Research papers Authors Grazia Pacillo, Cuong Nguyen Viet, Silvi Hafianti, Kseniya Abanokova, Hai-Anh Dang, Harold Armando Achicanoy Estrella, and Peter Laderach Coordination CĂŠcilia Poggi
SEPTEMBER 2020 No. 147
Who bears the burden of climate variability? A comparative analysis of the impact of weather conditions on inequality in Vietnam and Indonesia
Agence française de développement
Papiers de recherche
AFD Research Papers
Les Papiers de Recherche de l’AFD ont pour but de diffuser rapidement les résultats de travaux en cours. Ils s’adressent principalement aux chercheurs, aux étudiants et au monde académique. Ils couvrent l’ensemble des sujets de travail de l’AFD : analyse économique, théorie économique, analyse des politiques publiques, sciences de l’ingénieur, sociologie, géographie et anthropologie. Une publication dans les Papiers de Recherche de l’AFD n’en exclut aucune autre.
AFD Research Papers are intended to rapidly disseminate findings of ongoing work and mainly target researchers, students and the wider academic community. They cover the full range of AFD work, including: economic analysis, economic theory, policy analysis, engineering sciences, sociology, geography and anthropology. AFD Research Papers and other publications are not mutually exclusive. The opinions expressed in this paper are those of the author(s) and do not necessarily reflect the position of AFD. It is therefore published under the sole responsibility of its author(s).
Les opinions exprimées dans ce papier sont celles de son (ses) auteur(s) et ne reflètent pas nécessairement celles de l’AFD. Ce document est publié sous l’entière responsabilité de son (ses) auteur(s).
1
Who bears the burden of climate variability? A comparative analysis of the impact of weather conditions on inequality in Vietnam and Indonesia Grazia PACILLO International Center of Tropical Agriculture (CIAT), Rome, Italy Cuong NGUYEN VIET Mekong Development Research Institute, Hanoi, Vietnam Silvi HAFIANTI DefinitAsia, Yojakarta, Indonesia Kseniya ABANOKOVA NSU Higher School of Economics, Moscow, Russia Hai-Anh DANG Development Data Group, World Bank Harold Armando ACHICANOY ESTRELLA International Center of Tropical Agriculture (CIAT), Cali Colombia Peter LADERACH CCAFS CGIAR research programme
Abstract Is climate variability regressive? One argument could be as follows: People living in areas with high risk of climate hazards usually correspond to the most disadvantaged populations. Due to existing structural inequalities, they have limited opportunities to cope with climate hazards and often fall into a spiral of further poverty and social exclusion. In this paper, we investigate whether climate variability indeed has a regressive effect in Vietnam and Indonesia where both climate variability and inequality have been increasing. We directly analyse the effect of annual and seasonal temperature on income and income inequality across years. We do so by looking at the Vietnamese and Indonesian populations as a whole and also investigating more in-depth how these impacts change for the most vulnerable and marginalised groups. Our results suggest that climate variability increases inequality and that its biggest burden is bore by existing vulnerable groups. In Indonesia, these groups are rural, farming, low educated, female headed households, whose income is significantly reduced because of changes in climate conditions. Similarly, in Vietnam, ethnic minorities, rural, farming, and agricultural households bear the biggest impact of climate variability. Interestingly, some households in Vietnam are able to completely offset short-term impact of climate variability, using remittances and transfer as an insurance, but our findings also show that their coping strategy does not withstand longer term impacts of persistent climate variability.
2
Despite the remarkable efforts of the national governments in supporting most vulnerable and marginalised groups in the Vietnamese and Indonesian societies in the past decades, specific interventions are needed to address the needs of those who are still bearing the biggest burden of climate impacts to finally allow even the “last mile� groups to escape poverty and exclusion. Keywords Inequality; Climate variability; Indonesia; Vietnam. Acknowledgements This research project is part of the EU-AFD Research Facility on Inequalities, with the financial support of the European Union. The authors are thankful to Nam Viet Nguyen and Lara Barange for their great support in data and content analyses. The views expressed herein reflect those of the authors and do not reflect the official position of the European Union, AFD or other institutions involved in the study. JEL Classification D12, D130, O1, O120, O130, O5, Q1, Q12 Original version English Accepted August 2020
Résumé Le changement climatique et la variabilité sont-ils régressifs? Un argument pourrait être le suivant : les personnes vivant dans une zone à haut risque de variabilité climatique correspondent généralement aux populations plus défavorisées qui, en raison des barrières socioéconomiques structurelles et des inégalités existantes, ont des possibilités limitées d'améliorer leur bien-être et tombent souvent dans une spirale de pauvreté et d’exclusion sociale. Les aléas climatiques pourraient avoir donc des impacts disproportionnellement plus élevés sur ces personnes, avec une grande vulnérabilité et moindre de capacité à absorber et de se remettre de ces dommages. Dans cet article nous investigons si la variabilité climatique aurait effectivement un effet régressif dans deux des économies les plus dynamiques d'Asie du Sud-Est : le Vietnam et l'Indonésie. Dans ces pays la variabilité climatique et les inégalités se sont accrues. Nous analysons l'effet de la température annuelle et saisonnière sur les revenus et les inégalités de revenus au fil du temp. Nous examinons les populations vietnamiennes et indonésiennes dans leur ensemble et en étudions plus en profondeur comment ces impacts changent pour les groupes les plus vulnérables et marginalisés. Nos résultats suggèrent que la variabilité climatique aggrave les inégalités et que son plus gros fardeau est senti par les groupes vulnérables ou les groupes qui, en raison des barrières structurelles existantes, sont devenus de moins en moins capables de faire face aux aléas
