income tax policies lead to locational inefficiences
and find that cost-of-living and local wage adjustments tend to mitigate them.
While the aforementioned studies rely on theoretical macroeconomic models to explain
geographic inefficiences, this paper focuses more specifically on empirical estimates of the
effects of federal taxation on homeownership decisions. Differences in state and local tax
rates combined with variation in cost-of-living across the United States lead to varying
levels of the TCJA policy exposure, particularly with regards to the SALT deduction cap.
My analysis focuses on college-educated individuals who have recently moved and uses the
instrumental variable method to show that the SALT deduction limit has minimal average
effects, if any, on the probability of homeownership. I then restrict my analysis to subsets of
the population who may be on the margin of homeownership decisions. I find no evidence of
effects specific to young adults or young and married adults, but I do find strong evidence of
reductions in metropolitan area homeownership probabilities. In these areas, a 10 percent
increaseinaSALTlimitimpactleadstoa0.19percentagepointdecreaseintheprobabilityof
homeownership. Weak evidence of very small average effects combined with strong evidence
of larger effects indicate the SALT deduction limit’s possible contributions to the affordable
housing crises seen in some metropolitan cities already.
24
The results may be overstated if the change to the mortgage interest deduction also
affects housing tenure, although previous literature discussed in section 2.2 suggest that
these effects are small, if at all significant. Additionally, the other aspects of the TCJA such
as its increases in standard deductions and decreases in marginal income tax rates may offset
some of the SALT deduction limit’s effects. However, the inclusion of control variables and
fixed effects in the analysis help in reducing concerns of omitted variable bias.
The remainder of this paper is organized as follows. Section 2.2 reviews related literature
and how they have shaped the current analysis. Section 2.3 provides information regarding
the details of the Tax Cuts and Jobs Act and the SALT deduction. Section 2.4 gives theo-
retical considerations which help motivate the analysis and give insight on the mechanism.
Section 2.5 describes the data. Section 2.6 explains the identification strategy and reviews
the empirical results. Finally, section 2.7 concludes.
2.2 Previous Literature
The current study relates to two broader sets of literature. The first set examines the
impactoftheTaxCutsandJobsAct, withparticularfocusonitsuneveneffectsacrossstates.
It also relates to the strand of literature which analyzes housing tenure and its relationship
with tax deductions, especially the mortgage interest deduction. The Tax Cuts and Jobs
Act made an adjustment to this deduction, which is similar to that of SALT in the context
of housing decisions except it exclusively benefits itemizing homeowners while the SALT
deduction benefits all taxpayers who itemize and pay state and local taxes of any kind.
Since the Tax Cuts and Jobs Act passed, studies have recognized the disproportionate
effects on high-tax, high cost-of-living states. Altig et al. (2020) analyzes the difference
in effects between red and blue states and finds an average lifetime spending increase of
1.6 percent in Republican states as opposed to 1.3 percent in Democrat states, with the
gap widening when focusing on the top 10 percent of households. Coen-Pirani and Sieg
25
(2019) and Fajgelbaumet al. (2019) take theoretical approaches and develop dynamic spatial
equilibrium models to predict the relocation of highly-productive households to low-cost
cities in response to the TCJA, and welfare losses and heterogeneous effects resulting from
theSALTdeductionlimit, respectively. Tong(2021)usesdifferences-in-differenceswithmore
granular, county-level data to show that the SALT deduction cap has a significant negative
effect on home prices in counties where the average SALT deduction exceeds the $10,000
limit. The more expensive homes in these counties felt the largest effects.
While the literature on the TCJA and the SALT deduction cap is limited due to their
recent introductions, previous works examined the effects of owner-occupancy tax treatment
on homeownership rates and housing tenure. Rosen (1979) uses cross-sectional data to
estimatetheeffectsofpersonalincometaxationontenurechoiceandhousingdemand. Rosen
and Rosen (1980) examines the determinants of housing tenure choice and finds that the tax
system’s favorable treatment of home ownership is responsible for about one quarter of the
growth in homeownership rates since World War II. To find the determinants of housing
tenure choice, Goodman (1988) uses a probit model and found that permanent income is of
high significance in the decision.
Particular interest in the effects of the mortgage interest deduction on homeownership
rates and tenure has led researchers to debate the costs and benefits of the policy. Glaeser
(2003) perform a time series analysis to argue that the mortgage interest deduction, which
they describe as an ownership subsidy, has close to no effect on homeownership rates because
marginal groups rarely use the deduction. They show that while the ownership subsiy moves
significantly over time, homeownership rates have remained relatively stable. Following these
results, Bourassa and Yin (2008) study the mortgage interest and property tax deductions’
effectsonyoungadulthomeownershiprates. Theyshowthatalargemarginalgroupchoosing
between renting and owning includes young adults and find that these deductions actually
reduce homeownership rates due to their effects on home prices. Hanson (2012) uses a variety
of empirical methods to find no relationship between the mortgage interest deduction and
26
homeownership rates, although he does find that the deduction leads an increase of up to
18.4 percent in the size of homes purchased.
Previous literature on these topics establish evidence of the dispropotionate effects of
federal tax policies across states based on factors such as state and local tax rates, state
income distributions, and differences in costs-of-living. This motivates the current analysis
to add to the discussion by studying the SALT deduction limit’s effect on homeownership
decisions with special attention to differences in state income tax rates, building on Tong
(2021) results of its effects on home prices. Since the literature also suggests the mortgage
interest deduction’s minimal effect on overall homeownership levels, I restrict my analysis to
college-educated individuals under the assumption that the majority of taxpayers directly
affected by the SALT deduction limit have at least two years of college education, which
highly correlates with income levels. Additionally, in section 2.6.3 I follow Bourassa and Yin
(2008) and repeat my analysis on young adults who are likely on the margin when deciding
between buying and renting.
2.3 Background
In this section, I outline the relevant changes to the tax code following the Tax Cuts and
Jobs Act and discuss the state and local tax deduction and the debate surrounding it.
2.3.1 Tax Cut and Jobs Act of 2017
In 2017 Congress passed the Tax Cuts and Jobs Act (TCJA), with former President
Donald Trump signing it into law in December of that year. This major federal tax overhaul
impacts just about every taxpayer in the country with major changes to both the personal
and business tax codes. The TCJA was met with mixed reviews, partially due to the
polarizing politics present throughout the Trump presidency, but also because certain states
felt disproportionate effects which led 12 House Republicans to vote against their party
27
despite the bill ultimately passing along party lines. Many of the provisions are set to expire
in 2025, unless policy makers choose to extend their lifespans, making economic analysis of
the effects particularly relevant. Since this study focuses on household-level decisions, this
section highlights the main changes to the personal tax code and omits the business tax
alterations which I assume to have little or no direct effect on the following analysis.
