University of Southern California Dissertations and Theses

Essays on the economics of climate change adaptation in developing countries

(USC Thesis Other)

Essays on the economics of climate change adaptation in developing countries

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content ESSAYS ON THE ECONOMICS OF CLIMATE CHANGE ADAPTATION
IN DEVELOPING COUNTRIES
by
Mohammad Islamul Haque
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulllment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ECONOMICS)
May 2023
Copyright 2023 Mohammad Islamul Haque
Dedication
I dedicate this thesis to my wife, Deedhiti Dola, my sister, Runa Laila, my mother, Laijuara
Khanam, and my grandmother, Laily Begum. Without their unconditional support and
unwavering trust in me, it would have never been possible to come this far. I am forever
grateful for their presence in my life.
ii
Acknowledgements
I would like to express my heartfelt gratitude to my main advisors, Matthew Kahn and
Paulina Oliva, for their unwavering support and guidance throughout my research journey.
Their invaluable feedback and encouragement have been instrumental in improving my re-
search. I am also deeply grateful to my advisor, Richard Green, for his guidance during
this journey. I would like to acknowledge the the support that I have recieved from Jerey
Weaver and Vittorio Bassi from the very beginning of this research project. Their valuable
feedback has been immensely helpful in improving the quality of my work. I am also grateful
to Bosen Shao for being an amazing coauthor. I would like to thank Prashant Bharadwaj,
Teevrat Garg, Vernon Henderson, Mark Jacobsen, Rajat Khochar, Rob Metcalfe, Ahmed
Mushq Mobarak, Nicholas Ryan, and Ruozi Song for their helpful comments and advice.
Finally, I am grateful to all the participants of my talks at USC Applied Micro Reading
Group Meetings, Yale Development Lunch Seminar, UCSD Environmental Economics Sem-
inar, and AERE Summer Conference for their insightful feedback and helpful discussions.
iii
Table of Contents
Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
Chapter 1: Can Farmers Adapt to Climate Change in Developing Countries? Causal
Evidence from Coastal Bangladesh . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Background and Motivations . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Sea Level Rise and Salinity Intrusion in Bangladesh . . . . . . . . . . 5
1.2.2 Adverse Eect of Salinity on Crop Productivity . . . . . . . . . . . . 6
1.2.3 The Puzzle: Why Has There Been No Apocalypse? . . . . . . . . . . 7
1.2.4 Adaptation to Salinity in Farming Sector . . . . . . . . . . . . . . . . 8
1.2.5 Tidally Active Delta: Sharp Increase of Salinity at the Boundary . . 9
1.3 Data and Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.1 Salinity Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.2 Agricultural Census Data . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.3 Land Use Data from Landsat-5 Satellite Images . . . . . . . . . . . . 12
1.4 Empirical Strategy & Identication . . . . . . . . . . . . . . . . . . . . . . . 13
1.4.1 Fuzzy Spatial Regression Discontinuity Approach . . . . . . . . . . . 13
1.4.2 Instrumental Variable Approach . . . . . . . . . . . . . . . . . . . . . 15
1.4.3 Two Way Fixed Eect Approach . . . . . . . . . . . . . . . . . . . . 16
1.4.4 Adaptation to Climate Change: Short Run vs. Long Run . . . . . . . 17
1.5 Results and Interpretations . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.5.1 First Stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.5.2 Eect of Salinity on Land Use . . . . . . . . . . . . . . . . . . . . . . 18
1.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
iv
Chapter 2: The Role of Land Market in Achieving the Scale for Climate Change
Adaptation in Developing Countries . . . . . . . . . . . . . . . . . . . . . 32
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.2.1 Dierence in Economies of Scale and Need for Consolidation . . . . . 37
2.2.2 Role of Land Market in Achieving the Scale for Adaptation . . . . . . 38
2.3 Data and Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.4 Empirical Strategy and Identication . . . . . . . . . . . . . . . . . . . . . . 40
2.4.1 Instrumental Variable Approach . . . . . . . . . . . . . . . . . . . . . 40
2.5 Results and Interpretations . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.5.1 Eect of Salinity on Farm Size . . . . . . . . . . . . . . . . . . . . . . 41
2.5.2 Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Chapter 3: When Does Public Investment Complement or Crowd Out Private
Adaptation? Evidence from Hydraulic Structures in Coastal Bangladesh . 52
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.2.1 Coastal Embankments and Sluice Gates . . . . . . . . . . . . . . . . 56
3.2.2 Sluice Gate Types and Discretion Over Operation . . . . . . . . . . . 57
3.2.3 Impact of Automated Gate Operation on Land Use . . . . . . . . . . 57
3.3 Data and Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.3.1 Data on Hydraulic Structures . . . . . . . . . . . . . . . . . . . . . . 59
3.4 Empirical Strategy and Identication . . . . . . . . . . . . . . . . . . . . . . 59
3.4.1 Dierence in Discontinuity Approach . . . . . . . . . . . . . . . . . . 59
3.5 Results and Interpretations . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
v
List of Tables
1.1 Range of Salinity Across Study Villages . . . . . . . . . . . . . . . . . . . . . 24
1.2 Summary Stats of Agricultural Census 2008 . . . . . . . . . . . . . . . . . . 24
1.3 Increased Probability of High Salinity within Tidal Delta . . . . . . . . . . . 26
1.4 Eect of Salinity on Land Allocation to Aquaculture . . . . . . . . . . . . . 29
1.5 Robustness to Bandwidth Selection . . . . . . . . . . . . . . . . . . . . . . . 29
1.6 Robustness to Placebo Outcomes . . . . . . . . . . . . . . . . . . . . . . . . 30
1.7 Eect of Salinity on Land Allocation to Rice Farming by Season . . . . . . . 30
1.8 Eect of Salinity on Land Allocation to Dry Season Rice Farming by Variety 30
1.9 Longitudinal Evidence on Eect of Salinity on Land Allocation to Aquaculture 31
2.1 Eect of Salinity on Farm Size . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.2 Eect of Salinity on Ownership Consolidation . . . . . . . . . . . . . . . . . 50
2.3 Evidence on Consolidation through Rental Market . . . . . . . . . . . . . . . 50
2.4 Evidence on No Dierential Rental Behavior in Housing Market . . . . . . . 50
2.5 Evidence on No Change in Long Run Land Supply . . . . . . . . . . . . . . 51
3.1 Construction Time of Drainage Sluice Gates . . . . . . . . . . . . . . . . . . 65
3.2 Distance of sluice Gates from Study Villages . . . . . . . . . . . . . . . . . . 67
3.3 Type of Sluice Gates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
vi
3.4 Eect of Gate Types on Land Allocated to Aquaculture . . . . . . . . . . . . 68
3.5 Eect of Gate Types on Dry Season Rice Farming . . . . . . . . . . . . . . . 68
3.6 Eect of Gate Types on Monsoon Season Rice Farming . . . . . . . . . . . . 69
3.7 Evidence on No Strategic Gate Placement Post Liberalization . . . . . . . . 69
3.8 Evidence on Eect of Gate Types with Pre Liberalization Sample . . . . . . 69
vii
List of Figures
1.1 Trend of Salinity in the Study Area . . . . . . . . . . . . . . . . . . . . . . . 22
1.2 Seasonality in Salinity Level . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.3 Aquaculture Farms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.4 Copernicus Ocean Monitoring Satellite Data on Salinity . . . . . . . . . . . . 23
1.5 Landsat 5 Images of Dry Season Aquaculture: 1988 vs 2009 . . . . . . . . . 24
1.6 Tidally active delta . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.7 Coastal Contour Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.8 Soil Salinity Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.9 Regression Discontinuity Plot: First Stage . . . . . . . . . . . . . . . . . . . 28
1.10 Placebo Cuto: First Stage . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.1 Farm Size Across Countries . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.2 Misallocation in Agriculture . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.3 Productivity vs. Scale in Agriculture . . . . . . . . . . . . . . . . . . . . . . 46
2.4 Productivity vs. Scale in Aquaculture . . . . . . . . . . . . . . . . . . . . . . 47
2.5 Scale Comparison: Agriculture vs. Aquaculture . . . . . . . . . . . . . . . . 47
2.6 Plot Size Comparison: Agriculture vs. Aquaculture . . . . . . . . . . . . . . 48
2.7 Farm Size Distribution: High vs. Low Salinity . . . . . . . . . . . . . . . . . 48
viii
2.8 Ownership Distribution: High vs. Low Salinity . . . . . . . . . . . . . . . . . 49
2.9 Consolidation Visual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1 Coastal Embankment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.2 Sluice Gate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.3 Hoist Operated Vertical Sluice Gate . . . . . . . . . . . . . . . . . . . . . . . 66
3.4 Automated Radial Sluice Gate . . . . . . . . . . . . . . . . . . . . . . . . . . 67
ix
Abstract
This thesis provides novel evidence on three key issues related to climate change adaptation in
the coastal regions of developing countries: (1) can farmers adapt to soil salinity intrusion?
(2) what mechanism enables the farmers to consolidate lands when adaptation involves
switching to a farming with higher scale? (3) when does investment in coastal protection
crowd out private adaptation? I study these questions in the context of salinity intrusion
in coastal Bangladesh. While a technique to adapt to salinity intrusion is reallocating lands
from crop farming to salinity resistant aquaculture, the latter has relatively higher economies
of scale compared to that of the former. However, farmlands in developing countries are
highly fragmented and land market frictions have caused historical stagnancy in farm size
consolidation. Hence, if there is no consolidation, farmers operating fragmented land parcels
would fail to adapt, resulting in mass loss of livelihoods. Exploiting an exogenous variation
in soil salinity level arising from a hydrological feature of coastal Bangladesh, I employ both
fuzzy regression discontinuity design and instrumental variable approaches and nd that
there is a signicant increase the amount of land allocated to aquaculture in the areas aected
by high level of salinity intrusion. Moreover, this reallocation involved a signicant increase
in farm size. Furthermore, I nd strong evidence that rental market for land performs
substantial consolidation role in achieving the scale required for doing aquaculture. Finally,
x
exploiting the plausibly exogenous variation in sluice gate types, I nd evidence that the
extent of adaptation doubles when local people have discretion over the operation of the
hydraulic structures deployed for coastal protections. Together, all the evidence in this thesis
suggest that farmers can indeed adapt to salinity intrusion by reallocating their lands and
a functioning land market and
exibly designed protective infrastructures can complement
the adaptation process.
xi
Chapter 1
Can Farmers Adapt to Climate Change in Developing
Countries? Causal Evidence from Coastal Bangladesh
Mohammad Islamul Haque, Bosen Shao
1
1.1 Introduction
One of the major threats posed by global warming is the rise of sea level and the resulting
intrusion of salinity in the coastal areas across the world, and particularly in the low lying
deltaic regions such as Ganges-Brahmaputra-Meghna (GBM) delta in Bangladesh and West
Bengal of India, Mekong delta in Vietnam, etc. (Rahman et al., 2019). Multiple risk
assessment reports by US intelligence agencies have predicted/warned that even a moderate
rise in sea level would result in mass loss of livelihoods, forced migration, and con
icts in the
low lying deltaic regions of South Asia (NIC, 2017; NSMIP, 2020). However, since 1973 the
high water levels have increased at a rate of 15 millimeter (mm) per year in the tidal rivers
of southwestern coast of Bangladesh (Pethick and Orford, 2013). Furthermore, this rise in
1
Bosen Shao is a graduate student of Data Analytics at McDaniel College. He contributed in this chapter
by constructing land use data from satellite images.
1
tidal water level has resulted in 10 to 15 fold increase in the amount of farmlands with a
dry season salinity level above 8 dS/m, a level known to be extremely harmful for crops, in
some of those coastal districts between 1973 and 2009 (SRDI, 2010)
2
. Despite such increase
in tidal high water level and salinity, we haven't seen the apocalyptic scale of livelihood loss
and con
icts predicted by the security risk assessment reports. This paper tries to explain
that puzzle by investigating whether farmers in the developing countries can adapt to soil
salinity intrusion.
In fact, a widely practiced technique of adapting to high salinity level during the dry
season (January to May) is to create earthen dykes around the crop eld and trap tidal
water in order to cultivate shrimp, crabs, and species of shes that thrive in high salinity
environment
3
. This technique of adaptation primarily takes the form of inter-cropping where
the farmers harvest the aquaculture outputs and drain out the saline water before the onset
of the monsoon and do rain-fed rice farming during the wet season (July to October) when
persistent exposure to rainfall leaches away the soil salinity.
However, estimating causal eects of salinity intrusion is extremely challenging for both
reverse causality issue and omitted variable biases. Firstly, rapid adoption of aquaculture
started after the liberalization of exchange rate and trade in the early 1980s opened up the
opportunity to export shrimps and crabs to Europe, North America, and Japan at lucrative
prices. Hence, it is not obvious whether widespread adoption of dry season aquaculture in
coastal Bangladesh and elsewhere is a strategy to adapt to high level of soil salinity or just
2
dS/m is the short form for decisiemens per meter, a unit of electrical conductivity that is widely used
to measure salinity level in water and soil.
3
This method of adaptation to salinity intrusion in coastal farms is practiced in almost all the countries
in South Asia, South-East Asia, China and many other countries of the world
2
an adoption of a protable farming practice in response to high export demand. In the latter
case, it might be possible that the farmers allow in
ow of saline water in their elds, that
are otherwise suitable for crop farming, in order to increase their farm income. This means
that it is aquaculture that drives the increase in salinity level and not the other way around.
Secondly, there might be unobserved meteorological, topological, and soil features that aect
both salinity level and aquaculture productivity. Given the above two challenges, it is not
surprising that there has not been any paper in the literature that provides reasonable causal
estimates of land use adaptation to soil salinity. This chapter of my thesis attempts to ll
this gap in the literature on climate change adaptation.
In particular, I provide causal evidence on this question by exploiting the plausibly exoge-
nous frontier between tidally active delta and mature delta in the coastal Bangladesh. The
frontier between the tidally active delta and the mature delta is roughly determined by the
the extent of reach of the semi-diurnal tidal waves from the Bay of Bengal. As a result, there
is increased salinity in this region due to the capillary movement and occasional over
ows of
highly salinated tidal waters. Firstly, using a regression discontinuity (RD) approach I nd
that the probability of having a salinity level above 8 dS/m in the dry season increases by
50 percentage point at the frontier between mature delta and tidally active delta. Further-
more, the fuzzy RD estimate suggest that an increase in salinity level above 8 dS/m cause
a 64 percentage point increase in the percentage of land allocated to aquaculture. Secondly,
we also use a standard two stage least square (2SLS) approach using the frontier of tidally
active delta as an instrument for high salinity level. The estimates from the IV approach are
similar to that of the fuzzy RD approach highlighting the fact that the ndings from fuzzy
RD approach are not due to any hyperlocal aberration.
3
An important threat to the exclusion restriction both in the fuzzy RD and IV approach
is that the dierence in land allocation to aquaculture might arise due to dierences in soil
characteristics other than salinity. I rule out this concern by showing that there is no change
in the amount of land allocated to rice cultivation during wet season when the salinity level
is low and that the increase in allocation of land to aquaculture in the dry season comes
from a corresponding decline in the amount of land allocated to dry season rice farming. If
there were other dierences in permanent soil characteristics across the frontier that aect
rice productivity, the amount of land allocated to wet season rice farming would also decline
across the frontier. Furthermore, I rule out season specic meteorological confounders by
showing that the decline in the dry season rice cultivation is much lower for relatively salinity
tolerant local varieties of rice compared to that of the sensitive high yielding varieties.
Adding up all these evidence provide strong indication that farmers in the coastal areas
are indeed reallocating their lands from crop farming to aquaculture during the dry season
in response to salinity intrusion.
This chapter contributes to the litearture on adaptation to climate change in the farm-
ing sector. There is mixed evidence of adaptation in agriculture with respect to extreme
temperature and precipitation (Burke and Emerick, 2016; Desch^ enes and Greenstone, 2007;
Schlenker and Roberts, 2009; Taraz, 2018), and to lack of ground and surface water avail-
ability (Blakeslee et al., 2020; Hagerty, 2021; Hornbeck and Keskin, 2014). However, there
are only a few papers that investigate farmers' adaptation to salinity intrusion in the coastal
regions. The closest paper to this to this topic is Chen and Mueller (2018). This study nds
evidence that coastal salinity intrusion results in lower crop revenue that gets compensated
4
by a corresponding increase in aquaculture revenue. However, due to data limitation and
identication challenges they use within location variation in salinity exposure over just a
10 year time period. Hence, their results primarily correspond to behavioral response to
only medium term changes in salinity. But, in the absence of a 30 to 40 year long panel
dataset, we need to take advantage of cross sectional variation in salinity in order to capture
the full range of adaptive responses to long term changes in salinity level. However, even
though cross sectional estimates are inclusive of long term adaptation, these are susceptible
to omitted variable bias concerns. This is a key challenge in estimating adaptation to cli-
mate change and has been widely documented in the literature (Auhammer, 2018; Hsiang,
2016). I overcome this challenge by employing a spatial RD approach exploiting a spatial
discontinuity in salinity level in the coastal Bangladesh (Hagerty, 2021). To the best of my
knowledge, I provide the rst causal evidence on land use adaptation in response to salinity
intrusion in the coastal areas of developing countries.
1.2 Background and Motivations
1.2.1 Sea Level Rise and Salinity Intrusion in Bangladesh
Bangladesh is one of the most vulnerable countries to the threats from sea level rise. The
south-west coastal region is located within the Ganges-Brahmaputra-Meghna (GBM) delta
and has an elevation only between one to three meters above the sea level (L az ar et al., 2020).
Moreover, the high water levels have been increasing at an average rate of 15.9 millimeter
(mm) per year and a maximum rate of 17.2 mm per year in the same region over the last three
decades (Pethick and Orford, 2013). Low elevation combined with gradual increase in water
5
levels have resulted in massive increase in soil salinity level in the coastal districts. Figure 1.1
shows a comparison between the amount of farmlands with more than 8 dS/m salinity level,
a level known to be extremely harmful for traditional crops, in 1973 with that of in 2009 for
the three southwestern coastal districts of Bangladesh. We can see that from 1973 to 2009
the amount of farmlands with more than 8 dS/m salinity has increased by a factor of two
to twelve in these districts. Furthermore, gure 1.2 shows that there is a seasonal pattern
in the rise and fall of salinity within a year. In particular, the salinity level starts rising in
the dry season starting from December and peaks around April/May. Thereafter, it starts
declining with the beginning of monsoon in June and remains at relatively low level until
the end of November. Hence, the greatest impact of salinity increase falls on dry season crop
farming.
1.2.2 Adverse Eect of Salinity on Crop Productivity
Bangladesh is a rice eating country and hence traditionally an overwhelming majority of
the farmlands have been used for rice farming. However, rice farming is highly sensitive
to soil salinity. For example, traditional high yielding varieties (HYV) of rice can tolerate
upto only 3 dS/m soil salinity and its productivity declines by almost 40 percent at 8 dS/m
(Tanji and Kielen, 2002). Even one of the most resilient rice varieties, BRRI Dhan-47,
developed by Bangladesh Rice Research Institute (BRRI) can tolerate upto only 8 dS/m
of soil salinity without any decline in productivity. Furthermore, its productivity declines
by almost 43% at a salinity level of 12 dS/m (Radanielson et al., 2018). In addition, other
resilient alternative dry season crops such as wheat is only resilient upto 6 dS/m and its
productivity falls by 14.2% at a salinity level of 8 dS/m (Tanji and Kielen, 2002). This
6
shows that salinity intrusion beyond 8 dS/m makes the farmlands almost entirely unsuitable
for traditional crop farming.
1.2.3 The Puzzle: Why Has There Been No Apocalypse?
It is evident from the preceding subsection that rapid intrusion of extreme level of salin-