climatiques. En Indonésie, ces groupes sont des ménages ruraux, agricoles, peu scolarisés et dirigés par des femmes, dont revenus sont considérablement réduits en raison de changements des conditions climatiques. De même, au Vietnam les ménages ruraux et agricoles sont les plus touchés par la variabilité climatique. De plus, nos résultats montrent que les ménages des minorités ethniques sont en fait à la traîne du reste de la population au Vietnam. Il est intéressant de noter que certains ménages au Vietnam sont en mesure de compenser complètement l'impact à court terme de la variabilité climatique, en utilisant les envois de fonds et les transferts comme une assurance, mais nos résultats montrent également que leur stratégie d'adaptation ne résiste pas aux impacts à long terme de la variabilité climatique persistante. Malgré les efforts remarquables des gouvernements nationaux pour soutenir les groupes les plus vulnérables et marginalisés au Vietnam et en Indonésie au cours des dernières décennies, nous constatons que des interventions spécifiques sont nécessaires pour répondre aux besoins de ceux qui supportent encore le plus grand fardeau des impacts climatiques pour enfin permettre même aux groupes du « dernier kilomètre » d'échapper à la pauvreté et à l'exclusion. Mots-clés Inégalités, Variabilité climatique, Indonésie, Vietnam
3
Introduction Climate change and variability in South
and 0.76 째C for 1985-2005, and by the
East Asia (SEA) have been remarkable in
2060s it is expected to increase between
the past 20 years. The Germanwatch, for
0. 9째 C a nd 2 . 2째 C fr o m 20 12 le vel s
example, estimates that half of SEA falls in
(Karmalkar et al., 2012). Annual rainfalls
the top 10 most affected areas by climate
have also increased (by 12% in the past
change in the past 20 years (Eckstein et
30 years). With an estimated further
al, 2019). Among these areas, Vietnam
increase of 15% in the next 3 or 4 decades,
and Indonesia have experienced an
wetter and drier seasons are expected
increased incidence of extreme climate
especially for the regions south of the
hazards and uncertainties.
equ at or (i nc lu di ng Ja v a an d B ali) (Climate Service Center, 2015).
In Vi et n a m, a ve ra g e su r f ac e temperatures have increased by 1 degree
The impact of climate variability is bound
Celsius, over the last 40 years, with
to be non-negligible in the SEA region. By
southern provinces of the Central
the end of the 21st century, it is estimated
Highlands and Central Coast provinces
that the region will lose about 11% of its
warming even more. Between the end of
GDP because of climate change (ADB,
2015 and early 2016, El Nino events caused
2015). Empirical evidence suggests that
extensive droughts and consequent
the biggest losses will be faced by those
reduction of groundwater availability in
sectors that rely more on climatic
many provinces, especially in the Central
conditions, such as agriculture (First,
Highlands (FAO, 2016). Mountainous areas
2019). Agriculture is, however, the biggest
in the northern regions are increasingly
economic source of income for poorer
experiencing devastating floods and
people in SEA (Booth, 2019). In Vietnam,
finally, with more than 70% of its
for instance, 96% of the poor population
population living in coastal areas and
derives its livelihood from agriculture
low -lyi n g delt as, V ie tn a m is h ig hl y
(Pimhidzai, 2018). Similarly, in Indonesia,
exposed to riverine and coastal flooding
most of the farming households live
(Bangalore et al., 2018).
below the poverty line and heavily relies on agriculture for their subsistence
Similarly, about 40% of the population in
(Bresciani and Valdes, 2007; Suryahadi
Indonesia face high mortality risks due
and Hadiwidjaja, 2011). Therefore, climate-
to multiple climate hazards, such as
induced losses on crop and livestock
tsunami, floods, landslides, drought, and
pr odu c ti vit y ar e ex pec t ed t o b e
earthquakes (Leitmann, 2009). Climate
regressive in nature as they can severely
change has increased the occurrence of
worsen the life of the poorest, further
droughts especially in southern islands,
increasing economic inequality in these
the severity of floods and cyclone
countries (Fisher et al, 2002; Hallegatte et
intensity across the country, and sea-
al, 2014; Farbotko, 2020).
level rise effects in coastal areas (WB, 2014). Average temperatures have
Other studies, however, suggest that
increased steadily in the past 40 years,
agricultural is not the only sector that
within a range of 0.64 째C for 1960-2006
is an d will be i m pac ted b y c li mat e 4
variability. Dell et al (2009) and Hsiang
These areas are often remote, rural and
(2010), for example, find that the effects of
with infrastructure that is less resilient to
climate change are experienced very
climate impacts. Furthermore, more than
clearly also, and sometimes even more,
a third of the poor in Vietnam belongs to
by non-agricultural economic activities.
ethnic minorities (Pimhidzai, 2018). Scarce
There are two main channels for this
political representation, marginalisation
effect: labour productivity and supply. For
an d s oc ial n or m s o f te n hi n der t he
example, Graff Zivin and Neidell (2014)
capacity of ethnic minorities to access a
claim that the actual amount of work
diversified set of productive assets ,
time was significantly reduced during
improve their current economic status
warmer days and that the effect spans
and reduce their vulnerability to climate
across different sectors. Moreover,
hazards (MDRI & Oxfam, 2020). Finally,
Seppannen et al (2013) find that a 2%
women generally lack equal employment
reduc ti o n in l ab ou r pr odu c ti vit y is
and social opportunities, as the existence
associated with temperature above 25
of conservative social norms often
degrees Celsius. The rapid development
relegate them to their traditional role in
of countries like Vietnam and Indonesia
the households, such as chores and
has also moved a lot of people out of
child-care (Lawler & Patel, 2012; Leichenko
farm activities to wage and non-farm
and Silva, 2014).
employment. Some of these people are
Due to lack of resources and oppor-
most marginalised and disadvantaged
tu nitie s, p o or es t a n d m ost margi-
people, such as migrants. In Vietnam, for
nalised groups are bound to be the most
example, a significant number of poor
exposed and vulnerable to climate
women and ethnic minority migrants are
im pac ts ( Sil v a, 2 016 ). S oc iet al a nd
often employed in construction activities
economic pressure renders these groups
as wage employees far away from their h o me s.