One of the biggest changes induced by the Tax Cuts and Jobs Act decreased five of the
seven federal income tax bracket rates, leaving only the bottom and second highest brackets
untouched. This, coupled with standard deductions almost double their 2017 amounts,
simplified the tax filings of many Americans and those benefitting from the newly heightened
standard deductions saw reductions in their 2018 tax payments. The caveat that offset
some of the gains from higher standard deductions came in the form of suspended personal
exemptions, which previously allowed taxpayers to exempt $4,150 from their adjusted gross
income on which they pay taxes. Despite that, the child tax credit increased to $2,000 and
non-child dependants provided an additional $500 in tax credits to their main providers.
Itemized deductions were also heavily impacted by the Tax Cuts and Jobs Act, with
the suspension of many previously deductible expenses such as those associated with mov-
ing, home offices, laboratory breakage fees, licensing and regulatory fees, union dues, and
work clothing. Additionally the limit on mortgage interest deductions for married couples
decreased from allowing up to $1,000,000 worth of debt to only $750,000 worth of debt.
This change primarily impacts the wealthy and home buyers in very high cost-of-living areas
such as Manhattan or San Francisco. The biggest change however, imposed a cap on the
previously unlimited state and local tax deduction. This aspect of the TCJA continues to
be at the center of debates amongst lawmakers as it disproportionately impacts high-income
taxpayers in high-tax states. The remainder of this section discusses the cap to the state
and local tax deduction and its potential effects in further detail.
28
State and Local Tax Deduction
The state and local tax deduction allows taxpayers to deduct the amount of state and
local real property, and income or sales taxes, on their federal tax filings. This deduction was
previously unlimited until the Tax Cut and Jobs Act imposed a temporary cap of $10,000
per individual or married joint-filers through 2025, which hurts high-income taxpayers living
and working in high-tax states. The original Senate bill attempted to eliminate the SALT
deduction entirely, but was amended to allow for the capped amount to win the support of
representatives of high-tax states.
The debate surrounding the SALT deduction limit has remained politically charged, with
some critics accusing the rule of being an attack on left-leaning states. Indeed, previous liter-
ature has established that federal taxation already has disproportionate effects across states,
leading to locational inefficiencies and findings suggest that deductions such as SALT provide
some amount of correction for differences in costs-of-living reducing these inefficiencies. Lim-
iting or abolishing the SALT deduction could lead to losses in total welfare due to relocation
to less-productive areas. The other side argues that the federal government should not grant
leniency to states which choose to impose higher taxes on their residents, particularly given
SALT’s high costs. Perhaps the most unifying argument amongst lawmakers asserts that
the SALT deductions primarily benefit top-earners whom critics believe should bear higher
burdens of taxation.
The heat maps in figures 2.1 and 2.2 show the percent of returns that deducted state
and local income taxes for each state in 2017 and 2018, respectively. The darkest states
in both maps, Alaska, Florida, Nevada, South Dakota, Tennessee, Texas, Washington, and
Wyoming, have no state income taxes and all had less than 5 percent of filers deducting
state and local income taxes in 2017. While the percent of returns deducting state and local
income taxes decreases nationwide, few states such as California, Oregon, Utah, Colorado,
Virginia, and Maryland all remain relatively light in color indicating a significant number
of SALT deductions from these states even after the TCJA increased standard deductions.
29
Figure 2.1: Percent of SALT Deductions by State, 2017
Together, theunevencolorchangesacrossthemapsprovidevisualevidenceoftheinequalities
of federal taxation policies.
To see who benefits the most from SALT deductions, I break down all 2017 tax returns
by income buckets. Figure 2.3 shows the percent of filers who claimed SALT deductions by
incomeusinghistoricalStatisticsofIncomedatafromtheIRS.Thedollaramountaboveeach
bar is the average state and local income tax and real property tax paid by persons who filed
in each of the income buckets. As expected, the average amount deducted increases with
income. Taxpayers with adjusted gross income between $100,000 and $200,000 represent
the largest proportion of deductors, claiming about 26 percent of all SALT deductions with
an average deduction of $12,596 in SALT income and real property taxes. Comparing with
the percent of returns by income in the same year, figure 2.4 shows that the same $100,000
- $200,000 income bucket represents only about 13 percent of all returns despite claiming
30
Figure 2.2: Percent of SALT Deductions by State, 2018
double the percentage of SALT deductions. This confirms that higher-income individuals
tend to benefit more from the deduction, although it doesn’t take varying costs-of-living into
account.
2.4 Theory
To my knowledge, previous literature has not touched on the relationship between the
SALT deduction and homeownership decisions. The closest link relates the mortgage interest
deduction (MID) to housing decisions as briefly discussed in section 2.2 and researchers do
not reach a common consensus regarding the effects on homeownership. The MID allows
taxpayers to deduct the amount of mortgage interest paid on their owner-occupied housing
from their federal taxes each year.
31
Figure 2.3: Percent of SALT Deductions by Income, 2017
The SALT deduction has similar effects to the MID in that it reduces taxable income only
ifanindividualitemizes, anditdecreasesthecostofhomeownershipsinceSALTincludesreal
property taxes. In addition to real property taxes, SALT also includes state and local income
taxes so individuals who itemize in high-tax states may offset some of these tax payments by
using the SALT deduction for their federal taxes. If we assume that an itemizing taxpayer
will first deduct their SALT income taxes then make the decision to buy or rent their home,
then it’s possible to examine the budget constraints in the different SALT cases.
Following Hanson (2012), which examines the tradeoff between housing consumption and
other goods with respect to the MID, I focus on how the SALT deduction limit kinks an
individual’s budget constraints, assuming one does not deduct sales taxes. Three possible
cases help illustrate SALT’s effect on a taxpayer’s budget constraints, assuming they benefit
from itemizing and choose to do so regardless of their homeownership decision. The first
32
Figure 2.4: Percent of Returns by Income, 2017
case is shown in Figure 2.5, where no SALT deduction exists so there is an even tradeoff
between housing and other consumption. This budget constraint also describes the situation
in which an individual’s state income tax alone reaches the $10,000 limit, leaving no room
for real property tax deductions.
Figure 2.6 shows the case in which the taxpayer never feels the SALT impact, meaning
their SALT deduction never exceeds $10,000 due to limited income budget and/or low state
and local tax rates. It also describes the counterfactual, where the SALT deduction is
unlimited as it was prior to the TCJA. In this case, consumption spending costs exceed
housing purchase costs because there are tax incentives to own a home. When comparing
the relative cost of housing consumption to that in figure 2.5, it becomes apparent that the
SALT deduction effectively reduces the relative cost of purchasing housing which could result
in a substitution effect towards homeownership.
33
Figure 2.5: Budget Constraint: No SALT Deduction
Figure 2.7 illustrates the third case where the SALT impact occurs once the individual
crosses a threshold due to real property tax amounts increasing with housing consumption.
The $10,000 limit on the SALT deduction results in a kink in the budget constraint where
the taxpayer hits that limit and recieves no further benefit from SALT deductions. This
occurs when:

Real
P
H
+
State
I = 10; 000
where
Real
is the real property tax rate, which is a percentage of P
H
, the purchase price
of the home,
State
is the effective state income tax rate and I is income. Once housing
consumption reaches the point at which SALT exceeds the $10,000 limit, additional housing
consumption becomes relatively more expensive since the real property tax deduction does
not increase any further. If the kink occurs at a level of housing consumption that is close to
market rates in an area, the SALT impact may negatively affect homeownership decisions.
34
Figure 2.6: Budget Constraint: No SALT Impact
Lastly, I consider how a higher state income tax rate would affect the budget constraint.
Holding all else equal, a higher state income tax rate would effectively shift the entire budget
constraint down and the kink would occur earlier along the housing consumption axis as seen
in figure 2.8. This would reduce overall consumption, and the relative price of housing would
increase at a lower price point. Therefore, higher marginal tax rates and a limit to the SALT
deduction reduces financial benefits of homeowners and their real income since they reach
the maximum SALT benefit earlier and pay more of their income in both state and federal
taxes.
These different scenarios help to illustrate how the SALT deduction limit could affect
housing consumption. The cost of housing plays a significant role in how quickly individuals
reach the kink point in their budget constraint. In markets where even the entry-level homes
comeathighcostsresultinginthousandsofdollarsinrealpropertytaxpayments, purchasing
a home could ensure that even modest-incomes for the area may feel heavy SALT impact.
35
Figure 2.7: Budget Constraint: SALT Impact
This could deter these individuals from choosing to own their homes.
2.5 Data
This paper uses microdata from the U.S. Census and American Community Surveys,
obtained through IPUMS USA to analyze the effects of the SALT deduction limit on home-
ownership decisions. I extract a repeated cross-section of individual-level data spanning 2012
through 2019, excluding 2020 due to the rise of Covid-19, with information such as the in-
dividuals’ demographics, marital status, income, homeownership and metropolitan status.
Using the person’s marital status, state, and income, I calculate an estimate of state in-
come taxes paid by cross-referencing with data on state tax rates from the Internal Revenue
Service (IRS). I then restrict the sample to married heads of households and adult single
persons under the assumption that those who are married with their spouses present file
36
Figure 2.8: SALT Impact & Higher State Income Tax Rate
jointly and other adults who do not fall in that category file individually. Additionally, I
include household survey weights provided in the data to avoid over- or under-representing
particular types of households in my analysis.
To determine an individual’s SALT payment, I combine their real property tax payments,
if any, reported in the IPUMS USA data and state income taxes as calculated using reported
incomes and the state income tax rates. I define the SALT impact as the amount of SALT
payment exceeding the $10,000 deduction limit.
SALT impact =
8
>
>
<
>
>
:

State
+
Real
10; 000;
State
+
Real
10; 000 0
0;
State
+
Real
10; 000< 0
Since the SALT deduction limit primarily affects high-income individuals, I only include
thosewhohaveatleasttwoyearsofcollegeeducation. Indoingso, Iassumethatthesepeople
37
Table 2.1: SALT Impact Summary
Median Mean
Full college-educated sample
SALT exceeding $10,000 $5,985 $9,830
Additional federal taxes $1,536 $2,479
N = 333,032
Age 25-34
SALT exceeding $10,000 $3,911 $6,759
Additional federal taxes $1,074 $1,769
N = 34,176
Age 25-34 & Married
SALT exceeding $10,000 $4,936 $7,937
Additional federal taxes $1,185 $1,905
N = 20,977
Metropolitan Area
SALT exceeding $10,000 $6,062 $9,654
Additional federal taxes $1,562 $2,441
N = 203,217
All samples include only those with 2 years of college education or more.
are more likely to be directly impacted by the SALT deduction limit due to the correlation
between income and education. However, if the housing market is impacted, it is possible
that individuals have homeownership decisions indirectly affected by the SALT deduction
limit, even if their state and local tax payments do not exceed the $10,000 cap. This would
occur if those directly impacted opted to downsize their homes and purchase more affordable
housing than they would have in the environment without the SALT deduction limit. Then,
lower-income people would have to compete with higher-income individuals for the same
homes and force some into renting when they otherwise would have preferred to purchase
their homes.
Table2.1showsasummaryoftheSALTdeductionlimit’simpactasseeninthemicrodata
for affected individuals with at least two years of college education. The average amount
of additional federal taxes assumes that an individual affected by the SALT deduction limit
38
paysamarginaltaxrateof32%ifsingleor24%ifmarried. Thefullsampleaverageadditional
federal tax payment due to the policy change was $1,536. The young respondents had the
leastimpactofthesubgroupsinthetablewhilemetropolitanhouseholdershadhighermedian
impacts than the full sample.
Lastly, I use data from the Pew Research Center on political party affiliation by state
from 2014 as a measure of a state’s party leanings. Using pre-policy data ensures that the
measure is not endogenous, capturing resident reactions to the policy’s effects. Political
party affiliation is central to the identification strategy used in this study which is described
in the following section.
2.6 Empirical Strategy
In this section I go through the empirical strategy and discuss the identification, method-
ology, and results of the study. After examining the average treatment effect across the full
sample of college-educated respondents, I then consider whether the magnitude of effects
change for particular subgroups of the population.
2.6.1 Identification Strategy
The primary focus of this analysis is to determine the effects of the SALT deduction limit
on choice of homeownership, with varying levels of policy exposure across states due to differ-
ences in income levels, costs of living, and state and local tax rates. Ideally, I could regress an
individual’s SALT impact directly on their homeownership decision to analyze these effects
but those impacted by the SALT deduction limit likely pay high taxes already and may have
more incentive to purchase their homes, resulting in possible endogeneity. Additionally, the
SALT impact measurement includes property taxes which individuals only pay if they own
real estate. For these reasons, I use an instrumental variable to capture variation in SALT
impact across states and individuals to determine the effects on homeownership.
39
All else equal, a taxpayer’s state of residence determines their exposure to the SALT
deduction limit. Naturally, a state’s income and property tax rates play a large role in
determining an individual’s SALT cap impact. Table 2.2 provides some descriptive statistics
of states grouped by political affiliation. Democrat-leaning states tend to have higher state
income tax rates, with a mean top marginal tax rate of 6.10 percent, while that of the
Republican leaning states is 4.85 percent. The 2018 median household income of Democrat-
leaning states also exceeds Republicans by $12,800. The combination of higher tax rates
and greater income suggests taxpayers residing in Democrat-leaning states tend to pay more
in state and local income taxes. Additionally, the higher population density in Democrat-
leaning states likely correlates with greater home values due to the scarcity of land. This
would result in higher real property taxes, even at comparable rates.
Table 2.2: Descriptive Statistics by State Political Affiliations
Democrat Republican All
Top Marginal State Income Tax Rate 6.10% 4.85% 5.75%
Household Income $99,800 $87,000 $96,000
Property Tax $3,000 $1,000 $2,000
Population Density 6,584 1,894 5,380
Political affiliation of states determined by party with larger proportion of affiliants in 2014. Vari-
ables measures by median of 2018 respondents. Property taxes rounded to the nearest thousand.
Population density is persons per square mile.
Motivated by these characteristics, as well as the results of previous literature, I use the
interaction of a year (or post-policy) dummy and the state of residence’s 2014 Democrat-
identifying percent of population as an instrument for an individual’s SALT cap impact.
z
it
= Post-Policy
t
Percent Democrat-Affiliated
2014;i
The post-policy dummy variable indicates whether the respondent answered the survey in a
year after the SALT deduction limit became enforced. The political affiliation captures the
individual’s exposure to the SALT deduction limit given a fixed income. In other words, if
40
the individual lives amongst more Democrats, they likely pay higher state and local income
taxes than they would if they lived in a state with less Democrats. Being subject to higher
state and local tax rates means these individuals are more likely to have a higher SALT
impact.
The main identifying assumption behind this instrument is that people have comparable
preferences for homeownership in Democrat and Republican states in the absence of SALT.
Possible threats to the validity of this assumption could include the argument that residents
of Democrat-leaning states tend to have higher mobility so they may not be as inclined to
’settle down’ and purchase homes. Differences in demographics may also affect preferences
for homeownership. The inclusion of control variables and state-level fixed effects mitigate
concerns of omitted variable bias. Another possible argument against the validity of the
instrument is that it may correlate with other aspects of the Tax Cuts and Jobs Act and
through that, homeownership decisions. In that case, the effects described by the analysis
result from the entirety of the tax overhaul, with SALT cap impact acting as a proxy for an
individual’s overall exposure.
The instrument as described is exogenous to a person’s SALT deduction impact because
in 2014 the tax code allowed for unlimited SALT deductions so a household’s choice of
geographic residence would not have been affected by the TCJA. A problem would arise if
people moved states in response to the TCJA to reduce their tax burden, which would imply
endogeneity of the instrument. However, the decision to move states typically requires
planning and since this analysis focuses on short-run effects with only two years of data
following the policy change, migration between states in response to the TCJA is plausibly
negligible in this context. Additionally to address this concern, I restrict the sample to
individuals who have moved in the past year, but only within their same state, as this
captures the homeownership decision made after the policy enactment. This eases concerns
of endogeneity stemming from individuals who may have moved in response to the new SALT
deduction limit.
41
The basis of this analysis relies on the assumption that choice of homeownership is not
systematically different for people in states with differing levels of political preferences in the
absenceoftheSALTdeductionlimit. Inotherwords, peopleinDemocrat-leaningstateshave
comparable homeownership preferences to those in Republican-leaning states, all else equal.
I check this assumption by looking at differences in homeownership before and after the
SALT deduction limit. Table 2.4 shows the mean ownership in Democrat- and Republican-
leaning states in the years before and after the Tax Cuts and Jobs Act was implemented. In
both time periods, people in Democrat states had slightly lower probabilities of owning than
those in Republican states. The same holds true in table 2.3, however we see a much smaller
difference-in-difference estimate when comparing homeownership for the people who did not
move homes in the past year. These findings support the identifying assumption that choice
of homeownership does not differ systematically for people in states with different political
preferences and provides some evidence of a possible effect following the Tax Cuts and Jobs
Act.
Table 2.3: Political Leanings on Ownership (Same House)
Democrat Republican Difference
2018-2019 0.8016 0.8412 -0.0359
(0.0004) (0.0007) (1.40e-06)
2016-2017 0.7994 0.8384 -0.0390
(0.0004) (0.0007) (2.23e-12)
Difference 0.0022 0.0028 -0.0005847
Table 2.4: Political Leanings on Ownership (Moved Within State)
Democrat Republican Difference
2018-2019 0.4782 0.5157 -0.0375
(0.0018) (0.0031) (1.99e-05)
2016-2017 0.4694 0.5037 -0.0343
(0.0018) (0.0031) (4.28e-10)
Difference 0.0088 0.012 -0.0033
42
2.6.2 Effect of SALT Deduction Limit on Home Ownership
Based on the identifying assumption and instrumental variable introduced in section
2.6.1, I proceed to use two-stage least squares (2SLS) estimation to measure the causal effect
oftheSALTdeductionlimitonchoiceofhomeownership. Thefirst-stageregressionestimates
the relationship between the instrument and a person’s SALT impact:
S
ijk
=c
1
+
1j
+
1k
+
2019
X
l=2018
(D
j
I
il
)
l
+C
i