ity (e.g., beyond 8 dS/m) can potentially destroy the livelihoods of millions of agricultural
farmers in the coastal region of Bangladesh. Hence, it is understandable that US intelligence
agencies have outlined in multiple security risk assessment reports that a modest sea level
rise could result in loss of livelihoods, mass migration, and con
ict in dierent deltaic regions
of the world including the GBM delta in South Asia (NIC, 2017; NSMIP, 2020). However,
despite a 15-fold increase in the amount of farmlands with extreme level (above 8 dS/m) of
salinity in some of the districts between 1973 and 2009, we haven't seen the kind of chaos,
con
ict, and mass migration predicted in various reports prepared by the US intelligence
communities. This poses a puzzle that the rst part of this paper attempts to solve. In
particular, the limitation of the predictions from climate scientists and the intelligence ex-
perts is that they don't take into account the full range of behavioral responses to salinity
intrusion. Accounting for full range of adaptive behaviors is specially challenging because
people not only choose from the existing set of choices for adaptation, but also use their
ingenuity to innovate and expand the choice set over time. In the following subsection we
discuss such an innovative farming technique for adapting to salinity intrusion in the coastal
areas.
7
1.2.4 Adaptation to Salinity in Farming Sector
Many farmers in the worst aected southwestern coast are reallocating their lands to salinity
resistant aquaculture in the dry season. This involves creating earthen embankments sur-
rounding their plots to trap saline water and farm shrimp, crabs, swamp eel, etc. that can
grow in high salinity aquatic environment
4
. Furthermore, since shrimps/crabs can't survive
in low salinity environment, farmers drain out the water and do rain-fed rice farming during
the monsoon season when salinity level declines naturally due to heavy rainfall
5
. However,
rapid adoption of this innovative adaptation technique in farming started after the liberal-
ization of exchange rate and trade in the early 1980s opened up the opportunity to export
shrimps and crabs to Europe, North America, and Japan at lucrative prices. Hence, it is not
obvious whether widespread adoption of dry season aquaculture in coastal Bangladesh and
elsewhere is a means to adapt to high level of soil salinity or just an adoption of a protable
farming practice in response to high export demand. In that case it might be possible that
the farmers allow in
ow of saline water in their elds, that are otherwise suitable for crop
farming, in order to do more lucrative forms of farming. This would also mean that it is
aquaculture that drives the increase in salinity level and not the other way around. Hence,
it has been very challenging to generate rigorous causal evidence on the eect of salinity
intrusion on farming choices.
4
Figure 1.3 shows how the transformed crop elds for doing aquaculture look like
5
Shrimps and crabs both require at least 15 ppt salinity to grow.
8
1.2.5 Tidally Active Delta: Sharp Increase of Salinity at the Bound-
ary
Majority areas of coastal Bangladesh fall within a deltaic plane formed by the sedimentation
of alluvium carried by the rivers originated in Himalayas range and eastern hilly regions
of India. As gure 1.6 shows, the entire deltaic plane can be divided into four regions:
active delta, tidally active delta, mature delta and moribund delta (Islam and Gnauck, 2008;
Passalacqua et al., 2013). For the purpose of identifying the causal impact of salinity I
will take advantage of the exogenous dierence in soil salinity between tidally active delta
and mature delta. To be specic, there are two basic dierences between mature delta and
tidally active delta. Firstly, tidally active delta contains the major portion of the network
of tidal rivers and channels which help
ow the semi-diurnal tides from the Bay of Bengal
to the inland. Hence, the eect of tides are relatively more prevalent in the tidally active
delta. Secondly, there is a discontinuous decrease in elevation roughly at the frontier of the
tidally active delta. In gure 1.7, we can see that there is a 1 meter contour line that passes
almost along the border between tidally active delta and mature delta. Hence, elevation
decreases abruptly after that line. Because, of lower elevation, areas within tidally active
delta are more likely to get salinated by the capillary movement and occasional over
ow of
tidal waters. I argue, and later prove, that the above two features of tidally active delta
cause a discontinuous increase in soil salinity as one moves from mature delta to tidally
active delta across the border of these two regions. This is also supported by the soil salinity
map in gure 1.8. We can see that the region with the highest level of salinity almost
exactly coincides with the tidally active delta and there is an abrupt change in salinity
9
almost along the frontier of the tidally active delta. I exploit this fuzzy border between the
mature delta and tidally active delta to design a fuzzy spatial RD model in order to identify
the causal eect of increased soil salinity on land use choices. I also use the same border
as an instrument for high level of salinity (beyond 8 dS/m) in an IV specication used to
supplement the fuzzy RD results.
1.3 Data and Descriptive Statistics
1.3.1 Salinity Data
I use two sources for salinity data in this paper. The rst data set is from a survey con-
ducted by the soil resources development institute (SRDI) in 2009 in the coastal regions of
Bangladesh. A eld survey was conducted in the south-western coastal farmlands excluding
Sunderbans by multiple soil survey teams from SRDI in May, 2009. The sample points were
chosen through traverse line methods. Selected traverse lines were changed or modied ac-
cording to necessity during the eld survey. The adjacent traverse lines were at an interval of
3-4 kilometers from each other. Along all the traverse lines 2500 soil samples were collected
and analyzed in the laboratory to measure the salinity level. For more details on the survey
and the salinity testing methods please refer to SRDI (2010). On the basis of this survey
information, coastal land areas were classied into ve dierent categories: (i) 0 to 4 desi-
Siemens per meter (dS/m), (ii) 4.1 to 8 dS/m, (iii) 8.1 to 12 dS/m, (iv) 12.1 to 16 dS/m,
and (v) above 16 dS/m. Table 1.1 shows that out of more than 50 percent of the villages in
my sample have a dry season soil salinity level above 8 dS/m, a level beyond which even the
most resilient variety of crops/rice suer substantial productivity damage.
10
The second source of salinity data comes from Copernicus Marine Environment Monitoring
Service (CMEMS). This is a gridded dataset with a spatial resolution of approximately
8km*8km and contains monthly data on ocean salinity, surface temperature, etc. from 1993
to 2020. Moreover, water salinity is expressed in practical salinity unit (psu) in this dataset.
Figure 1.4 shows the trend of salinity in the month of April, time of the year when salinity
level peaks, from 1993 to 2020. We can see that there has been a steady increase in ocean
salinity during this time period. However, the magnitude of increase seems small. This is
not surprising because ocean salinity doesn't change by large magnitude within tractable
time period. But, a small change in ocean salinity might correspond to larger change in the
surface and groundwater salinity, and soil salinity.
1.3.2 Agricultural Census Data
The primary analysis in this paper is based on agricultural census data of 2008. The agricul-
tural census of Bangladesh 2008 was a full-count based household level census which collected
data on household composition, land ownership, dierent types of land uses, agricultural as-
sets, livestock etc. Data of all households of the country in both rural and urban areas
were collected simultaneously. Entire country was divided into 1,53,945 Enumeration Areas
(EAs). Each enumeration area had around 200 households in rural areas and around 300
households in urban areas and also a location map. The map was used to identify EAs cor-
rectly on ground. One enumerator was assigned the responsibility of one or two enumeration
areas for data collection. Moreover, there were 19,529 supervisors, meaning one supervisor
for 7 to 8 EAs. This was the most strictly monitored census in the history of Bangladesh.
Moreover, after the completion of eld work of full-count census, a post enumeration check
11
was done in order to assess the census data quality. For further details on the agricultural
census of 2008 please refer to BBS (2010).
Table 1.2 provides summary statistics of a set of variables for the region of our interest.
The table shows that most of the land, asset, and livestock variables have distributions
skewed to the right.
1.3.3 Land Use Data from Landsat-5 Satellite Images
We use Landsat 5 satellite images to detect dry season aquaculture in the coastal region.
Landsat 5 images have a resolution of 30 meter by 30 meter (approximately 20 decimals of
land). The publicly available images are at a frequency of 16 days, from 1988 to 2013. Firstly,
we identify and exclude all the permanent water bodies from the study area. Thereafter, we
calculate Normalized Dierence Water Index (NDWI) from Landsat 5 bands following Rokni
et al. (2014) for the dry seasons, February to April. The NDWI index for dry season help us
detect water in land surface during dry season. To be specic, we classify every 30 meter by
30 meter grid with a NDWI over 0.4 in the dry season as an aquaculture farm . This helps
us to measure the percentage of farmlands that are allocated to aquaculture for each of the
administrative unions in the coastal belt of Bangladesh. Figure 1.5 shows the comparison of
the amount of land employed in aquaculture between 1988 and 2009. We can see that there
has been a rapid increase in the amount of land used for dry season aquaculture between
these years.
12
1.4 Empirical Strategy & Identication
As discussed before, one of the main challenges in identifying adaptation to soil salinity
intrusion is the potential reverse causality issue. For example, since aquaculture is highly
protable, farmers might allow the saline water to
ow into their lands and this can in
turn increase the salinity level. Furthermore, other observed and unobserved meteorologi-
cal, topological, and soil characteristics might aect both salinity level and productivity of
aquaculture and crops. Finally, note that strategic sorting of farmers depending on the level
of salinity is not a concern given that the focus of this paper is solely on ecient land use.
In the presence of strategic sorting I will be capturing both the intensive and the extensive
margins of the eects of salinity on land use, which is in fact the policy relevant parameter
in this context.
1.4.1 Fuzzy Spatial Regression Discontinuity Approach
I exploit the plausibly exogenous border between the mature delta and the tidally active
delta to employ a fuzzy spatial regression discontinuity (RD) design. As described in the
background section, while very few areas within mature delta have more than 8 dS/m, all
the areas within the tidally active delta have a salinity level above that threshold. Therefore,
there is a sharp increase in the probability of having a soil salinity level above 8 dS/m as we
move from the mature delta to the tidally active delta right at the border between them. I
exploit this geographical feature to employ a fuzzy RD design in order to identify the eect
of salinity on land use choices. I use a standard local linear specication as follows:
13
Y
hv
=
N
v
+d
v
+d
v
N
v
+
l
+ X
h
+
hv
(1.1)
In equation 1.1, Y
hv
is the outcome of interest for household h in village v. Our main
dependent variables are the amount of farmlands that are employed in dierent types of land
use. Moreover, in the above equation,d
v
represents the nearest distance from the centroid of
villagev to the frontier of tidally active delta in kilometers. I use positive distances for villages
within tidally active delta and negative distances for villages in mature delta. In addition,
X
h
is a vector of farm covariates and
l
represents longitude-quartile xed eects. I use
longitude-quartile xed eects, following Ito and Zhang (2020), in order to avoid comparing
villages that are located in distant longitudes. Finally, the treatment variableN
v
is a dummy
variable indicating whether the centroid of villagev falls within the areas with a salinity level
of more than 8 dS/m. Here, the parameter of interest is
. I use the 8 dS/m threshold for
dening treatment because there is no variety of rice that is fully resilient beyond this level of
salinity (Radanielson et al., 2018). As discussed earlier, the treatment variable N
v
could be
endogenous both due to reverse causality and omitted variable concerns. The identication
strategy is to use the dummyZ
v
indicating whether the centroid of villagev falls within the
tidally active delta as an instrument for the treatment variable to circumvent these problems.
The exclusion restriction is that within a small neighborhood around the frontier between
these two deltaic regions, being within tidally active delta only aects the land use choices
through increased salinity level. Furthermore, I use local linear regression with triangular
kernel weighting and Mean Square Error (MSE) optimal bandwidth to estimate the RD
treatment eects (Calonico et al., 2014). Finally, since treatment varies at village level, I
14
cluster the standard errors at village level, but the results are robust to clustering standard
errors at a higher administrative level such as unions.
1.4.2 Instrumental Variable Approach
I also use a two stage least square approach (2SLS) using the same exogenous border between
the mature delta and the tidally active delta as an instrument of whether a village falls within
an area with above 8dS/m salinity to estimate the treatment eects. I use the following
standard specication for the 2SLS approach:
Y
hvd
=
N
v
+
d
+ X
h
+
hvd
(1.2)
In equation 1.2, Y
hv
is the outcome of interest for household h in village v in district d.
The outcome variables are as described earlier. Furthermore,
d
represents xed eect for
districtd andX
h
represents a vector of household specic covariates. Finally, the treatment
variable N
v
is a dummy variable indicating whether the centroid of village v falls within
the areas with a salinity level of more than 8 dS/m. Here, the parameter of interest is
.
As mentioned before, I use the 8 dS/m threshold for dening treatment because there is
no variety of rice that is resilient beyond this level of salinity (Radanielson et al., 2018).
Furthermore, as discussed earlier, the treatment variable N
v
could be endogenous both due
to reverse causality and omitted variable concerns. The identication strategy is to use the
dummy Z
v
indicating whether the centroid of village v falls within the tidally active delta
as an instrument for the treatment variable to circumvent these problems. The exclusion
15
restriction is that being within tidally active delta only aects the land use choices through
increased salinity level.
The instrumental variable (IV) approach is used in addition to the fuzzy RD approach
to ensure that the results have more external validity. Moreover, as expected, fuzzy RD
approach lacks sucient power in most of the tests. Hence, IV approach works as a nice
supplement in verifying the statistical signicance of some results. However, the identication
assumption is stronger for the 2SLS approach
6
. Nonetheless, I rule out a host of threats to
the exclusion restrictions in my analysis.
1.4.3 Two Way Fixed Eect Approach
In this approach I only use the coastal unions and look at the eect of within union change
in ocean salinity (copernicus monitoring data) at the nearest coastal point on land allocation
of that area. I use a standard two way xed eect (TWFE) estimation approach as follows:
Y
ut
=
Salinity
ut
+
u
+
t
+
ut
(1.3)
Here, Y
u
is the outcome of interest for union u at time t,
u
is the union xed eect,
and
t
is the time xed eect. The treatment variable is Salinity
u
t which shows the ocean
salinity level at the nearest coastal point of union u at time t. Our parameter of interest is
again
. The identication assumption is that there is no time varying unobservable that
aect both the ocean salinity and land use in the coastal unions. However, with union and
6
Essentially, the exclusion restriction has to be satised for a larger bandwidth in case of 2SLS than that
of the fuzzy spatial RD approach
16
year xed eects, this approach only captures the behavioral response to short run year to
year
uctuations in the salinity level. I discuss this issue in the next subsection.
1.4.4 Adaptation to Climate Change: Short Run vs. Long Run
There is a large literature that use TWFE models exploiting within location variation in
weather variables over time to estimate adaptation to climate damage (Desch^ enes and Green-
stone, 2007; Schlenker and Roberts, 2009; Dell et al., 2012). However, the weakness of this
approach is that the economic agents would perceive the short term weather
uctuations as
transitory and therefore not adjust their choices in the way they would have done in response
to a long term change in climatic conditions. Hence, estimates from this approach are not
inclusive of full range of adaptive behavioral responses to climate change. On the contrary,
while cross sectional estimates are inclusive of adaptation to long term dierence in climatic
conditions, these are highly susceptible to omitted variable bias concerns (Mendelsohn, Nord-
haus, and Shaw, 1994; Schlenker, Michael Hanemann, and Fisher, 2005). However, we can
take advantage of the cross sectional variation in climatic conditions and get around the
challenges from omitted variable bias concerns with either a spatial discontinuity (Hagerty,
2021) or a ne scale quasi-random variation (Blakeslee et al., 2020) in the climatic vari-
able/marker of interest. The fuzzy RD and IV approaches described above helps us to do
that in this chapter. However, we still present the TWFE results to illustrate the dierences
between short term and long term responses to climate change.
17
1.5 Results and Interpretations
1.5.1 First Stage
Firstly, gure 1.9 shows that there is a sharp increase in probability of being in the high
salinity zone inside the tidally active delta. In fact, all the villages within the tidal delta
have a salinity level above 8 dS/m. Furthermore, gure 1.10 shows that the discontinuity is
only at the frontier and once we move 3-4 kilometers in either direction from the frontier, the
discontinuity goes away. Table 1.3 conrms the visual evidence on the rst stage relationship.
Column 1 and 2 provide the rst stage of the RD and the 2SLS approaches, respectively. We
can see that the probability of having more than 8 dS/m salinity goes up by between 18 to
50 percentage point inside the tidally active delta depending on the specication. Finally, as
expected, since RD approach focus on a small bandwidth around the frontier, the rst stage
F-statistic for the RD approach is relatively smaller indicating weak instrument issue. Hence,
we will report the Anderson Rubin (AR) test statistics for all the second stage estimates
that are obtained from the fuzzy discontinuity approach.
1.5.2 Eect of Salinity on Land Use
Table 1.4 shows the second stage results on the eect of having more than 8 dS/m salinity
on amount of land allocated to aquaculture. The rst two columns provide results from the
fuzzy RD approach and the second two columns provide estimates from the 2SLS approach.
Firstly, the table shows statistically signicant positive impact of being within high salinity
areas on the amount of land allocated to aquaculture across all the specications, with or
without farm controls. Considering the mean farm size in the control and the treatment
18
areas, we can see that the percentage of farmland allocated to aquaculture increases between
64 percentage points to 29 percentage points depending on the specication.
An important threat to the exclusion restriction both in the fuzzy RD and IV approach
is that the dierence in land allocation to aquaculture might arise due to dierences in soil
characteristics other than salinity. Results in table 1.7 rule out this concern by showing that
there is no change in the amount of land allocated to rice cultivation during wet season when
the salinity level is low and that the increase in allocation of land to aquaculture in the dry
season comes from a corresponding decline in the amount of land allocated to dry season
rice farming. If there were other dierences in permanent soil characteristics across the
frontier that aect rice productivity, the amount of land allocated to wet season rice farming
would also decline across the frontier. Furthermore, estimates from table 1.8 rule out other
season specic meteorological confounders by showing that the decline in the dry season rice
cultivation is much lower for relatively salinity tolerant local varieties of rice compared to
that of the sensitive high yielding varieties. Finally, table 1.5 and table 1.6 shows that the
reduced form RD estimates are not sensitive to bandwidth choices and are robust to placebo
outcomes.
Given that the above RD and the 2SLS estimates provide behavioral response to long run
dierences in salinity, it is worthwhile to look at the response to short run
uctuations in
salinity level. Table 1.9 provides estimates of the two way xed eect model in equation 1.3
using within union variation in exposure to coastal ocean salinity over time
7
. Firstly, we
7
We had to use this pure reduced form estimate because we do not have data to study the rst stage
relationship between ocean salinity and soil salinity for the coastal unions.
19
can see that 1 psu increase in salinity in the nearest coastal point results in between 2 to
2.46 decimal increase in the amount of land allocated to aquaculture in a union. However,
since that the annual peak ocean salinity increased by only 0.03 psu between 1993 and 2020
(gure 1.4), the TWFE estimates imply an economically insignicant response to short term