T hei r
li vi ng
an d
less able to make adequate investment
w o rki n g
choice to protect themselves from the
conditions are precarious and highly
impact of climate hazards while the lack
exposed to weather conditions. Therefore,
of voice and representation in the
the overall impact of climate variability
political decision-making process and
on these sectors might also be non-
societal norms hinders their control over
negligible for the most disadvantaged part of the population.
a more equal distribution of national
Economic activities are only one pathway
education, infrastructure and the judicial
whereby the impact of climate could
system).
resources and services (such as health,
widen the gap between the poorest and
This self-reinforcing relationship of
the richest. There other factors that could
unequal relations in roles, functions,
increase exposure, vulnerability to, and
decision rights, and opportunities is often
capacity to cope with climate hazards,
referred as “structural inequality� (Dani
for instance remoteness, ethnicity and
an d de Ha a n, 2 00 8). St ruc tu ral ine-
gender (Leichenko and Silva, 2014). Most
qualities are the product of the inte-
poor and marginalised groups in both
ractions of discriminations based on
Indonesia and Vietnam, for instance, can
gender, age, ethnicity, race, religion ,
often afford to live in less desirable areas.
culture, unequal access to basic services 5
and unequal opportunity for participation
Despite these remarkable results ,
and choice. Several studies show how
inequality has been steadily increasing
structural inequalities significantly reduce
(Gini coefficient +5% in Vietnam and 10%
opportunities to escape poverty
in Indonesia, 1990-2014) (UN, 2018). In
(Andrews and Leigh, 2009; Wilkinson and
Vietnam, inequality has widened even
Pickett, 2009; Kerry et al., 2010; Dang et al.,
more in rural and remote areas, mostly
2020). The interactions between rising
populated by minority ethnic groups (Le
inequality, lower social mobility and
and Booth, 2013; Nguyen et al., 2015; Bui et
higher climate variability can further
al., 2017; Nguyen et al., 2017; Nguyen and
strengthen existing barriers that limit
Nguyen, 2017). Indonesia has become the
disadvantaged and marginalised groups
sixth country with the greatest wealth
ability to cope with and adapt to climate
inequality in the world, with the four
hazards1 (Beck, 2010).
richest people in the country having more wealth than 100 million poor people, all
Studies on the direct impact of climate
together (Asra, 2000; Akita, 2002; Oxfam,
change and variability on structural
20172; Ananda and Pulungan, 2019).
inequalities are limited, as the literature mostly focuses on the direct effects on
Our major contribution in this paper is
poverty and treats inequality as a
that we directly investigate the
secondary and consequential issue
relationship between climate variability
(Leichenko and Silva, 2014). Poverty and
and drivers of structural inequalities
inequality are, however, very distinct
using a within country approach .
phenomena and often follow different
Understanding the direct relationship
patterns. Vietnam and Indonesia are
bet wee n str uc tu ral ine qu alit y a n d
perfect examples. These two countries
climate variability will be of paramount
are among the fastest-growing lower-
importance for the policy development in
middle-income countries in the region.
Vietnam and Indonesia. Hence, although
Their effort in reducing poverty rates has
policies in both countries were designed
been outstanding in the past decade and
to target and support low-income groups
as of now about 10% and 5.8% of the
in coping with emergencies, they often
population lives under the poverty line,
have little relevance to the needs, rights
respectively in Indonesia and Vietnam.
and priorities of the most marginalised people. The lack of participation and voice the places of power has in the recent years worked in favour of the better-off and widened the gap between these and the most vulnerable, reducing even further their ability to face climatic
Inequality is not all bad. Studies show that that some degree of inequality could provide the incentives for investments, human capital accumulation and future economic growth (Lazear and Rosen 1981; Barro 2000; DablaNorriet et al., 2015). Nonetheless, high level of sustained inequality can cause large social costs, reduced individuals’ educational and occupational choices, increased resource misallocation and overall reducing economic growth (Stiglitz, 2012). 1
c h alle ng es ( Ox fa m, 201 7 ; Ox fa m i n Vietnam, 2015; Nguyen Tran Lam et al., 2016; Muhtadi and Warburton, 2020).
https://www.oxfam.org/en/indonesia-evenit/inequality-indonesia-millions-kept-poverty. 2
6
Furthermore, a large majority of the
Synthetic panels approach (Dang and
previous studies analyse the relationship
Lanjouw, 2013; Dang et al, 2019) provides a
of climate and inequality across
robust alternative to real panel data. We
countries, with less attention to how
use this approach to test the relationship
different groups within each country are
between mobility, income and income
impacted by climate hazards (Acemoglu
inequality for climate variability impacts
et al., 2001; Dell et al., 2009; Auffhammer et
and use Vietnam as a case study.