1
+
ijk
(2.1)
where S
ijk
is the log SALT impact of person i in state j in year k. State and year fixed
effects are represented by
1j
and
1k
, respectively. D
j
is the percent of Democrat residents
in statej, which is interacted with dummy variables for each post-policy year. C
i
is a vector
of control variables characterizing person i and
ijk
is the error term.
An alternative variant of the instrument interacts the proportion of Democrat-leaning
residents in the state to a post-policy dummy variable, Post
k
, as in equation 2.2.
S
ijk
=c
1
+
1j
+
1k
+ (D
j
Post
k
) +C
i

1
+
ijk
(2.2)
In both specifications, the coefficient on the instrument, , represents the percentage in-
crease in SALT impact resulting from a percentage point increase in the Democrat-leaning
population of the home state.
Table 2.5 shows the first-stage estimates from equations 2.1 and 2.2. After the SALT
deduction limit of $10,000 becomes enforced, the coefficients for percent Democrat and years
2018 and 2019 are positive and significant. Likewise, when using a single post-policy dummy
in place of individual year dummies for the instrument, the coefficient remains positive and
highly significant. This indicates that, on average, the SALT impact of college-educated
people increase by about 10 percent when an additional percent of the 2014 state population
affiliates with Democrats. These results follow those of related literature which has found
that the SALT deduction affects Democrat states more than Republican states, particularly
43
for residents with higher incomes.
Table 2.5: Political Leanings on SALT Impact
Instrument (1) (2)
Post*Party 0.1001***
(0.0180)
2019*Party 0.1027***
(0.0197)
2018*Party 0.0974***
(0.0167)
F-Statistic 18.06 31.10
R
2
0.186 0.186
Dependent variable is log(SALT). Standard errors are clustered at state-level and reported in paren-
theses.
Given the relevance of the instrument in the first stage regression, I next move onto the
second stage and regress the predicted SALT impact on homeownership. The second stage
regression takes the form of a linear probability model due to the binary dependent variable,
which indicates whether the individual rents or owns their home.
H
ijk
=c
2
+
2j
+
2k
+
k
^
S
ijk
+C
i