uctuations in ocean salinity. But, this is exactly what we would expect based on the theory
of adaptation. Because, farmers wouldn't make large xed investments in order to reallocate
their lands to aquaculture simply in response to short term
uctuations in salinity. They
would expect the salinity level to come back to its long run average level.
Together, the above estimates make a convincing case that farmers can indeed adapt to
changes in long run salinity level by reallocating their lands from crop farming to aquaculture.
This might be the potential explanation behind why there were no mass loss of livelihoods,
and con
icts despite having rapid increase in salinity in the coastal regions of Bangladesh.
1.6 Conclusion
There has been rapid increase in salinity intrusion in the coastal Bangladesh since 1973. The
salinity level has gone beyond the threshold above which no traditional crop is fully resilient.
Nonetheless, I nd strong evidence that farmers adapt to this challenge by reallocating
their lands from crop farming to aquaculture during the season of high salinity. Hence,
the predictions by climate experts and intelligence agencies that increased salinity intrusion
would result in mass loss of livelihoods seems to have not taken into account the full range of
adaptive responses of the coastal population. The ndings of this chapter make a strong case
20
for careful consideration of land use adaptation to salinity intrusion in models for assessing
future damages from sea level rise.
21
Appendix
Figure 1.1: Trend of Salinity in the Study Area
Figure 1.2: Seasonality in Salinity Level
22
Figure 1.3: Aquaculture Farms
Figure 1.4: Copernicus Ocean Monitoring Satellite Data on Salinity
23
Figure 1.5: Landsat 5 Images of Dry Season Aquaculture: 1988 vs 2009
Salinity range (dS/m) Frequency Percent
0 to 4 924 36.89
4.1 to 8 228 9.10
8.1 to 12 292 11.66
12.1 to 16 667 26.63
Above 16 394 15.73
Total 2505 100
Table 1.1: Range of Salinity Across Study Villages
Variables Mean SD Median
Household Size 5 2 4
Homestead Size 6 9 5
Owned Land 88 210 17
Rented Land 25 111 0
Leased out Land 21 110 0
Net Cultivated Land 68 164 8
Net Irrigated 27 83 0
Fallow Land 1 19 0
Temporary Crop Land 60 156 0
Land in Aquaculture 20 123 0
Observations 1263118 1263118
Land variables are in decimals (1 acre=100 decimals)
Table 1.2: Summary Stats of Agricultural Census 2008
24
Figure 1.6: Tidally active delta
The above map shows division of the Bengal Delta into active, tidally active, mature and
moribund deltas. Source: Islam and Gnauck (2008)
25
Figure 1.7: Coastal Contour Map
Source: Climate Change Cell (2016)
1fSalinity> 8dS=mg
RD First Stage IV First Stage
(1) (2)
1fTidalDeltag 0.184