al., 2013). While it is important to discern
The rest of the paper is structured as
differences at regional or global level,
follows: Section 1 describes data and
these studies fail to capture the main
methods; Section 3 discusses the results
socio and economic barriers to climate
of the analysis and Section 4 concludes.
responses. In this paper, we specifically analyse the impact of temperature and rainfall changes on the income and income inequality distribution. In addition to the average country effect, we also investigate whether this relationship changes due to selected drivers of structural inequalities, such as economic activities, gender, ethnicity and remoteness. Finally, a technical contribution. To assess the linkages between social mobility and the impact of climate variability on inequality, we use a synthetic panel approach. We use Vietnam as a case study. Panel data are often regarded as ideal to estimate household level impacts of different nature. This is more so in the case of climate variability studies as one would ideally like to control for those specific households’ features that could confound the results of the estimations across a long-time span. However, long, and large panel datasets are rare3.
VHLSS includes a rotating panel of communes and households, but only for a third of each round sample. Synthetic panels allow us to use all the data available, and not only the ones that are included in the panel. 3
7
I – Data and methods We use five rounds of the regionally representative Indonesian Family Life Survey (IFLS) (1993/94, 1997, 2000, 2007, 2014/15) and nine rounds of the provincial, regional and national representative Vietnam Household Living Standard Survey (VHLSS), collected every two years from 2002 to 2018. Our sample include a panel of 4,909 HHs for Indonesia, for a total of 24,545 observation, and 99,723 households/repeated cross-section observations for Vietnam. Climate data sources are Climate Prediction Center (CPC) of the NOAA ESRL Physical Sciences Division (PSD) for daily min and max temperature and the Climate Hazards Group InfraRed Precipitation with Station (CHIRPS) for daily rainfalls amounts 4. There are many methodological challenges that need to be accounted for when estimating the impact of weather on income and inequality. First, the lack of linearity of the relationship, which means that unusual variability in weather conditions do not have a uniform impact on the income response. For instance, it is reasonable to assume that 1degrees temperature increase around the mean values of the temperature distribution will not have the same effect as a 1-degrees increase at the extreme ends of the distribution (colder or hotter temperatures) (Burke et al., 2015; Dell et al. ,2012; Skoufias et al.,2013; Schlenker and Roberts, 2009; Deryugina and Hsiang, 2017) and this effect might change significantly across seasons (Nurfolk, 2016).
In this paper, we follow the Deryugina and
Hsiang (2017) who use the number of temperature- and rainfall- days to control for nonlinearity in the income and income inequality response function to climate. Climate variables are estimated on an annual and seasonal bases. This approach effectively controls for both non-linearity between climate, weather, and income and inequality responses and for within-year heterogeneity of impacts. Another challenge in the analysis of the impact of weather on income and inequality is the existence of a significant heterogeneities across the units of analysis. Thus, it is reasonable to assume that weather variability does not have the same effect in all countries or across different groups in the population. Unobservable characteristics, such as contextual historical heritage and households’ characteristics, as for example adaptive capacity or creativity (Acemoglu et al.,2001) might confound naïve estimations that do not control for specific fixed effects, as in Dell et al., (2009). Panel data are ideal in this context as they get rid of all location specific unobservable characteristics and they allow for a correct and unbiased comparisons across years of the income and inequality responses (Auffhammer et al, 2013; Dell et al.2012). When panel d ata are not available, multiple dimensions fixed effects might reduce the omitted variable bias in the estimation. We use panel data for Indonesia and for Vietnam we use multiple fixed effects dimensions to control for specific within-commune unobserved characteristics and households’ unobservable characteristics.
4
More details on the data can be found in the Appendix. 8
Following Deryugina and Hsiang (2017), we use the local random deviation of household income and provincial level Gini coefficients to estimate the marginal response to changes in weather, controlling for unobserved heterogenity for locations and households; for spatial auto-correlation across-locations; within-location auto-correlation and nonlinear climate trends. We control for non-linearity of the impact of weather on income and inequality using annual within-communes/sub-district level variations in the distribution of daily temperatures. This approach allows us to estimate the marginal effect of a single day’s temperature on end-of-year income, conditional on temperatures experienced during the rest of the year (Deschenes and Greenstone, 2011). Our empirical equation is the following: �
đ?‘”
đ?‘š đ?‘šâ„Ž đ?‘Œđ?‘–đ?‘Ą = đ?œ‡đ?‘– + đ?œƒđ?‘Ą + ∑đ??ť (đ?‘‡đ?‘–đ?‘Ąđ?‘š )â„Ž + đ?›ž đ?‘šâ„Ž (đ?‘‡đ?‘–đ?‘Ąâˆ’1 )â„Ž ]] + ∑đ?‘š[đ?œ— đ?‘› (đ?‘ƒđ?‘–đ?‘Ą ) + đ?œ‘đ?‘› (đ?‘ƒđ?‘–đ?‘Ąâˆ’1 )] + đ?›˝đ?‘–đ?‘Ą đ?‘?đ?‘–đ?‘Ą â„Ž=1[∑đ?‘š[đ?›˝
(1)
Where Y is the log of income per capita of household i in time t, or the provincial level Gini coefficient for province i in time t, and đ?‘?đ?‘–đ?‘Ą are selected household level characteristics, such as remoteness, gender, education, ethnicity 5, age of the head of the households, proportion of children and females in the household. We use year, đ?œƒđ?‘Ą , and commune (in the case of Vietnam) or household (in the case of Indonesia) fixed effects, đ?œ‡đ?‘– . Location fixed effects controls for unobserved constant differences across location, such as elevation, and households, such as ability and entrepreneurship. Year fixed effects, on the other hand, control for common trends, such as trends in climate or technological innovations. Our main parameters of interest are đ?›˝ đ?‘šâ„Ž and đ?œ— đ?‘› which show the impact of an increase of 1 day of certain temperature or rainfall bins on the growth of income per capita and income inequality at provincial level. The model allows for non-linearity for each dimension of temperature and rainfalls. The dimensions are constructed as eight 3-degrees-temperature bins: 0-12; 12-15; 15-18; 18-21; 21-24; 24-27; 27-30 and more than 30 degrees Celsius6. Similarly, we constructed 8 rainfall bins based on the daily amount of rain experienced: 00-00 mm; 00:05 mm; 05-10 mm; đ?‘”
10-15 mm; 15-20; 20-25 mm; 25-50 mm and more than 50 mm. đ?‘‡đ?‘–đ?‘Ąđ?‘š and đ?‘ƒđ?‘–đ?‘Ą are the number
of days where the 24 hours temperature average and total rainfalls falls in the mth temperature bin and gth rainfall bin. The average effect of daily weather in each of these bins can be identified if the number of days in each of the bins is assumed to be orthogonal to the other potential confounders in the model and conditional to other explanatory variables, so that the effect of, say, an additional 30 degrees Celsius day is estimated by comparing the commune (in the case of Vietnam) or the sub-district (in the case of Indonesia) to itself across the years when the number of days in the 30 degrees Celsius bin was different.