2
+
ijk
(2.3)
where H
ijk
is the homeownership decision which takes a value of 1 if person i owns their
home and 0 if they rent their home and
^
S
ijk
is the predicted SALT cap impact from equation
2.1. Since I estimate a probability by regressing the log-transformed SALT cap impact
variable, the 2SLS coefficient represents the percentage point increase in the probability of
homeownership from a one percent increase in SALT impact.
The control variables included in the analysis stem from the housing tenure literature. In
previous studies, researchers have modeled the probability of homeownership using probit,
logit, and linear probability models. Most studies include a similar set of control variables
measuring a person’s demographics and household composition. Due to the lack of precise
data, I exclude the user cost of owning relative to renting from my analysis and instead
include state- and year- fixed effects to account for bias stemming from omitted variables.
44
Table 2.6: Effect of SALT Deduction Limit on Homeownership
Instrument (1) (2) (3) (4)
Panel A: Recently moved college sample
Post-policy dummy*democratic -0.0069. -0.0076* -0.0073. -0.0066.
percent (0.0040) (0.0038) (0.0037) (0.0037)
Year dummy* democrat -0.0068. -0.0075* -0.0072. -0.0065.
percent (0.0040) (0.0038) (0.0037) (0.0037)
Panel B: Same house college sample
Post-policy dummy*democratic -0.0011 -0.0010 -0.0008 -0.0009
percent (0.0008) (0.0008) (0.0008) (0.0007)
Year dummy* democrat -0.0011 -0.0010 -0.0007 -0.0009
percent (0.0008) (0.0008) (0.0008) (0.0007)
Control variables
Marital status No Yes Yes Yes
Number of Children No Yes Yes Yes
Age No No No Yes
Metropolitan status No Yes Yes Yes
Black No No Yes No
Hispanic No No Yes No
Dependent variable is ownership of dwelling, which takes a value of 0 if they own and 1 if they rent.
Robust standard errors are in parentheses.
Besides that, I control for the other characteristics commonly included in the housing tenure
model literature such as marital status, number of children, age, metropolitan status, and
race.
Since I am interested in how the SALT deduction limit impacts homeownership decisions,
I estimate the effects for all persons who moved to a new home within the same state in
the year before they were surveyed. The results are reported in panel A of table 2.6. The
estimates provide weak evidence of minimal effects, with 5-percent level of significant only
the second specification. Based on the second specification’s results, a 10 percent increase in
a person’s SALT impact leads to a 0.076 percentage point decrease in the probability of that
person choosing to own their home rather than rent it. These magnitudes of the estimates
remain relatively constant for the different combinations of the control variables, however
the effects are statistically insignificant for the other specifications.
45
To ensure these results are not driven by unobserved factors related to the instrument,
I perform a falsification test by running an identical analysis on respondents who have not
moved homes in the year before they were surveyed and therefore made their homeownership
decisions prior to the TCJA. I report the results of this control experiment in panel B of
table 2.6. Unlike in the analysis on the recently moved sample, the 2SLS coefficients on the
SALT cap impact are insignificant in all specifications, which suggests my analysis and use
of the instrumental variable are valid in the context of this study. Based on that, I argue
that any effects found in panel A are indeed due to changes in the personal tax policy.
The results shown in table 2.6 resemble those in the literature studying the effects of the
mortgage interest deduction on homeownership. In other words, the magnitude of effects of
theSALTdeductionlimitareminimal. Thisleadstoquestionsaboutwhosebehaviorchanges
in response to the SALT deduction limit. In areas with surging house prices, taxpayers on
the margin may get priced out while those more heavily impacted end up downsizing or not
feeling the effects as much due to an excess of disposable income. In the following section,
I consider the effect of the SALT deduction limit for households on the margin of their
homeownership decision.
2.6.3 Who is Primarily Affected?
Section 2.6.2 establishes a weak causal link between SALT impact and homeownership
decisions for all college education respondents who moved in the past year, however, house-
holds on the margin likely experience stronger effects than those with state and local taxes
well above the limit. Following Bourassa and Yin (2008), in this section I focus more closely
on the effects of the SALT deduction limit on young adults between the ages of 25 and 34, as
they are likely first-time homebuyers if they choose to own. I also look separately at young
married adults since the SALT deduction limit does not increase for married joint-filers.
Lastly, I consider metropolitan dwellers both within the principle city and outside as these
areas tend to be more densely populated with higher costs-of-living and incomes. I report
46
results for the three subsets of the population in table 2.7.
Table 2.7: Effect of SALT Deduction Limit on Homeownership
Instrument (1) (2) (3) (4)
Panel A: Young
Post-policy dummy*democratic -0.0075 -0.0095 -0.0086 -0.0097
percent (0.0074) (0.0071) (0.0070) (0.0072)
Panel B: Young & Married
Post-policy dummy*democratic -0.0150* -0.0129. -0.0117. -0.0126.
percent (0.0072) (0.0070) (0.0068) (0.0070)
Panel C: Metropolitian Area
Post-policy dummy*democratic -0.0187** -0.0195*** -0.0192*** -0.0188***
percent (0.0060) (0.0057) (0.0057) (0.0057)
Year dummy* democrat -0.0187** -0.0195*** -0.0192*** -0.0188***
percent (0.0060) (0.0057) (0.0057) (0.0057)
Control variables
Marital status No Yes Yes Yes
Number of Children No Yes Yes Yes
Age No No No Yes
Metropolitan status No Yes Yes Yes
Black No No Yes No
Hispanic No No Yes No
Dependent variable is ownership of dwelling, which takes a value of 0 if they own and 1 if they rent.
Robust standard errors clustered at state-level in parentheses.
Relative to the findings in section 2.6.2, the magnitudes of the 2SLS coefficients estimated
for the young adults and the young married adults are greater than those of the full sample,
however almost all specifications are statistically insignificant. Restricting the sample to
those living in metropolitan areas, whether within or outside of the principal city, results in
much stronger and statistically significant estimates. The magnitudes of the metropolitan
area sample almost triple those of the full sample, with a 10 percent increase in SALT limit
impact leading to a 0.187 percentage point decrease in the probability of the respondent
owning their home. This suggests that the responses of metropolitan households strongly
contribute to the average overall effects.
Assuming metropolitan areas have relatively higher costs-of-living and taxes, these re-
47
sults confirm the notion that the SALT deduction limit more heavily impacts high-income
individuals in high-tax states (and localities). If it reduces homeownership rates in densely
populated metropolitan areas that have low supplies of housing relative to demand, the
SALT deduction limit may also contribute to the affordable housing crises in cities which
already struggle to house their residents. While the policy may not necessarily reduce the
total supply of housing in these areas, it might lead to lower owner-occupancy rates if private
investors and companies have greater purchasing power and keep renters from entering the
home buying market.
2.7 Final Thoughts
The Tax Cuts and Jobs Act imposed a $10,000 limit on the state and local tax deduction
which resulted in disproportionate effects across states with varying tax rates and a heavier
burden on high-income households in high-tax states. Amongst the college educated, empir-
ical estimates suggest that a one percentage point increase in Democrat-identifying residents
in a person’s home state correlates with a 10 percent increase in SALT cap impact. Based
on this relationship and the identifying assumption that people have comparable preferences
for homeownership across states with different levels of political party affiliations absent the
SALT limit, I use 2SLS to estimate the effect of the SALT deduction limit on homeownership
decisions.
The 2SLS results for the college-educated sample provide mild evidence that a 10 percent
increase in SALT cap impact leads to a 0.07 percentage point decrease in the probability
of owning a home. When focusing on the young adult and young married adult subgroups
withinthepopulation,Ifindnosignificantevidenceofanimpactonhomeownershipdecisions.
However, the magnitude of effect increases by almost three times that of the full sample when
I restrict the analysis to persons living in a metropolitan area, which includes those within
and outside of the principal city. These results suggest that the SALT deduction limit has