0.495

(0.0809) (0.0165)
F-Stats 4.60 848
Longitude quartile FE Yes No
District FE No Yes
N 116 2505
Control Mean 0.44 0.44
MSE Optimal BW 4.814 NA
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 1.3: Increased Probability of High Salinity within Tidal Delta
26
Figure 1.8: Soil Salinity Map
The above map shows the soil salinity level of the coastal areas. The extreme high salinity
zone in south-west coast almost coincides with the tidally active delta. Source: Soil
Resources Development Institute (2010)
27
Figure 1.9: Regression Discontinuity Plot: First Stage
Figure 1.10: Placebo Cuto: First Stage
28
Amount of Land in Aquaculture
Fuzzy RD 2SLS
(1) (2) (3) (4)
1fSalinity> 8dS=mg 72.17

77.74

40.13

36.09

(42.96) (45.60) (9.011) (8.449)
AR Wald Test: Chi Sq 7.84 8.37 17.75 16.58
p-val [0.019] [0.015] [0.00] [0.00]
Longitude quartile FE Yes Yes
District FE Yes Yes
Farm Controls No Yes No Yes
Control Mean 11 11 11 11
Control Mean Farm Size 62.40 62.40 62.40 62.40
Treatment Mean Farm Size 93.19 93.19 93.19 93.19
N 92144 92144 1166498 1166498
MSE Optimal BW 4.814 4.814 NA NA
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 1.4: Eect of Salinity on Land Allocation to Aquaculture
Land in Aquaculture
(1) (2) (3) (4) (5) (6) (7)
RD Estimate 11.07 17.59

22.19

22.21

19.63

16.54

13.83

(8.668) (8.508) (8.538) (8.342) (7.891) (7.509) (7.222)
Longitude-quartile FE Yes Yes Yes Yes Yes Yes Yes
Farm Controls Yes Yes Yes Yes Yes Yes Yes
Manual BW (km) 3.25 3.75 4.25 4.75 5.25 5.75 6.25
Eective N 90198 102314 110793 124227 136603 147833 149651
Control Mean 28.510 28.510 28.510 28.510 28.510 28.510 28.510
(3.899) (3.899) (3.899) (3.899) (3.899) (3.899) (3.899)
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 1.5: Robustness to Bandwidth Selection
29
Temporary Crops Bamboo Clumps Permanent Trees
(1) (2) (3)
RD Estimate -17.20

0.0119 -0.446
(8.241) (0.0997) (0.769)
Longitude-quartile FE Yes Yes Yes
Farm controls Yes Yes Yes
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 1.6: Robustness to Placebo Outcomes
Rabi Season Kharif Season
Rice Wheat Rice
(1) (2) (3)
1fSalinity> 8dS=mg -38.25

-0.951

14.18
(3.324) (0.152) (11.24)
District FE Yes Yes Yes
Control Mean 40.63 1.10 40
N 1166498 1166498 1166498
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 1.7: Eect of Salinity on Land Allocation to Rice Farming by Season
Rabi Season Rice Cultivation
Tolerant Rice Sensitive Rice
(1) (2)
1fSalinity> 8dS=mg -7.119

-31.14

(1.235) (4.126)
District FE Yes Yes
N 1166498 1166498
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 1.8: Eect of Salinity on Land Allocation to Dry Season Rice Farming by Variety
30
Percentage of Land in Aquaculture
(1) (2)
Salinity (psu) 2.023

2.455

(1.015) (1.108)
Union FE Yes Yes
Year FE Yes Yes
Weather Control No Yes
N 5260 5260
Mean in 1993 18.49 18.49
Standard errors in parentheses are clustered at union level