We only add ethnicity as explanatory variables in the model for Vietnam as this feature has not been historically relevant for inequality in Indonesia. 6 Deyugina and Hsiang (2017) use 17 bins for United States where the range of temperature is much higher than Vietnam and Indonesia, where the average temperatures is between 24 and 27 degrees Celsius. Therefore, we grouped the lowest bins into one, 0-12, and then created the following ones with a difference of 3 degrees, as in Deryugina and Hsiang (2017). For Indonesia, which has much average higher temperatures than Vietnam, very few days had temperature less than 18 degrees. Nonetheless, for comparison purposes, we kept the definition of the temperature bins for both countries. 5
9
We exploit the small distortions in the distribution of daily weather conditions across years to estimate their impact on income and income inequality, while controlling for weather systematic patterns and their random variations in each location. Our direct focus is on contemporaneous effects of weather, defined as marginal effect of a single day’s temperature on end-of-year income, conditional on temperatures experienced during the rest of the year (Deschenes and Greenstone, 20117). Temperature and rainfall are, however, serially correlated. Therefore, we add lagged values of both temperature and rainfall to control for any potential effect of past temperature and rainfalls on current output 8. Lagged weather variables are also included in the model to control for the temporal displacement of income and foresight of adaptive capacities across years. In this model we assume constant marginal effects of adaptation efforts, captured by H. Therefore, we keep H=1. Some studies argue that, especially in developed economies, the contemporaneous net effect of weather variability on income might be negligible (e.g. Gallup, Sachs, and Mellinger, 1999; Nordhaus, 2006). This is because the direct, short-term effect on outputs, prices, and revenues, might be offset by adaptive and mitigation measures, national and household level defensive investments and effective risk transfer mechanisms. With the increasing penetration of crop risk insurance schemes and the remarkable climate adaptation and mitigation efforts in both Vietnam and Indonesia, we might expect a certain degrees of temporal displacement of wealth across years due to climate adaptation and mitigation. Several studies have argued that annual measures of climate variability might hide intraannual variation and specific impacts that timing of weather instability can generate (Hsiang, 2010; Mendelshon et al., 1994; Welch et al., 2010; Yang and Choi, 2007; Narloch, 2016). To test whether seasonality affects the relationship between weather and income in our analysis, we also estimate each temperature and rainfall bins for the dry (November to April) and wet (May to October) seasons. Finally, we use standard errors that are clustered in two dimensions (Deryigina and Hsiang, 2017; Cameron et al., 2011) by year and location, communes for Vietnam and sub-districts for Indonesia. This allows to control for both spatial correlation across contemporary locations auto-correlation within locations. We also use population sampling weights. We use this model for each country, Vietnam, and Indonesia, separately, and estimate the average effect across the entire sample as well as across different dimensions of the structural inequalities that are prevalent in the two countries. A repeated cross-section model is used for Vietnam and a panel model for Indonesia. Summary statistics of the main explanatory variables can be found in Appendix 1 (table A1-A2).