48
little effect on homeownership decisions when averaged across all impacted taxpayers, but it
may have a more relevant impact in metropolitan areas, many of which have a shortage of
housing supply already.
Thisanalysisonlycontainstwopost-policyyearssofindingsarelimitedtotheshort-term.
If prices adjust to the policy, the effects may diminish or even become more homeowner-
friendly in the long-run. This outcome seems more likely after Covid-19 led many employees
to remote work. Since then, ease of mobility has allowed more remote workers to live in
lower-tax and lower cost-of-living states while still keeping the same jobs.
49
Bibliography
Albouy, D. (2009). The unequal geographic burden of federal taxation. Journal of Political
Economy, 117 (4), 635–667.
Altig, D., Auerbach, A., Higgins, P., Koehler, D., Kotlikoff, L., Terry, E. and
Ye, V. (2020). Did the 2017 tax reform discriminate against blue-state voters? National
Tax Journal, 73 (4), 1087–1108.
Arkhangelsky, D., Athey, S., Hirshberg, D. A., Imbens, G. W. and Wager, S.
(2021). Synthetic difference-in-differences. American Economic Review, 111 (12), 4088–
4118.
Barzel, Y. (1976). An alternative approach to the analysis of taxation. Journal of Political
Economy, 84 (36), 1177–1197.
Bourassa, S. C. and Yin, M. (2008). Tax deductions, tax credits and the homeownership
rate of young urban adults in the united states. Urban Studies, 45 (5&6), 1141–1161.
Chernick, H. and Merriman, D. (2013). Using littered pack data to estimate cigarette
tax avoidance in nyc. National Tax Journal, 66 (3), 635–668.
Chiou, L. and Muehlegger, E. (2008). Crossing the line: Direct estimation of cross-
border cigarette sales and the effect on tax revenue. The B.E. Journal of Economic Anal-
ysis and Policy, 8 (1).
— and — (2014). Consumer response to cigarette excise tax changes. National Tax Journal,
67 (3), 621–650.
Coen-Pirani, D.andSieg, H.(2019).Theimpactofthetaxcutandjobsactonthespatial
distribution of high productivity households and economic welfare. Journal of Monetary
Economics, 105 (C), 44–71.
Colas, M. and Hutchinson, K. (2021). Heterogeneous workers and federal income taxes
in a spatial equilibrium. American Economic Journal: Economic Policy, 13 (2), 100–134.
Cornelius, M. E., Driezen, P., Fong, G. T., Chaloupka, F. J., Hyland, A.,
Bansal-Travers, M., Carpenter, M. J. and Cummings, K. M. (2014). Trends
in the use of premium and discount cigarette brands: findings from the itc us surveys
(2002–2011). Tobacco Control, 23 (suppl 1), i48–i53.
50
DeCicca, P.,Kenkel, D. andLiu, F. (2013a). Excise tax avoidance: The case of cigarette
excise taxes. Journal of Health Economics, 32, 1130–1141.
—, — and — (2013b). Who pays cigarette taxes? the impact of consumer price search. The
Review of Economics and Statistics, 95 (2), 516–529.
Espinosa, J. and Evans, W. N. (2013). Excise taxes, tax incidence, and the flight to
quality: Evidence from scanner data. Public Finance Review, 41 (2), 147–176.
Fajgelbaum, P. D., Morales, E., Suarez Serrato, J. C. andZidar, O. (2019). State
taxes and spatial misallocation. Review of Economic Studies, 86 (1), 333–376.
Glaeser, S. J. M., Edward L. (2003). The benefits of the home mortgage interest de-
duction. Tax Policy and the Economy, 17, 37–82.
Goodman, A. C. (1988). An econometric model of housing price, permanent income, tenure
choice, and housing demand. Journal of Urban Economics, 23, 327–353.
Hanson, A. (2012). Size of home, homeownership, and the mortgage interest deduction.
Journal of Housing Economics, 21, 195–210.
Harding, M., Liebtag, E. and Lovenheim, M. F. (2012). The heterogeneous geographic
and socioeconomic incidence of cigarette taxes: Evidence from nielsen homescan data.
American Economic Journal: Economic Policy, 4 (4), 169–198.
Merriman, D. (2010). The micro-geography of tax avoidance: Evidence from littered
cigarette packs in chicago. American Economic Journal: Economic Policy, 2 (2), 61–84.
Nargis, N., Fong, G. T., Chaloupka, F. J. and Li, Q. (2014). The choice of discount
brand cigarettes: A comparative analysis of international tobacco control (itc) surveys in
canada and the united states (2002-2005). Tob Control, 23 (0 1).
Nesbit, T. (2007). Excise taxation and product quality: The gasoline market. Economic
Issues, 12 (2), 11–14.
Orzchowski, W. and Walker, R. C. (2013). The Tax Burden on Tobacco. Tech. rep.,
Orzchowski and Walker.
Rosen, H. S. (1979). Housing decisions and the u.s. income tax. Journal of Public Eco-
nomics, 11, 1–23.
— and Rosen, K. T. (1980). Federal taxes and homeownership: Evidence from time series.
Journal of Political Economy, 88 (1), 59–75.
Sobel, R. S. and Garrett, T. A. (1997). Taxation and product quality: New evidence
from generic cigarettes. Journal of Political Economy, 105 (4), 880–887.
Stehr, M. (2005). Cigarette tax avoidance and evasion. Journal of Health Economics, 24,
277–297.
51
Sullivan, R. S. and Dutkowsky, D. H. (2012). The effect of cigarette taxation on prices:
An empirical analysis using local-level data. Public Finance Review, 40 (6), 687–711.
Summer, M.T.andWard, R.(1981).Taxchangesandcigaretteprices.JournalofPolitical
Economy, 89 (6), 1261–1265.
Tong, Z. Y. (2021). The impact of the state and local tax (salt) deduction cap on u.s. home
prices. OCC Working Papers - New Frontiers in Bank Risk, Working Paper.
52
Appendix A
Appendix to Chapter 1
A.1 Cigarette Summary Statistics
Table A.1 contains some summary statistics for the scanner data. These were calculated by
aggregating the sales data by store-weeks and computing the statistics for subsets of products’
observed sales and prices. The statistics for distance to closest taxzone were calculated based on
the geographical locations of the 84 stores included in the sample and are measured in kilometers.
Table A.1: Summary Statistics
Mean Std. Dev. Min Max
Packs Sold 8283.811 8015.001 25 27349
Cartons Sold 1055.294 1187.341 19 5365
Discount Packs Sold 809.2771 265.6767 25 1675
Discount Cartons Sold 138.7337 71.67661 19 292
Premium Packs Sold 15758.34 4064.586 844 27349
Premium Cartons Sold 1971.854 1064.694 589 5365
Pack Price 2.308015 0.2388754 1.861538 2.818185
Carton Price 18.90927 1.892866 13.89248 22.85148
Discount Pack Price 2.1548 0.2000707 1.861538 2.633467
Discount Carton Price 18.02369 1.648922 13.89248 20.90583
Premium Pack Price 2.46123 0.1648551 2.159271 2.818185
Premium Carton Price 19.79486 1.698085 15.04429 22.85148
Distance to Closest Taxzone 4690.109 4702.859 18.05148 25893.6
Note: Quantities sold and average prices are at the week-level.
Distance to closest taxzone is measured in meters.
53
A.2 Difference-in-Differences: Falsification Test
The difference-in-differences specification requires the identifying assumption that stores sales
would have evolved similarly absent the Cook County tax increase. To argue the plausibility of this
assumption, I perform a falsification test to see whether the Cook County tax increase had an effect
on total grocery sales. Since Dominick’s Finer Foods is a supermarket chain, the usefulness of this
falsification test relies on the assumption that consumers have not changed their grocery store habits
due to an increase in cigarette excise taxes, only their cigarette purchasing habits. The results of the
difference-in-difference estimation that uses total grocery sales as the dependent variable instead of
quantity of cigarette products sold is shown in table A.2.
Table A.2: Falsification: Grocery Sales DID
Dependant Variable Treatment Quarter DID
log(Total Grocery Sales) 269.02 -0.003 -0.13
(271.28) (0.007) (0.138)
.significant at 10%; *significant at 5%; **significant at 1%; ***significant at 0.1%
Standard errors clustered at the store level in parenthases.
54
A.3 Synthetic Difference-in-Differences Plots
The following plots illustrate the residuals of synthetic control groups relative to those of treated
stores which experience the Cook County tax increase. The control groups are generated by the
synthetic difference-in-difference methodology through the time and unit weights as described in
section 1.4. The differences between the dotted blue lines and the solid blue lines represent the
SDID estimates.
Figure A.1: SDID Discount Pack Sales Residuals
55
Figure A.2: SDID Discount Carton Sales Residuals
56
Figure A.3: SDID Premium Pack Sales Residuals
57
Figure A.4: SDID Premium Carton Sales Residuals
58
A.4 Traditional Difference in Differences
One of the main drawbacks plaguing the difference-in-differences approach is consumers’ antici-
pation of the tax increase. If consumers respond to the announcement of the tax increase before the
change actually occurs, then the estimator is likely biased. I respond to this in the main analysis
by including stockpiling dummy variables and also weekly grocery sales. To check whether the
inclusion of such variables affect the estimates, I report the unconditional difference-in-differences
estimates below with no controls.
Table A.3: Tax Increase on Quantities Sold