p< 0:10,

p< 0:05,

p< 0:01
Table 1.9: Longitudinal Evidence on Eect of Salinity on Land Allocation to Aquaculture
31
Chapter 2
The Role of Land Market in Achieving the Scale for
Climate Change Adaptation in Developing Countries
Mohammad Islamul Haque
2.1 Introduction
Reallocation of farmlands according to evolving comparative advantage can potentially miti-
gate the impact of climate change on farm income to a substantial extent (Costinot, Donald-
son, and Smith, 2016). However, if the reallocation involves switching to a farming technique
with relatively higher economies of scale compared to that of the existing one, the transition
process would require consolidation of multiple contiguous land parcels in any developing
country context with high fragmentation of land ownership
1
. But, such consolidation might
not be feasible due to cultural and institutional barriers in the land markets of developing
countries (Sood, 2020). Furthermore, even in the absence of any barrier, the consolida-
tion would require multiple land market transactions that can potentially lead to hold-up
1
For instance, while only less than 10 percent of farms in the US are below 10 acres, more than 80 percent
of the farms in India, China, Bangladesh, and Indonesia are less than 10 acres.
32
problems that would fully extract the rents which could have otherwise incentivized farmers
to put together a large farm. Such barriers to consolidation has been one of the key fac-
tors behind historically low rate of technology adoption and high productivity dispersion in
the farming sector of developing countries (Foster and Rosenzweig, 2022) and climate change
can potentially worsen the extent of dispersion if these barriers prevents reallocation of lands
according to evolving comparative advantage (Mendelsohn, 2008).
As described in chapter 1, sea level rise has caused a massive increase in the amount of
farmlands with more than 8 dS/m salinity, a level beyond which no crop is fully resilient, in
the coastal areas of Bangladesh. Furthermore, we have also seen evidence in chapter 1 that
farmers adapt to salinity by reallocating their lands to aquaculture in the dry season. In this
chapter I provide causal evidence on the mechanism through which the reallocation process
works. In particular, adaptation through reallocation of lands to aquaculture should require
substantial land consolidation due to the relatively higher economies of scale in aquaculture
in compared to that of agriculture, and the high level of land parcel fragmentation in these
regions. Figure 2.5 displays the relationship between prot per decimal of land and plot
size for rice farming and coastal aquaculture
2
. The graph shows that while rice farming
can be done protably even at a very small scale, it requires at least a 35 decimal plot to
obtain positive economic prot from aquaculture. As a result, reallocating crop elds to
aquaculture would require consolidation of multiple contiguous plots which in turn would
involve multiple land market transactions. However, there are cultural and institutional
barriers in land market transactions in developing countries that might limit the extent of
2
This graph is produced using nationally representative data from Bangladesh Integrated Household
Survey of 2011-2012. For more details on this survey data please refer to Ahmed (2015).
33
consolidation (Sood, 2020). Furthermore, even in the absence of any barriers, multiple land
market transactions could easily result in hold-up problems and slow down the process of
reallocation in practice. On the contrary, salinity intrusion increases the opportunity cost of
not reallocating lands to aquaculture. Hence, the extent to which land market can help in
achieving the scale required for transitioning from traditional crop farming to aquaculture
is fundamentally an empirical question.
Similar to chapter 1, in this chapter I exploit an exogenous variation in soil salinity level
arising from a hydrological feature and employ a two stage least square (2SLS) approach to
investigate whether the land market can perform the consolidation role in achieving the scale
required for land use adaptation to salinity intrusion in the coastal regions of Bangladesh.
Firstly, I investigate whether the reallocation process involved consolidation of farmlands
by estimating the eect of salinity intrusion on farm size. I nd that an increase in salinity
level above 8 dS/m result in a 52 percent increase in farm size. Moreover, there is an
associated increase in the gini coecient of farm size in these areas by around 23 percent.
This provides strong evidence that the reallocation of land from crop farming to aquaculture
was made possible by substantial consolidation of farmlands. Nonetheless, one important
threat to the exclusion restriction is that the dierence in both the farm size and land
allocation could arise from pre-existing dierences in farm size distribution between tidally
active delta and mature delta. I rule this out by showing that the increase in farm size
happens even after controlling for land ownership, suggesting that consolidation primarily
occurs through rentals.
34
Furthermore, I also investigate the mechanisms through which the farm size consolidation
takes place. The scale for adaptation might be achieved either through ownership consol-
idation via the land sales market and/or through consolidation of operation via the rental
market for land. However, historically the market for land sales has been very inactive in the
developing countries due to the cultural barrier in selling lands inherited from generation to
generation
3
. In consistence with this historical pattern I nd that salinity intrusion results
in only a moderate increase in the total land ownership, and the gini coecient of total land
ownership. This also suggests that majority of the land consolidation potentially occurs
through the rental market.
The rental market for land can perform both allocative role and consolidation role. Con-
solidation through rental market would imply that a smaller number of farmers would be able
to rent-in lands in high salinity aected areas, and conditional on being a renter, each farmer
would rent in a higher amount of lands. The empirical estimates support this mechanism
by showing that increase in salinity level above 8 dS/m results in a 20 percent decline in the
number of households who rent-in lands. Furthermore, conditional on being a renter, the
amount of land rented increases by 90 percent. An alternative interpretation of this result
based on purely allocative role of the land market would be that a limited number of people
have the skills of doing aquaculture and therefore, more people rent out lands for which only
a few renters are available. I rule out this interpretation by showing that increase in salinity
level above 8 dS/m doesn't cause any signicant change in the number of households who
rent out lands, or in the amount of land that is being rented out, implying that the long
3
Ownership of more than 95 percent of the farmlands in India are derived from inheritance (Foster and
Rosenzweig, 2022)
35
run land supply doesn't change in response to salinity intrusion. An important threat to
the exclusion restriction is that the rental market is more active in tidally active delta in
general. I rule this out by showing that there is no dierence in the rental market activity in
the housing sector between the two areas. Altogether, these results provide strong evidence
that the consolidation role of the rental market was crucial in achieving the scale required
for adaptation to salinity intrusion in coastal Bangladesh.
This chapter contributes to a classic literature in development economics on the historical
stagnancy in land consolidation and the resulting productivity dispersion in the farming
sector of developing countries (Foster and Rosenzweig, 2022; Adamopoulos and Restuccia,
2014). It is worthwhile to read the last few sentences from the conclusion section of Foster
and Rosenzweig (2022) to understand the severity of the problem as perceived by the leading
experts on this topic:
"despite the potential gains to farm productivity and output per worker of
transitioning to an equilibrium of fewer but larger farms, small farms are likely to
be the dominant force of production in low-income countries for the foreseeable
future without external intervention..."
With the increased risks from climate change this historical stagnancy in land consolida-
tion in the developing countries raises an important question that no one has studied in the
literature before: if climate change adaptation in the farming sector of developing countries
involve switching to a higher scale, would farmers be able to consolidate lands and achieve
the scale? To the best of my knowledge this is the rst and only paper that rigorously
investigates this issue?
36
This paper also contributes to the growing literature that shows that farms/rsm can
collectively achieve the scale required for mechanization through the rental market (Bassi
et al., 2022; Caunedo and Kala, 2021). This chapter build on their works to provide the rst
rigorous evidence that farmers can achieve the scale required for climate change adaptation
through the rental market for land.
2.2 Background
2.2.1 Dierence in Economies of Scale and Need for Consolidation
As described in detail in chapter 1, there has been rapid increase in salinity in the coastal
areas of Bangladesh and many farmers in the worst aected southwestern coast are reallo-
cating their lands to salinity resistant aquaculture in the dry season. This involves creating
earthen embankments surrounding their plots to trap saline water and farm shrimp, crabs,
swamp eel, etc. that can grow in high salinity aquatic environment
4
.
However, given the highly fragmented land parcels in developing countries, reallocation
of lands from traditional crop to aquaculture is not a straightforward process. In particular,
aquaculture has relatively much higher economies of scale compared to that of rice farming.
Figure 2.5 compares the relationship between prot per decimal and the scale of operation
measured in decimals in rice farming with that of
5
aquaculture. Two stylized facts emerge
from the gure: (i) while aquaculture requires a minimum scale (approximately 30 decimals)
for protable operation rice farming has no such minimum scale constraint, (ii) beyond the
4
Figure 1.3 shows how the transformed crop elds for doing aquaculture look like
5
1 Acre=100 Decimals
37
minimum scale, aquaculture has relatively much higher economies of scale compared to that
in rice farming. Furthermore, the median agricultural plot size in Bangladesh is around
15 decimals (Ahmed, 2015). Therefore, adapting to soil salinity intrusion by shifting from
rice farming to aquaculture in the dry season would require substantial amount of land
consolidation.
2.2.2 Role of Land Market in Achieving the Scale for Adaptation
In a functioning land market, farmers with comparative advantage in aquaculture can
rent/purchase contiguous plots to consolidate. However, given the historical stagnancy in
land consolidation in many developing countries including in developing countries, it is not
obvious if the land market can achieve the scale required for adaptation (Foster and Rosen-
zweig, 2022). For example, agriculture in Bangladesh is still dominated by labor intensive
farming techniques characterized by low scale and low productivity. In gure 2.2, I superim-
pose the relationship between prot per acre and scale of operation in rice farming against
the distribution of plot size. We can see that the overwhelming mass of the plots are very
small and are characterized by low productivity. This productivity relationship also implies
that there would be a substantial gain in income from consolidation. Despite this arbitrage
opportunity we haven't seen any trend of consolidation in farm size in the recorded history
of Bangladesh.
This stagnancy in land consolidation could be the result of various frictions in the land
market of developing countries (Sood, 2020). For example, there are cultural barriers in
selling lands inherited across generations. Moreover, given the weak institutional environ-
ment, property rights are weak and land owners might be reluctant to participate in a rental
38
contract due to the fear of losing their land rights. Furthermore, if the contracts are not
enforceable, it might not be worthwhile for the renters to rent lands and make large xed
investments there. Finally, even in the absence of any frictions in the land market, there
might be hold up problems that would extract away the surplus which could otherwise in-
centivize farmers to consolidate. To summarize, the extent of misallocation in farm size
distribution implied by gure 2.2 can only be rationalized if there are substantial barriers
in the consolidation process such that the aggregate costs from all the frictions from the
consolidation process must outweigh the aggregate gains from it.
Given the above discussion, we can understand that it is no easy feat for a developing
country land market to help in consolidation and achieve the scale required for adaptation
specially when a large number of those cases would involve consolidation of more than two
plots. On the contrary, if climate change increases the cost of misallocation to the extent that
now the benets of consolidation would outweigh the aggregate cost from all the frictions,
it might be possible to achieve the scale for adaptation through land market overcoming
all the potential barriers that we have mentioned above. This chapter sheds light on this
important question: can land markets in developing countries overcome historical barriers
in consolidation faced with increased risks from climate change? I provide the rst rigorous
causal evidence that rental market for land can help in achieving the scale for climate change
adaptation in the farming sector of developing countries.
39
2.3 Data and Descriptive Statistics
I use the the agricultural census of 2008 described in chapter 1 for data on land use. Fur-
thermore, similar to chapter 1, I also use salinity data from the salinity survey of the soil
resources development institute (SRDI) in 2009.
2.4 Empirical Strategy and Identication
2.4.1 Instrumental Variable Approach
I use a two stage least square approach (2SLS) using the exogenous border between the
mature delta and the tidally active delta as an instrument of whether a village falls within
an area with above 8dS/m salinity to estimate the treatment eects. I use the following
standard specication for the 2SLS approach:
Y
hvd
=
N
v
+
d
+ X
h
+
hvd
(2.1)
In equation 1.2, Y
hv
is the outcome of interest for household h in village v in district d.
The outcome variables are farm size, land ownership, gini coecient of farm size and land
ownership, amount of land rented in and leased out, etc.. Furthermore,
d
represents xed
eect for district d and X
h
represents a vector of household specic covariates. Finally, the
treatment variable N
v
is a dummy variable indicating whether the centroid of village v falls
within the areas with a salinity level of more than 8 dS/m. Here, the parameter of interest
is
. As mentioned before, I use the 8 dS/m threshold for dening treatment because there
is no variety of rice that is resilient beyond this level of salinity (Radanielson et al., 2018).
40
Furthermore, as discussed earlier, the treatment variable N
v
could be endogenous both due
to reverse causality and omitted variable concerns. The identication strategy is to use the
dummy Z
v
indicating whether the centroid of village v falls within the tidally active delta
as an instrument for the treatment variable to circumvent these problems. The exclusion
restriction is that being within tidally active delta only aects the land use choices through
increased salinity level.
The instrumental variable (IV) approach has been used instead of the fuzzy RD approach
of chapter 1 because fuzzy RD approach lacks sucient power in most of the tests. How-
ever, the results from fuzzy spatial RD approach are qualitatively similar in all the cases.
Nonetheless, the identication assumption is stronger for the 2SLS approach
6
. Nonetheless, I
rule out a host of threats to the exclusion restrictions in my analysis to solidify the credibility
of my ndings.
2.5 Results and Interpretations
2.5.1 Eect of Salinity on Farm Size
Firstly, column 1 in table 2.1 shows that increase in salinity level above 8 dS/m result in
a statistically signicant increase in farm size. In particular, the farm size increases by 52
percent over the control mean. Moreover, column 3 shows that there is an associated increase
in the gini coecient of farm size in these areas by around 23 percent. This provides strong
6
Essentially, the exclusion restriction has to be satised for a larger bandwidth in case of 2SLS than that
of the fuzzy spatial RD approach
41
evidence that the reallocation of land from crop farming to aquaculture was made possible
by substantial consolidation of farmlands.
Nonetheless, one important threat to the exclusion restriction is that the dierence in
farm size could arise from pre-existing dierences in land distribution between tidally active
delta and mature delta. But, column 2 and 4 in table 2.1 show that having a salinity
beyond 8 dS/m increases farm size and gini coecient of farm size even when we control
for land ownership and gini coecient of land ownership, respectively. This indicates that
consolidation must have been happening through rental market transactions. In the next
subsection I investigate the underlying mechanism in detail.
2.5.2 Mechanisms
In this subsection I investigate the mechanisms through which the farm size consolidation
takes place. The scale for adaptation might be achieved either through ownership consol-
idation via the land sales market and/or through consolidation of operation via the rental
market for land. However, historically the land sales market has been very inactive in the
developing countries due to the cultural barrier in selling lands inherited from generation to
generation
7
. In consistence with this historical pattern, table 2.2 shows that salinity intru-
sion results in relatively moderate and noisy increase in the total land ownership, and the
gini coecient of total land ownership. This suggests that majority of the land consolidation
potentially occurs through the rental market.
7
Ownership of more than 95 percent of the farmlands in India are derived from inheritance (Foster and
Rosenzweig, 2022)
42
The rental market for land can perform both allocative role and consolidation role. Con-
solidation through rental market would imply that a smaller number of farmers would be
able to rent-in lands in high salinity aected areas, and conditional on being a renter, each
farmer would rent in a higher amount of lands on average. The empirical estimates from
table 2.3 support this mechanism by showing that increase in salinity level above 8 dS/m
results in a 62 percent decline in the number of households who rent-in lands. Furthermore,
conditional on being a renter, the amount of land rented increases by 100 percent. An alter-
native interpretation of this result based on purely allocative role of the land market would
be that a limited number of people have the skills of doing aquaculture and therefore, more
people rent out lands for which only a few renters are available. Results from table 2.5 rule
out this interpretation by showing that increase in salinity level above 8 dS/m doesn't cause
any signicant change in the number of households who rent out lands, or in the amount of
land that is being rented out.
Nonetheless, an important threat to the exclusion restriction is that the result is driven
by pre-existing dierence in the pattern of rental market activity between farmers in the
tidally active delta and mature delta. But, estimates from table 2.4 rule this out by showing
that there is no dierence in the rental market activity in the housing sector between the two
areas. Altogether, these results provide strong evidence that the consolidation role of the
rental market was crucial in achieving the scale required for adaptation to salinity intrusion
in coastal Bangladesh.
43
2.6 Conclusion
Sea level rise and the resulting salinity intrusion changes the comparative advantage of the
coastal farmlands from traditional crop farming to salinity resilient aquaculture. However,
aquaculture has relatively higher economies of scale and reallocating lands from crop farming
to aquaculture would require substantial consolidation of fragmented land parcels in devel-
oping countries. The historical stagnancy in consolidation in the developing countries raises
a question on whether farmers can achieve the scale for climate change adaptation. In this
chapter I show strong evidence that the rental market for land in developing countries can
help in collectively achieving the scale for climate change adaptation in the farming sector.
An important implication of the ndings from this paper is that a functioning land market
can complement private adaption to climate change.
44
Appendix
Figure 2.1: Farm Size Across Countries
Figure 2.2: Misallocation in Agriculture
45
Figure 2.3: Productivity vs. Scale in Agriculture
Farm Size Gini of Farm Size
(1) (2) (3) (4)
1fSalinity> 8dS=mg 33.10