We decided to focus on temperature because temperature is highly correlated with rainfall, and therefore we only use rainfall to control for unexpected heterogeneity in the weather impact which is not accounted for by our temperature variables. 8 We also estimated other model specifications, including polynomials (total annual rainfall squared) and variables that capture the deviation from the historical averages of temperature and rainfalls (coefficient of variation). However, our estimates do not change much. Therefore, we prefer following the Deryugina and Hsiang (2017) approach. 7
10
II – Results 2.1. Trends of economic and structural inequalities Income disparities between the richest and poorest have widened through time in both Vietnam and Indonesia. Our data show that some households have been lacking behind, and their economic conditions have deteriorated in the past decades. These households are poor, often live in rural areas and have certain demographic characteristics that are usually associated with structural inequalities, such gender, ethnicity, education, and age of the household head. Nonetheless, peculiar discrepancies across the two countries can be noted. In Vietnam, for instance, crop and overall agricultural income has grown much slower than non-farm and wage employment income (Figure 1, Panel A1). Furthermore, although positive, income growth for the poorest has been much slower than the rest of the population (Panel A2). Agricultural and rural households are significantly lacking behind, as their income has been lower than the average since 2002 and this gap has steadily widened (Panel A3). Our data also show that ethnicity of the head of the household is an important barrier for economic prosperity at household level (Panel A4). Thus, since 2002, ethnic minority households have experienced the lowest income growth in absolute and relative terms. Similarly, households whose head has low education (primary or lower) have also experienced a slow income growth, while female headed households or those households headed by older people are closer to the average growth since 2002. Interestingly, it seems that young households’ heads have been struggling in catching up with other groups in terms of income growth, as their income has been consistently lower than other groups of households across time. Income inequality at national and sub-national level has been relatively stable since 2002, with even a slight reduction noticeable in 2018 (Figure 1, Panel B1)9. The decomposition analysis of income inequality across sources of livelihoods shows that wage income accounts for most of the inequality in Vietnam (+ 49%) followed by non-farm income (31%) (Panel B2). Since 2002, the contribution of wage income to total inequality has consistently increased (+16% in 2018), while non-farm income contribution to inequality has reduced from 2002 to 2004 and increased ever since, although at a much lower rate than wage income. Agricultural income, which includes crop, livestock, fisheries, and forestry income, and other income (transfers, remittances, etc.) contribute very little to total inequality (11% and 9%, respectively) with a slight reduction since 2002 (5% and 6%, respectively). Despite the relatively stable inequality levels at national level, inequality has, in fact, been increasing since 2002 mostly for those provinces where rural, poor, agricultural households are located and, even more, for those with the highest concentration of ethnic minority households. Thus, our data show that inequality increased in those regions more densely
Consistently with official WB figures, expenditure inequality, on the other hand, has increased in the past years, although on a relatively small scale. 9
11
populated by ethnic minority, such as Northern mountainous regions and Central highlands10. In these regions, inequality increased by 12% and 14% since 2002, respectively. To understand whether these patterns are explained by difference between or within groups for each of the most vulnerable groups in Vietnam, we decomposed the Theil’s L of the income per capita. Panel B3 of Figure 1 shows that since 2002, most of the total inequality can be explained by within-groups differences, except for education and ethnicity. For these households, between-groups inequality has increased. In line with previous literature (e.g. McCaig et al. 2015; Tuyen, 2016; Kompass et al., 2017; Pimhidzai, 2018), the largest increase in between inequality has occurred across ethnicities (Kinh vis-à-vis ethnic minorities). In Indonesia, disparities between the richest and the poorest have also been progressively increasing since 1993, faster and wider than in Vietnam. Like Vietnam, though, household income is more and more sourced from non-farm wage employment and less from agricultural activities (Figure 2 – Panel A1). Poorest households have experienced the lowest income growth in the whole population, while richest groups have been become progressively richer and richer at a much faster rate(Panel A2), especially in the Sulawesi and Kalimantan regions where income per capita grew 58% more than in Java and 39% more than in Sumatra since 1993. Poor, agricultural and rural households have generally been lacking behind (Panel A3), especially in the Java region, where farming and poor households have also experienced a reduction of income since 2007. Similarly, female headed households and households where the head is old (more than 60 years old) or with low education (primary or lower) have experienced lower income growth than the average (Panel A4), at national as well as across different regions Our data shows a consistent reduction of income inequality at national and provincial level since 199311, from above 70% to slightly less than 60% (Figure 2 – Panel B1). A spike in inequality occurred in concomitance to the financial crisis in 1997 but reduced significantly since 2000. The source of livelihoods and income that accounts mostly for the total inequality changed from 2000 to 2007 and overall, across all the years in the sample (Panel B2). While non-farm income accounted for about 40% to 50% of total inequality in 1993, 1997 and 2000, its contribution dropped to 14% in 2007 and slightly increased to 16% in 2014. In more recent years, the major source of inequality comes, instead, from other sources, such as remittances and transfers, and from non-farm wage income, both accounting for about 40% of total inequality, with a 18% increase since 1993 of the latter. Farm and farm labour income, on the other hand, do not contribute much to the total inequality. Like in
Regional level figures can be found in the Appendix. WB estimates indicate an equalizing effect of the 1997 financial crisis and an increase in inequality in the last years of the Gini coefficient. Our data shows, instead, an increase of inequality following the crisis and consistent reduction of the national income inequality since 2000. There might be three main reasons for these discrepancies: 1) the dataset used in our analysis, IFLS, is representative at regional level and not at national level. WB, instead, uses SUSENAS datasets which are nationally representative. We did not use SUSENAS data as are not publicly available 2) Furthermore, our panel includes a selection of the households that have been followed in the 5 rounds of the IFLS data collection, located in 13 of the 27 Indonesian provinces. 10 11
12
Vietnam, total inequality is mostly explained by differences within groups. The only exception is for inequality across households with different level of education, for whom we notice a slight increase of the between-groups inequality since 1993 (Panel B3). provinces, Remittances and other non-labour income transfers have significantly increased since 2000s in Indonesia, from 1.8 billion USD to 9.7 billion USD, according to WB estimates12. In our sample, the main recipients of transfers are middle aged widow, less educated, female headed households, with more dependents (children and adults in productive age) than households who do not receive any transfer. They are more likely to be poor and live in urban areas in the Sumatra region and their sources of livelihood are primarily non-labour and non-farm wage income. There exists evidence on migration patterns and lower and unequal employment opportunities that might explain why remittances, transfers and wage income have been the major contributors of inequality in Indonesia. Firstly, internal migration is a young people phenomenon. About 65% of the total migrant population are aged between 15 and 34. Most of them are single, more educated men, who look for better opportunities in the services sector, in urban areas in Java. Middle aged widows are, therefore, more likely to stay in their place of origin and receive remittances from their dependents, especially when poor, less educated and with more dependents (Sukamdi and Mujahid 2015). Women also have lower and unequal employment opportunities in Indonesia. In the past two decades, women labour market participation, and especially married or widows with more dependent children, has been consistently lower than international standards (Comola & de Mello, 2012; WB, 2018; Taniguchi and Tuwo, 2014; Cameron et al., 2019) 13. Unemployment is, however, relatively “unaffordable� for poor households who often rely on the informal sector for otherwise inaccessible job opportunities. About 75% of the Indonesia’s total female workforce is employed in the informal sector (WB, 2019), which pays lower salaries especially for low skilled workers and provides lower quality working environment than the formal sector. There also exists a significant gender wage pay gap in the formal sector, where women earn on average 30% less than men (Weni et al., 2019). Therefore, poor, and less educated women might face significant barriers in accessing equal earning opportunities than the rest of the population, both in the formal and informal sectors.