did
Discount Packs -0.198***
(0.045)
Discount Cartons -0.044.
(0.025)
Premium Packs -0.361***
(0.051)
Premium Cartons -0.042
(0.051)
.significant at 10%; *significant at 5%; **significant at 1%; ***significant at 0.1%
The estimates in table A.3 exhibit downwards biases relative to those in the main analysis with
the control variables.
59
A.5 Synthetic Difference in Differences
(^
sdid
;^ ;^ ;
^
) = argmin
;;;
N
X
j=1
T
X
t=1
(Q
jt

j

t
W
jt
)
2
^ !
sdid
j
^

sdid
t
(A.1)
I also estimate the synthetic difference-in-differences estimators directly, without taking control
variables into account with equation A.1. In this specification, I observe differences in the quantities
sold rather than those in the residuals of predicted quantities. These estimates are less precise than
those in the main analysis and, like the standard difference-in-differences estimates in appendix A.4,
they seem to have a downwards bias without the control variables.
Table A.4: SDID Tax Increase Effects on Quantities Sold
Treatment Effect S.E. 95% CI
Discount Pack -0.25 0.09 (-0.42, -0.08)
Discount Carton -0.05 0.05 (-0.14, 0.04)
Premium Pack -0.36 0.15 (-0.65, -0.07)
Premium Carton -0.32 0.88 (-2.05, 1.41)
Jackknife standard errors in parenthesis.
log(Quantity) is the dependent variable.
60
Figure A.5: SDID Overall Store Cigarette Product Sales
61
Figure A.6: SDID Discount Pack Sales
62
Figure A.7: SDID Discount Carton Sales
63
Figure A.8: SDID Premium Pack Sales
64
Figure A.9: SDID Premium Carton Sales
65
A.6 Market Shares of Cigarette Product Types
In the following plots, I separate the stores based on their neighborhood characteristic’s income
quantiles and graph the market shares of the four product types over time. The vertical lines
represent tax changes, with the Cook County tax increase being the last in the dataset. The lower-
income quantile stores seem to experience more dramatic stockpiling behavior relative to higher-
income quantile stores, purchasing higher quantities of discount packs before the Cook County tax
increase and after more premium packs. Additionally, the higher-income stores tend to have larger
market shares of premium cartons, which could explain why the excise tax elasticities for those
products are insignificant in the main analysis. Consumers with more money are less likely to
change their behaviors due to tax increases than those with less money.
66
Figure A.10: Market Shares of First Income Quartile
67
Figure A.11: Market Shares of Second Income Quartile
68
Figure A.12: Market Shares of Third Income Quartile
69
Figure A.13: Market Shares of Fourth Income Quartile
70
A.7 Cigarette Quantity Estimates
In the central analysis, I include store fixed effects which prohibits the inclusion of time-invariant
store characteristics in the regression. To understand the relationship between certain store charac-
teristics and cigarette sales quantities, I now exclude store fixed effects and use each store’s market
characteristics of its surrounding neighborhoods as independent variables in the analysis.
Table A.5: Cigarette Quantity Estimates
Disc. Packs Disc. Cartons Prem. Packs Prem. Cartons
Tax -1.203*** -1.409*** -0.214 -1.048*
(0.332) (0.381) (0.422) (0.493)
Rel. Tax 0.859 -0.923* 0.267 -1.293*
(0.571) (0.446) (0.411) (0.567)
Distance 0.060 -0.204* 0.082 -0.126
(0.092) (0.081) (0.115) (0.105)
Distance: -0.428* 0.209 -0.187 0.305
Rel. Tax (0.198) (0.142) (0.186) (0.201)
Store FE No No No No
Controls Yes Yes Yes Yes
Robust standard errors clustered at store-level reported in parentheses
.significant at 10%; *significant at 5%; **significant at 1%; ***significant at 0.1%
Store distance to the closest lower-tax border has a negative and significant effect on the sales
quantities of discount carton products but is insignificant for all other products. This means that
if a store is closer to the border, it generally sells fewer discount cartons than it would if it were
farther from the border but other cigarette product sales remain unchanged.
Table A.6 shows the quantiles of relative tax rates seen in the data, with half of all stores having
a tax amount that is at least 21.7 percent higher than the closest lower tax zone. The coefficient on
relative tax rate is negative and significant for both discount and premium cartons, indicating that
larger tax differentials between the current store and its closest lower tax zone lead to lesser carton
sales in the current store.
These results exhibit mild support for the border-crossing hypothesis, where consumers travel
across borders to purchase cheaper products due to tax differentials. However, insignificant coeffi-
cients on the interaction term in the carton specifications suggest that the effects of distance does
71
not change with different levels of relative tax rates. While I am unable to verify this as an expla-
nation for my findings due to the nature of my supply-side data, previous literature has studied this
type of consumer response to cigarette excise taxes and have found that border-crossing is quite
common, especially in areas of high taxation Chernick and Merriman (2013), DeCiccaet al. (2013b),
Stehr (2005), Merriman (2010).
Table A.6: Relative Tax Quantiles
0% 25% 50% 75% 100%
0.0625 0.1667 0.2174 0.4085 1.5823
72

Abstract (if available)

Abstract This thesis comprises of two research papers which empirically explore the consumption and behavioral responses to two different, real-world taxation policies. The first chapter measures the causal effect of an increase to the cigarette excise tax on the quantities of cigarettes sold using supply-side scanner data from a Chicago supermarket chain. It also estimates excise tax elasticities. The second chapter examines the effect of the state and local tax (SALT) deduction limit on individuals’ homeownership decisions using a unique instrumental variable to measure a taxpayer’s SALT impact in a linear probability model for homeownership.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Essays on causal inference

PDF

Essays on econometrics

PDF

Essays on applied microeconomics

PDF

Three essays on the identification and estimation of structural economic models

PDF

Essays on treatment effect and policy learning

PDF

Essays on the estimation and inference of heterogeneous treatment effects

PDF

Essays on effects of reputation and judicial transparency

PDF

Essays on competition and antitrust issues in the airline industry

PDF

Essays on estimation and inference for heterogeneous panel data models with large n and short T

PDF

Sustaining open source software production: an empirical analysis through the lens of microeconomics

PDF

Essays on nonparametric and finite-sample econometrics

PDF

Essays on competition and strategy within platform industries

PDF

Three essays on the evaluation of long-term care insurance policies

PDF

Leveling the playing field: unbiased tests of the relative information content of book income and taxable income

PDF

Two essays in econometrics: large N T properties of IV, GMM, MLE and least square model selection/averaging

PDF

Essays on family planning policies

PDF

Applications of Markov‐switching models in economics

PDF

Essays on development and health economics: social media and education policy

PDF

Essays in information economics and marketing

PDF

Empirical essays on industrial organization

Asset Metadata

Creator Lee, Rachel Alexis (author)

Core Title Essays on the microeconomic effects of taxation policies

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Degree Conferral Date 2022-05

Publication Date 04/05/2022

Defense Date 03/02/2022

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

Tag applied microeconomics,causal inference,economics,excise taxes,microeconomics,OAI-PMH Harvest,Taxation

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Ridder, Geert (committee chair), Hsiao, Cheng (committee member), Nix, Emily (committee member)

Creator Email lee962@usc.edu,rachlee93@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC110879196

Unique identifier UC110879196

Document Type Dissertation

Format application/pdf (imt)

Rights Lee, Rachel Alexis

Type texts

Source 20220406-usctheses-batch-919 (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