20.33

0.166

0.114

(8.994) (5.153) (0.0251) (0.0155)
Owned Land 0.648

(0.0106)
Gini of Owned Land 0.899

(0.0219)
District FE Yes Yes Yes Yes
Control Mean 62.40 62.40 0.67 0.67
N 1166498 1166498 2505 2505
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 2.1: Eect of Salinity on Farm Size
46
Figure 2.4: Productivity vs. Scale in Aquaculture
Figure 2.5: Scale Comparison: Agriculture vs. Aquaculture
47
Figure 2.6: Plot Size Comparison: Agriculture vs. Aquaculture
Data Source: Bangladesh Integrated Household Survey, 2011
Figure 2.7: Farm Size Distribution: High vs. Low Salinity
48
Figure 2.8: Ownership Distribution: High vs. Low Salinity
Figure 2.9: Consolidation Visual
49
Owned Land Gini of Ownership
(1) (2)
1fSalinity> 8dS=mg 19.25

0.0586

(11.58) (0.0301)
District FE Yes Yes
Control Mean 78.85 0.74
N 1166498 2505
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 2.2: Eect of Salinity on Ownership Consolidation
1fLandRentedin> 0g Land rented-in
Full Sample Conditional Sample
(1) (2) (3) (4) (5) (6)
1fSalinity> 8dS=mg -0.218

-0.216

0.698 -0.568 72.45

50.56

(0.0466) (0.0460) (6.429) (6.620) (25.77) (20.10)
Owned Land -0.000124

0.0642

0.359

(0.0000133) (0.00442) (0.0162)
District FE Yes Yes Yes Yes Yes Yes
N 1166498 1166498 1166498 1166498 373006 373006
Control Mean 0.35 0.35 20.27 20.27 60.23 60.23
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 2.3: Evidence on Consolidation through Rental Market
1fRentedHouseg
(1)
1fTidalDeltag -0.0390
(0.0472)
N 1371460
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 2.4: Evidence on No Dierential Rental Behavior in Housing Market
50
1fLandLeasedOut> 0g Land Leased Out
Full Sample Conditional Sample
(1) (2) (3)
1fSalinity> 8dS=mg -0.0115 -4.405 -9.947
(0.0462) (6.361) (15.70)
District FE Yes Yes Yes
N 1166498 1166498 181113
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 2.5: Evidence on No Change in Long Run Land Supply
51
Chapter 3
When Does Public Investment Complement or Crowd
Out Private Adaptation? Evidence from Hydraulic
Structures in Coastal Bangladesh
Mohammad Islamul Haque
3.1 Introduction
There has been billions of dollars of investments in building protective infrastructures in
the coastal regions around the world. With the increased threats from climate change in-
duced sea level rise, storm surges, and salinity intrusion, governments and large development
nancing organizations, like the World Bank, are planning to invest even more in building in-
frastructures to protect the coastal communities across the world
1
. While public investment
in infrastructure has a key role in building resilience to climate change in the coastal regions,
it is also important to ensure that it doesn't end up substituting it. It's not uncommon to
have engineering designs that are based on naive assumptions about behavioral response and
1
WB (2022)
52
doesn't take into account the potential impacts of its features on the local farm productivity.
Given the scale of investment in protective infrastructures in the coastal areas around the
world, these mistakes could result in massive misallocation of productive resources.
For example, a major type of infrastructure that are built to protect coastal communities
across the globe are sluice gates at the connecting points between tidal rivers and tributary
channels along the route of the coastal embankments
2
. The role of the sluice gates is to reg-
ulate the
ow of tidal waters from the riverside to the tributary channels of the countryside.
However, there could be two dierent types of gates in terms of discretion over their opera-
tions: (1) fully automated (2) manipulatable. This poses an important design decision that
can have serious resource allocation consequences. While it might seem intuitive that more
discretion should result in better land allocation, it might not be the true with information
frictions and under weak institutional settings. For example, local in
uential people might
forcibly open the sluice gates even if an area is still suitable for crop farming and allowing
in
ow of tidal water might damage crop productivity in monsoon season by increasing the
salinity level permanently
3
. In this chapter, I investigate the impact of having discretion
over sluice gate operation on land use adaptation to soil salinity intrusion.
As discussed in chapters 1 and 2, there has been rapid increase in soil salinity in the
coastal Bangladesh over the last 3-4 decades and farmers can adapt to soil salinity intru-
sion by reallocating their lands from crop farming to salinity resilient aquaculture in the
dry season. However, adapting to soil salinity by reallocating lands to aquaculture would
2
Figure 3.2 shows a typical sluice gate in coastal areas.
3
Local elites might also abuse the discretion over the operation of sluice gates to
ood crop elds with
saline water and appropriate lands from smallholders. This might result in adverse distributional impact
and increased con
icts in the region. However, assessing the distributional impact is beyond the scope of
this project. My focus is on analyzing the impact of gate types on land use eciency as salinity level keeps
rising
53
require allowing the tidal water to
ow inside the farmlands. However, if the sluice gates
get automatically closed during tides, farmers can't trap the tidal saline water to do aqua-
culture. On the contrary, with hoist operated gates, farmers have discretion over allowing
tidal water
ow inside the elds and trap it for doing aquaculture. Therefore, discretion over
the operation of the sluice gates might result in better land use adaptation to soil salinity
intrusion. Furthermore, dry season salinity level is too high to do traditional crop farming
even if the worst aected areas of the coastal region are surrounded by embankments with
functional sluice gates. This is because one of the major mechanisms of soil salinization is
the capillary movement of saline water upward from shallow water tables (Jitrapinate, 2016)
and embankments and sluice gates can't stop the capillary movement. Therefore, automated
sluice gates might have the unintended consequence of crowding out private adaptation.
In this chapter, exploiting the exogenous border between the mature delta and the tidally
active delta, and the plausibly exogenous variation in the type of sluice gates, I use a dif-
ference in regression discontinuity (RD) approach to show that having an automated sluice
gate in an area result in under-adaptation to salinity intrusion. To be specic, having discre-
tion over the operation of the sluice gate almost doubles land use adaptation to soil salinity
intrusion. However, this increase in land allocation to aquaculture doesn't necessarily imply
more ecient land allocation. One particular argument could be that if the gates were not
manipulatable people could do crop farming and therefore manipulatable gates are essen-
tially substituting dry season crop farming by allowing in
ow of tidal water inside the crop
elds. But, I rule out this possibility by showing that gate type does not have a dierential
impact on the amount of land allocated to dry season rice farming. Nevertheless, another
argument could be that there is a permanent increase in salinity from allowing in
ow of
54
tidal water inside the eld in dry season. However, I rule this out by showing that there is
increased land allocation to rain-fed crop farming in areas with manipulatable gate in the
monsoon season. All these evidence give strong indication that manipulatable sluice gates
result in a pareto improvement in land allocation.
This paper contributes to the strand of literature that investigates the impact of public
policies and investment on private adaptation (Baylis and Boomhower, 2019; Wagner, 2022;
Bradt and Aldy, 2022). Bradt and Aldy (2022) is the most closely related paper to this
chapter. They nd a signicant positive impact of constructing
ood protection levees on
privately owned home values. This chapter gives a counterintuitive case where not having