12
Accessible here: https://data.worldbank.org/indicator/BX.TRF.PWKR.CD.DT?locations=ID&view=chart
In Indonesia, 53% of women in productive age (15 years or above) participate in the labour market, while, in comparison, the corresponding rate in Vietnam is 73% (WB, 2018). 13
13
Figure 1: Annual average real income per capita, by income type (Panel A1), quintiles (Panel A2) and across selected demographic groups (Panel A3 & A4) from 2002 to 2018 in Vietnam. Panel B shows the distribution of the income Gini coefficient, at national and provincial level (B1) its decomposition sources of income (B2) and the GE0 Theil’s L decomposition (B3). Source: Authors’ calculation using VHLSS (2002-2018)
Panel B: Gini and GE0 Theil's L distribution and decomposition Panel B2: Gini coeffient by sources of income
Panel B3: GE0 Theil's L decomposition
100
16% 16% 12% 8%
Gini - national level Gini - province level
W age inc
Farm inc
Nnfarm inc
Other inc
14
W ithin
Urban/Rural areas
Region
Sex HH head
Province
Poor HHs
Non-farm HHs
Education HH head Ethnicity of HH head
Agricultural HHs
Age HH head (>60 yrs old)
2018
2016
2014
2012
2010
2008
2006
2004
2002
2017
2012
2007
2002
0
0
20
20
2002 2018 2002 2018 2002 2018 2002 2018 2002 2018 2002 2018 2002 2018 2002 2018 2002 2018 2002 2018
0
40
40
20
60
60
40
60
80
80
80
100
100
Panel B1: Income Gini coefficient
Between
Figure 2: Annual average real income per capita, by income type (Panel A1), quintiles (Panel A2) and across selected demographic groups (Panel A3 & A4) from 1993 to 2014 in Indonesia. Panel B shows the distribution of the income Gini coefficient, at national and provincial level (B1) its decomposition sources of income (B2) and the GE0 Theil’s L decomposition (B3). Source: Authors’ calculation using IFLS data (1993-2014)
Panel B: Gini and GE0 Theil's L distribution and decomposition Panel B3: GE0 Theil's L decomposition
100
Panel B2: Gini decomposition by sources of income
3% 9%
0
Farm income 1995
2000
2005
2010
Gini - national level Gini - provincial level
2015
Nnfarm lab inc Other income
15
Farm lab inc
Urban/Rural areas
Region
Sex HH head
Province
Poor HHs
Non-farm HHs
Agricultural HHs
Education HH head
1993 1997 2000 2007 2014
Age HH head (>60 yrs old)
0
20
20
1993 2014 1993 2014 1993 2014 1993 2014 1993 2014 1993 2014 1993 2014 1993 2014 1993 2014
0
40
40
20
60
60
40
60
80
80
80
100
100
Panel B1: Income Gini Coefficient
Nnfarm inc Within
Between
2.2. Trends in climate variability Average temperatures and total rainfalls have been varying significantly across time. Our data show that temperature patterns in Vietnam have become increasingly more diverse since 2002, especially during the main dry season (November to April). This is at the exception of the years 2004 and 2006 when communes have experienced similar average temperature in the year In Indonesia, we observe higher temperature variability than in Vietnam, more so during the dry season, for lower temperature level and higher rainfall amounts14 (Figure 3). Figure 3: Average temperature in Vietnamese communes (Panel A) and Indonesian sub-districts (Panel B) during dry and wet seasons in the past 12 months. Source: Authors estimation using CPC-NOAA ESRL (temperature).