exibility in the operation of the protective infrastructures end up making people worse o.
This chapter also contributes to a growing literature on the impact of hydraulic structures
on farm productivity. For example, Du
o and Pande (2007) and Strobl and Strobl (2011)
nd that irrigation dams have positive impact on agricultural production in the downstream
locations but no such benets for the locations where these are situated. But, the literature
do not have any study that investigates whether the variation in discretion over the operation
of dams matter for farm productivity
4
. To the best of my knowledge this chapter gives rst
evidence on the impact of discretion over dam operation on land use eciency.
4
Dams are essentially larger versions of sluice gates that allow regulating the
ow of river water from the
upstream regions to the downstream regions.
55
3.2 Background
3.2.1 Coastal Embankments and Sluice Gates
The initiative to build coastal embankments and drainage sluice gates started in the late
1950s (Schmidt, 1969). A master plan was developed in 1964, and over 80 percent of the
existing drainage sluice gates in my sample area were built by the year of 1990
5
. Firstly, em-
bankments were built along the banks of tidal rivers to prevent tides from over
owing to the
countryside. Figure 3.1 shows the structure of a typical coastal embankment. Furthermore,
drainage sluice gates were built at the connection of tidal rivers and tributary channels that
runs across the countryside. Hence, drainage sluice gates regulate the in
ow of tidal waters
from rivers to the countryside channels. Figure 3.2 shows how the sluice gates are placed at
the locations of the embankment where tidal rivers connect with tributary channels. This
scheme of embankments and sluice gates are meant to protect low lying communities from
tidal
ooding, storm surges, and also to restrict soil salinity intrusion by stopping over
ow of
saline water to the crop elds in the countryside. The small low lying land areas surrounded
by embankments and sluice gates are known as polders and there are as many as 139 polders
in the coastal region of Bangladesh. Therefore, there are a large number of sluice gates as
well. Table 3.2 shows that 90 percent of the village centroids of our sample are within 3.35
miles (5388 meters) of a drainage sluice gate. Hence, sluice gates are ubiquitous in our study
area as well.
5
Please see table 3.1 for construction of sluice gates by decade
56
3.2.2 Sluice Gate Types and Discretion Over Operation
The most important component of a sluice structure is the gate which regulates the in
ow
and out
ow of water. There are dierent kinds of gates: radial, vertical,
ap. An interesting
dierence among these gates is the variation in the extent of discretion over the gate oper-
ation. In particular, vertical gates have hoists and therefore people in charge of operating
these gates have full discretion over the opening and closing times of the gates
6
. This in turn
means that they can also make decisions on whether to allow the tidal in
ows to enter from
the riverside to the countryside. On the contrary,
ap gates and radial gates do not have
any hoist. These types of gates open and close based on the relative water level between the
countryside and the riverside
7
. Therefore, a drainage sluice structure with a radial or
ap
gate is essentially automated.
3.2.3 Impact of Automated Gate Operation on Land Use
Adaptation to soil salinity intrusion in the farming sector requires discretion over the op-
eration of the sluice gates. Firstly, in the extremely salinity aected areas of the coastal
region dry season salinity level is too high to do traditional crop farming even if the areas
are surrounded by embankments with functional sluice gates. This is because one of the
major mechanisms of soil salinization is the capillary movement of saline water upward from
shallow water tables (Jitrapinate, 2016) and embankments and sluice gates can't stop the
capillary movement. Therefore, even within embankment protected areas the salinity level
rises in the dry season to the extent that it is not feasible to do crop farming protably. On
6
Figure 3.3 shows vertical gates with hoists
7
Figure 3.4 shows how a radial sluice gate automatically opens up when the water pressure on the other
side is relatively higher.
57
the other hand, adapting to soil salinity by reallocating lands to aquaculture would require
allowing the tidal water to
ow inside the elds. However, the radial and
ap gates get
automatically closed during tides and therefore farmers can't trap the tidal saline water to
do aquaculture. On the contrary, the vertical type gates have hoists that the farmers can
use to allow tidal water to
ow inside the elds and trap it for doing aquaculture. Therefore,
discretion over the operation of the sluice gates result in better land use adaptation to soil
salinity intrusion.
Furthermore, the regular semi-diurnal tides only cause shallow inundation of the low
lying farmlands and without causing any threats to their homesteads. Hence, farmers can
do controlled
ooding of their elds and trap the saline water to do aquaculture in the
dry season when the salinity level goes up. Moreover, I will later show evidence that this
deliberate seasonal
ooding do not increase the salinity level permanently because farmers
drain out the saline water and do rain-fed crop farming in the same land once the salinity is
washed away by rainfall at the start of the monsoon.
3.3 Data and Descriptive Statistics
In this chapter, I am using soil salinity data from SRDI soil salinity survey and land use
data from agricultural census of 2008 both of which are described in chapter 1 (BBS, 2010;
SRDI, 2010). Furthermore, I will use data on sluice gate location, construction time, gate
types, features, etc. from the following survey on coastal hydraulic structures.
58
3.3.1 Data on Hydraulic Structures
Institute of Water Modeling (IWM) conducted a survey on all the hydraulic structures of
Bangladesh back in 2016. This was a comprehensive survey that collected data on GPS
locations, all possible features of these hydraulic structures, their existing conditions, and
functionality. It also collected data on the condition of the channels approaching the sluice
gates on both the upstream and downstream sides. IWM also recorded pictures and videos
to corroborate the data of their survey. Table 3.1 shows that more than 50 percent and
80 percent of all the drainage sluice structures were constructed by the year 1979 and 1989
respectively. This highlights that majority of the sluice gates were constructed before rapid
increase in shrimp cultivation after the liberalization of the economy in early 1980s. Fur-
thermore, table 3.2 shows that 90 percent of the village centroids in our sample are within a
distance of 3.35 miles (5388 meters) from the nearest sluice gate. Hence, drainage sluice gates
are pretty ubiquitous in my study area. Finally, table 3.3 shows that 74 percent of sluice
structures have an automated gate, whereas just 26 percent of the gates are manipulatable.
3.4 Empirical Strategy and Identication
3.4.1 Dierence in Discontinuity Approach
I exploit the exogenous border between the mature delta and the tidally active delta, and the
plausibly exogenous variation in the type of sluice gates to employ a dierence in regression
discontinuity (RD) approach for the causal identication of the eect of gate types on land
use. As described in the background section of chapter 1, there is a sharp increase in
59
salinity at the border between mature delta and tidally active delta as one moves from
the former to the latter. Furthermore, as described in the preceding section on hydraulic
structures in coastal Bangladesh, whether a sluice gate is automated or manipulatable is
mostly predetermined. Moreover, I will provide evidence to rule out the possibility that
gate types are systematically decided based on crop or aquaculture productivity. Therefore,
sluice gate type is essentially exogenous. I exploit the above exogenous variations in salinity
and gate types to design a dierence in discontinuity model. I use a standard local linear
specication as follows:
Y
hv
=
N
v
+ N
v
M
v
+
1
d
v
+
2
d
v
M
v
+
1
d
v
N
v
+
2
d
v
N
v
M
v
+
l
+ X
h
+
hv
(3.1)
In equation 1.1, Y
hv
is the outcome of interest for household h in village v. Our main
dependent variables are the amount of farmlands that employed in dierent types of land
use. Moreover, in the above equation,d
v
represents the nearest distance from the centroid of
villagev to the frontier of tidally active delta in kilometers. I use positive distances for villages
within tidally active delta and negative distances for villages in mature delta. In addition,
X
h
is a vector of farm covariates and
l
represents longitude-quartile xed eects. I use
longitude-quartile xed eects, following Ito and Zhang (2020), in order to avoid comparing
villages that are located in distant longitudes. N
v
is a dummy variable indicating whether the
centroid of villagev falls within tidally active delta. Furthermore,M
v
is a dummy indicating
whether the sluice gate in villagev is manipulatable or not. Here, the parameters of interest
60
are
and . I use the 8 dS/m threshold for dening treatment because there is no variety
of rice that is fully resilient beyond this level of salinity (Radanielson et al., 2018).
There are two identication assumptions for this dierence in discontinuity specication.
First assumption is that within a small neighborhood around the frontier between tidally
active delta and mature delta, being located in either side only aects the land use choices
through the dierences in salinity level between these regions. We have provided a host
of tests in chapter 1 to rule out potential threats of violation of this assumption. Second
assumption is that there is no unobservable that determine both the placement of a specic
type of sluice gate and the productivity of crops or aquaculture. I will provide results
supporting this assumption in the next section.
Finally, I use local linear regression with triangular kernel weighting and Mean Square
Error (MSE) optimal bandwidth to estimate the RD treatment eects (Calonico et al., 2014).
Moreover, since treatment varies at village level, I cluster the standard errors at village level,
but the results are robust to clustering standard errors at a higher administrative level such
as unions.
3.5 Results and Interpretations
Firstly, I estimate the eect of having a manipulatable gate on the amount of land allocated
to aquaculture within the tidally active delta. Table 3.4 shows the estimates from the above
dierence in discontinuity approach for the parameters
and . The baseline specications
in column 1 and 4 show that having a manipulatable sluice gate results in a statistically
signicant increase in the amount of land allocated to aquaculture. In fact, the amount
61
of land allocated to aquaculture almost doubles in places with a manipulatable sluice gate.
Furthermore, this result is fully robust to controlling for a set of farm covariates (column
2 and 5) and sluice covariates (column3 and 6). Moreover, the results are not sensitive to
choices of dierent bandwidths (column 1, 2, 3 vs. columns 4, 5, 6).
One particular argument could be that if the gates were not manipulatable people could
do crop farming and therefore manipulatable gates are essentially substituting dry season
crop farming by allowing in
ow of tidal water inside the crop elds. But, table 3.5 shows
that gate type does not have a dierential impact on the amount of land allocated to dry
season rice farming. This indicates that the the automated gates can't stop salinity intrusion
(capillary movement) and lands mostly remain unused if the farmers can't do aquaculture.
Hence, manipulatable gates seem to have increased the eciency of land allocation with
season.
Nevertheless, another argument could be that there is a permanent increase in salinity
from allowing in
ow of tidal water inside the eld in dry season. However, table 3.6 shows
that if anything there is increased land allocation to rain-fed crop farming in areas with
manipulatable gate, implying that allowing in
ow of tidal water doesn't increase salinity
level permanently. The farmers can easily drain out the saline water before monsoon and
the salinity level goes down naturally due to heavy rainfall during monsoon. Hence, doing
aquaculture in the dry season doesn't necessarily reduce the cropping suitability in the
monsoon season. Hence, having manipulatable gates can allow farmers to have increased
farm income in the dry season without sacricing their income in the other seasons of the
year. This is essentially a strict pareto improvement in land allocation.
62
However, one of the major threats in the identication is that the gate types could
have been strategically chosen depending on the productivity of rice and aquaculture in the
specic regions. We rule out this possibility by taking advantage of the timing of economic
liberalization and expansion of aquaculture in Bangladesh. More than 95 percent of the
shrimps and crabs produced in the coastal aquaculture farms are exported to US, Europe,
and Japan. Hence, it is no surprise that allocation of land to aquaculture expanded after the
liberalization of the economy in the early to mid 1980s opened up opportunities to export
shrimps and crabs to the US and Europe. This is supported by our earlier observation
from the Landsat 5 satellite image in gure 1.5 that the dry season aquaculture was limited
even in the year 1988. Hence, the rst test we conduct is on whether the gate types varies
systematically for the sluice structures constructed before and after the liberalization of the
country. Table 3.7 shows that the gate types do not vary systematically before and after
the start of the liberalization of the economy in 1980. This result is robust to changes
in the cuto year as well. Furthermore, we test whether the results on increased land
allocation to aquaculture in areas with manipulatable gates hold if we focus on the sub-
sample where the sluice gates were constructed before 1980. Table 3.8 shows that there is
increased contemporary land allocation in aquaculture in places with manipulatable gates
even if we restrict our sample to places where the gates were constructed before the start
of the liberalization of the economy in early 1980s. Together, these two tests give strong
evidence that the the gate types are not driven by suitability of aquaculture and crop farming
and the under adaptation to salinity in places with automated gates is most likely a causal
eect of not having discretion over the operation of the gates.
63
3.6 Conclusion
Governments and large development nancing organizations are investing billions of dollars
in constructing coastal protections. However, often the engineering designs of these infras-
tructures don't take into account the impact of dierent features of these infrastructure on
local productivity and behavioral responses. One key consideration in many of the coastal
infrastructure projects would be to decide whether the local people should have discretion
over the operation of the sluice gates that regulate the in
ow and out
ow of tidal waters
from the ocean. In this chapter I show evidence that not having discretion over the opera-
tion of sluice gates result in under adaptation to salinity intrusion in the coastal regions of
Bangladesh. Given the scale of the investments in hydraulic structures for coastal protection
in Bangladesh, such crowding out of private adaptation imply massive misallocation in the
farming sector of coastal Bangladesh.
64
Appendix
Figure 3.1: Coastal Embankment
Decades Frequency Percentage Cumulative Percentage
1960 108 25.47 25.47
1970 120 28.30 53.77
1980 113 26.65 80.42
1990 20 4.72 85.14
2000 35 8.25 93.40
2010 28 6.60 100.00
Total 424 100.00
Construction year is missing for 24 Drainage Sluice Structures
Table 3.1: Construction Time of Drainage Sluice Gates
65
Figure 3.2: Sluice Gate
Figure 3.3: Hoist Operated Vertical Sluice Gate
66
Figure 3.4: Automated Radial Sluice Gate
Percentages of Villages Distances from Sluice Gates (meters)
1% 237.987
5% 504.2295
10% 699.2922
25% 1258.876
50% 2125.543
75% 3582.41
90% 5388.276
95% 6968.489
99% 11234.2
Mean 2741.196
Std. Dev. 2213.812
Table 3.2: Distance of sluice Gates from Study Villages
Frequency Percentage Cumulative
Automated Gates 329 74.10 74.10
Manipulatable Gates 115 25.90 100.00
Total 444 100.00
Table 3.3: Type of Sluice Gates
67
Amount of Land in Aquaculture (decimals)
(1) (2) (3) (4) (5) (6)
1fTidalDeltag 30.02

31.17

20.46

21.13

23.03

11.83
(13.36) (12.83) (11.59) (12.55) (12.00) (11.17)
1fTidalDeltag*Manipulatable 29.64

32.38

26.33

37.58

37.12

26.55

(15.94) (14.03) (14.54) (15.67) (13.79) (13.16)
BW 4.75 4.75 4.75 5.75 5.75 5.75
Longitude-quartile FE Yes Yes Yes Yes Yes Yes
Farm Controls No Yes Yes No Yes Yes
Sluice Controls No No Yes No No Yes
N 83943 83943 83943 101432 101432 101432
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 3.4: Eect of Gate Types on Land Allocated to Aquaculture
Amount of Land in Dry Season Rice Farming (decimals)
(1) (2) (3) (4) (5) (6)
1fTidalDeltag -1.222

-1.266

-1.497

-1.202

-1.241

-1.393

(0.523) (0.526) (0.656) (0.473) (0.472) (0.540)
1fTidalDeltag*Manipulatable 0.500 0.534 0.0820 0.359 0.363 0.0485
(0.419) (0.421) (0.575) (0.356) (0.360) (0.494)
BW 4.75 4.75 4.75 5.75 5.75 5.75
Longitude-quartile FE Yes Yes Yes Yes Yes Yes
Farm Controls No Yes Yes No Yes Yes
Sluice Controls No No Yes No No Yes
N 83943 83943 83943 101432 101432 101432
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 3.5: Eect of Gate Types on Dry Season Rice Farming
68
Amount of Land in Monsoon Season Rice Farming (decimals)
(1) (2) (3) (4) (5) (6)
1fTidalDeltag -7.263 -10.46 -8.444 -4.924 -6.822 -4.608
(14.12) (10.49) (10.78) (13.72) (10.32) (10.34)
1fTidalDeltag*Manipulatable 27.35 32.48

42.50

22.24 22.48 33.54

(30.24) (15.33) (19.77) (29.76) (14.36) (19.51)
BW 4.75 4.75 4.75 5.75 5.75 5.75
Longitude-quartile FE Yes Yes Yes Yes Yes Yes
Farm Controls No Yes Yes No Yes Yes
Sluice Controls No No Yes No No Yes
N 83943 83943 83943 101432 101432 101432
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 3.6: Eect of Gate Types on Monsoon Season Rice Farming
Manipulatable
(1)
1fPost 1980g 0.0171
(0.0273)
N 1134199
Standard errors in parentheses are clustered at union level

p< 0:10,

p< 0:05,

p< 0:01
Table 3.7: Evidence on No Strategic Gate Placement Post Liberalization
Amount of Land in Aquaculture (decimals)
(1) (2) (3) (4) (5) (6)
1fTidalDeltag 30.98

31.25

20.45 23.26 24.62

13.35
(14.77) (14.12) (12.60) (14.11) (13.48) (12.24)
1fTidalDeltag*Manipulatable 25.20 29.04

22.42 34.18 34.49

26.71

(19.84) (16.42) (15.84) (20.61) (16.61) (15.29)
BW 4.75 4.75 4.75 5.75 5.75 5.75
Longitude-quartile FE Yes Yes Yes Yes Yes Yes
Farm Controls No Yes Yes No Yes Yes
Sluice Controls No No Yes No No Yes
N 73590 73590 73590 87036 87036 87036
Standard errors in parentheses are clustered at village level

p< 0:10,

p< 0:05,

p< 0:01
Table 3.8: Evidence on Eect of Gate Types with Pre Liberalization Sample
69
Bibliography
Adamopoulos, T. and D. Restuccia (2014). The size distribution of farms and international
productivity dierences. American Economic Review 104 (6), 1667{1697.
Ahmed, A. (2015). Bangladesh integrated household survey (bihs) 2011-2012.
Auhammer, M. (2018). Quantifying economic damages from climate change. Journal of
Economic Perspectives 32 (4), 33{52.
Bassi, V., R. Muoio, T. Porzio, R. Sen, and E. Tugume (2022). Achieving scale collectively.
Econometrica 90 (6), 2937{2978.
Baylis, P. and J. Boomhower (2019). Moral hazard, wildres, and the economic incidence of
natural disasters.
BBS (2010). Census of agriculture 2008. national series, volume 1.
Blakeslee, D., R. Fishman, and V. Srinivasan (2020). Way down in the hole: Adaptation to
long-term water loss in rural india. American Economic Review 110 (1), 200{224.
Bradt, J. T. and J. E. Aldy (2022). Private benets from public investment in climate
adaptation and resilience.
Burke, M. and K. Emerick (2016). Adaptation to climate change: Evidence from us agricul-
ture. American Economic Journal: Economic Policy 8 (3), 106{40.
Calonico, S., M. D. Cattaneo, and R. Titiunik (2014). Robust data-driven inference in the
regression-discontinuity design. The Stata Journal 14 (4), 909{946.
Caunedo, J. and N. Kala (2021). Mechanizing agriculture. Technical report, National Bureau
of Economic Research.
Chen, J. and V. Mueller (2018). Coastal climate change, soil salinity and human migration
in bangladesh. Nature Climate Change 8 (11), 981{985.
Costinot, A., D. Donaldson, and C. Smith (2016). Evolving comparative advantage and the
impact of climate change in agricultural markets: Evidence from 1.7 million elds around
the world. Journal of Political Economy 124 (1), 205{248.
Dell, M., B. F. Jones, and B. A. Olken (2012). Temperature shocks and economic growth:
Evidence from the last half century. American Economic Journal: Macroeconomics 4 (3),
70
66{95.
Desch^ enes, O. and M. Greenstone (2007). The economic impacts of climate change: evi-
dence from agricultural output and random
uctuations in weather. American economic
review 97 (1), 354{385.
Du
o, E. and R. Pande (2007). Dams. The Quarterly Journal of Economics 122 (2), 601{646.
Foster, A. D. and M. R. Rosenzweig (2022). Are there too many farms in the world? labor
market transaction costs, machine capacities, and optimal farm size. Journal of Political
Economy 130 (3), 636{680.
Hagerty, N. (2021). Adaptation to surface water scarcity in irrigated agriculture. Working
paper.
Hornbeck, R. and P. Keskin (2014, January). The historically evolving impact of the ogal-
lala aquifer: Agricultural adaptation to groundwater and drought. American Economic
Journal: Applied Economics 6 (1), 190{219.
Hsiang, S. (2016). Climate econometrics. Annual Review of Resource Economics 8, 43{75.
Islam, S. N. and A. Gnauck (2008). Mangrove wetland ecosystems in ganges-brahmaputra
delta in bangladesh. Frontiers of Earth Science in China 2 (4), 439{448.
Ito, K. and S. Zhang (2020). Willingness to pay for clean air: Evidence from air purier
markets in china. Journal of Political Economy 128 (5), 1627{1672.
Jitrapinate, N. (2016). Capillary rise simulation of saline waters of dierent concentrations
in sandy soils. Engineering and Applied Science Research 43 (2), 78{84.
L az ar, A. N., R. J. Nicholls, J. W. Hall, E. J. Barbour, and A. Haque (2020). Contrasting
development trajectories for coastal bangladesh to the end of century. Regional Environ-
mental Change 20, 1{14.
Mendelsohn, R. (2008). The impact of climate change on agriculture in developing countries.
Journal of Natural Resources Policy Research 1 (1), 5{19.
Mendelsohn, R., W. D. Nordhaus, and D. Shaw (1994). The impact of global warming on
agriculture: a ricardian analysis. The American economic review, 753{771.
NIC (2017). Global trends: Paradox of progress. Technical report, Oce of the Director of
National Intelligence.
NSMIP (2020). A security threat assessment of global climate change. Technical report, The
Center for Climate and Security, an institute of the Council on Strategic Risks.
Passalacqua, P., S. Lanzoni, C. Paola, and A. Rinaldo (2013). Geomorphic signatures of
deltaic processes and vegetation: The ganges-brahmaputra-jamuna case study. Journal of
Geophysical Research: Earth Surface 118 (3), 1838{1849.
71
Pethick, J. and J. D. Orford (2013). Rapid rise in eective sea-level in southwest bangladesh:
Its causes and contemporary rates. Global and Planetary Change 111, 237{245.
Radanielson, A. M., O. Angeles, T. Li, A. M. Ismail, and D. S. Gaydon (2018). Describing
the physiological responses of dierent rice genotypes to salt stress using sigmoid and
piecewise linear functions. Field Crops Research 220, 46{56. Rice and Abiotic Stresses:
Part II.
Rahman, M., G. Penny, M. Mondal, M. Zaman, A. Kryston, M. Salehin, Q. Nahar, M. Islam,
D. Bolster, J. Tank, and M. M uller (2019). Salinization in large river deltas: Drivers,
impacts and socio-hydrological feedbacks. Water Security 6, 100024.
Rokni, K., A. Ahmad, A. Selamat, and S. Hazini (2014). Water feature extraction and
change detection using multitemporal landsat imagery. Remote sensing 6 (5), 4173{4189.
Schlenker, W., W. Michael Hanemann, and A. C. Fisher (2005). Will us agriculture really
benet from global warming? accounting for irrigation in the hedonic approach. American
Economic Review 95 (1), 395{406.
Schlenker, W. and M. J. Roberts (2009). Nonlinear temperature eects indicate severe
damages to us crop yields under climate change. Proceedings of the National Academy of
sciences 106 (37), 15594{15598.
Schmidt, O. (1969). East pakistan coastal embankment project. The Professional Geogra-
pher 21 (4), 252{254.
Sood, A. (2020). Land market frictions in developing countries: Evidence from man-
ufacturing rms in india. Technical report, Mimeo. https://aradhyasood. github.
io/Sood Land Frictions India1. pdf.
SRDI (2010). Saline soils of bangladesh.
Strobl, E. and R. O. Strobl (2011). The distributional impact of large dams: Evidence from
cropland productivity in africa. Journal of development Economics 96 (2), 432{450.
Tanji, K. K. and N. C. Kielen (2002). Agricultural drainage water management in arid and
semi-arid areas. FAO.
Taraz, V. (2018). Can farmers adapt to higher temperatures? evidence from india. World
Development 112, 205{219.
Wagner, K. R. (2022). Adaptation and adverse selection in markets for natural disaster
insurance. American Economic Journal: Economic Policy 14 (3), 380{421.
WB (2022). Continued investment in coastal resilience is critical for sustainable growth
in bangladesh. https://www.worldbank.org/en/news/press-release/2022/09/11/
world-bank-continued-investment-in-coastal-resilience-is-critical-for-sustainable-growth-in-bangladesh.
72

Abstract (if available)

Abstract This thesis provides novel evidence on three key issues related to climate change adaptation in the coastal regions of developing countries: (1) can farmers adapt to soil salinity intrusion? (2) what mechanism enables the farmers to consolidate lands when adaptation involves switching to a farming with higher scale? (3) when does investment in coastal protection crowd out private adaptation? I study these questions in the context of salinity intrusion in coastal Bangladesh. While a technique to adapt to salinity intrusion is reallocating lands from crop farming to salinity resistant aquaculture, the latter has relatively higher economies of scale compared to that of the former. However, farmlands in developing countries are highly fragmented and land market frictions have caused historical stagnancy in farm size consolidation. Hence, if there is no consolidation, farmers operating fragmented land parcels would fail to adapt, resulting in mass loss of livelihoods. Exploiting an exogenous variation in soil salinity level arising from a hydrological feature of coastal Bangladesh, I employ both fuzzy regression discontinuity design and instrumental variable approaches and find that there is a significant increase the amount of land allocated to aquaculture in the areas affected by high level of salinity intrusion. Moreover, this reallocation involved a significant increase in farm size. Furthermore, I find strong evidence that rental market for land performs substantial consolidation role in achieving the scale required for doing aquaculture. Finally, exploiting the plausibly exogenous variation in sluice gate types, I find evidence that the extent of adaptation doubles when local people have discretion over the operation of the hydraulic structures deployed for coastal protections. Together, all the evidence in this thesis suggest that farmers can indeed adapt to salinity intrusion by reallocating their lands and a functioning land market and flexibly designed protective infrastructures can complement the adaptation process.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

PDF

Essays in environmental economics

PDF

Essays on urban and real estate economics

PDF

The impact of minimum wage on labor market dynamics in Germany

PDF

Essays on narrative economics, Climate macrofinance and migration

PDF

Essays on development economics and adolescent behavior

PDF

Essays on environmental economics: education, employment, and experiments

PDF

Three essays on economics of early life health in developing countries

PDF

Essays on the new economic history of the Middle East

PDF

Three essays on the microeconometric analysis of the labor market

PDF

Essays on education and institutions in developing countries

PDF

Evolution of returns to scale and investor flows during the life cycle of active asset management

PDF

Economic policy-making in a changing environment: the politics of success in Peru vis-à-vis China

PDF

Signs of skilled adaptation in the co-speech ticking of adults with Tourette's

Asset Metadata

Creator Haque, Mohammad Islamul (author)

Core Title Essays on the economics of climate change adaptation in developing countries

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Degree Conferral Date 2023-05

Publication Date 10/20/2024

Defense Date 03/22/2023

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

Tag climate change adaptation,consolidation,farm productivity,farming scale,land market,OAI-PMH Harvest,salinity intrusion

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Kahn, Matthew (committee chair), Oliva, Paulina (committee chair), Green, Richard (committee member)

Creator Email islamulhaque30@gmail.com,mohammah@usc.edu

Unique identifier UC113057190

Identifier etd-HaqueMoham-11679.pdf (filename)

Legacy Identifier etd-HaqueMoham-11679

Document Type Dissertation

Format theses (aat)

Rights Haque, Mohammad Islamul

Internet Media Type application/pdf

Type texts

Source 20230421-usctheses-batch-1027 (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