Panel A: Vietnam
Dry season
2014
2007
2000
1997
1993
2014
2007
2000
Wet season
1997
2002 2004 2006 2008 2010 2012 2014 2016 2018
Dry season
1993
2002 2004 2006 2008 2010 2012 2014 2016 2018
18
23
20
24
22
25
24
26
26
27
28
28
Panel B: Indonesia
Wet season
In both Vietnam and Indonesia, precipitations do also vary significantly across years. In Vietnam, rainfall variability increased since 2010. 14
16
Average measures of weather conditions can, however, hide differences in variability at different temperatures and rainfall levels. In Vietnam, both during the dry and wet season, most of the days in the year have temperatures ranging between 21 and 30 degrees on average. The highest variability is at the lowest extreme of the distribution (00-12 degrees) and between 21 and 27 degrees, especially since the 2010s. The highest temperature bins (27-30 and more than 30) show the biggest variations. For example, in 2008 temperatures were in very few occasions higher than 30 degrees on average. The year 2010, on the other hand, was an extremely hot year, with more than 80 days of average temperature above 30 degrees in some parts of Vietnam. In 2012 this number dropped to a quarter of what was in 2010 to increase then steadily, but slowly, till 2016 and drop again in 2018 15. In Indonesia, we do not observe any days in the lowest temperature bins (0-12; 12-15 and for most years 15-18). Interestingly though, since 2007 more and more households have experienced days with temperature less than 21 degrees and more than 27 degrees, especially during the dry season. The highest variation is, however, for days with temperature between 21 and 24 degrees, less so in the year 2000, but otherwise consistently varied, again, more in the dry season that in the wet season. Within the most common temperature bin (24-27 degrees), the number of days were quite consistent in 1993 and 1997 but started decreasing since the 2000, with more fluctuation during the wet season. The number of days in the 27-30 degrees bin has been increasing since 2000 confirming the increasing temperature trends observed overall in the country
.
16 17
Rain patterns have also been volatile in Vietnam. The number of no rain days have been quite varied within the years but overall, we do not observe unusual trends. On the other hand, at higher amount of rainfalls (above 14 mm) we observe a much higher variation and occurrence of unusual rainfalls in both wet and dry season. Like the temperature trends, variability in rainfall seems to have increased since 2010, confirming the high correlation between these two variables. 15
Rainfalls have also been erratic since 1993. The number of days within each rainfall bin varies significantly across years. More variation is observed at the tails of the distribution, with more unusual observations for the number of days where rainfall have been more than 50 mm a day. In addition, whilst the number of no rain days fluctuates across years, the overall picture shows an increase in wetter days, especially since 2000. 16
Additional figures on the distribution of temperature and rain days in Vietnam and Indonesia can be found in Figure A1 and A2 in the Appendix. 17
17
Table 1 : Number of temperature and rainfall days from 2002 to 2018 in Vietnam (Panel A) and from 1993 to 2014 in Indonesia (Panel B). Source: Authors estimation using; CPC-NOAA ESRL (temperature); CHIRPS (rainfall) Panel A: Vietnam 2002
2004
2006
2008
2010
2012
2014
2016
2018
Variable
Description
Mean
Mean
Mean
Mean
Mean
Mean
Mean
Mean
Mean
Temperature 00-12 °C
Total number of days in the past 12 months when temperature was between 0 °C and 12 °C
0
0
0
1
0
0
1
2
1
Temperature 12-15 °C
Total number of days in the past 12 months when temperature was between 12 °C and 15 °C
2
6
3
10
2
6
9
4
6
Temperature 15-18 °C
Total number of days in the past 12 months when temperature was between 15 °C and 18 °C
16
11
15
14
12
20
13
13
11
Temperature 18-21 °C
Total number of days in the past 12 months when temperature was between 18 °C and 21 °C
26
26
25
24
22
22
26
23
26
Temperature 21-24 °C
Total number of days in the past 12 months when temperature was between 21 °C and 24 °C
37
43
37
47
37
38
43
30
40
Temperature 24-27 °C
Total number of days in the past 12 months when temperature was between 24 °C and 27 °C
87
93
82
105
101
108
97
83
92
Temperature 27-30 °C
Total number of days in the past 12 months when temperature was between 27 °C and 30 °C
178
164
181
153
150
161
151
176
174
Temperature >30 °C
Total number of days in the past 12 months when temperature was 30 °C or higher
21
23
22
11
43
11
24
36
17
N
27720
8521
8532
8503
9354
9354
9342
9320
9077
18
Panel B: Indonesia 1993
1997
2000
2007
2014
Variable
Description
Mean
Mean
Mean
Mean
Mean
Temperature 00-12 °C
Total number of days in the past 12 months when temperature was between 0 °C and 12 °C
0
0
0
0
0
Temperature 12-15 °C
Total number of days in the past 12 months when temperature was between 12 °C and 15 °C
0
0
0
0
0
Temperature 15-18 °C
Total number of days in the past 12 months when temperature was between 15 °C and 18 °C
0
0
0
0
0
Temperature 18-21 °C
Total number of days in the past 12 months when temperature was between 18 °C and 21 °C
0
0
0
0
0
Temperature 21-24 °C
Total number of days in the past 12 months when temperature was between 21 °C and 24 °C
7
8
8
7
5
Temperature 24-27 °C
Total number of days in the past 12 months when temperature was between 24 °C and 27 °C
246
236
167
173
155
Temperature 27-30 °C
Total number of days in the past 12 months when temperature was between 27 °C and 30 °C
112
121
53
185
205
Temperature >30 °C
Total number of days in the past 12 months when temperature was 30 °C or higher
1
1
1
1
1
N
4909
4909
4909
4909
4909
19
III – Income and income inequality sensitivity to contemporaneous weather trends Using a non-parametric approach, we explored the relationship between weather variability, income, measured as household level income per capita, and income inequality, measured as the Gini coefficient. We use average annual temperature of the past 12 months to compare contemporaneous weather, income, and income inequality. Our data show that the correlation between weather conditions, income and inequality is non-linear (Figure 4). We observe an overall positive correlation between average yearly temperature and household level income per capita in both Vietnam and Indonesia (Panel A & C). In Vietnam, however, household income decreases when the average annual temperature goes above 28 degrees Celsius. This negative impact seems to be mostly experienced in Central Highlands. This region has historically been one of the most sensitive to the El Nino effects since 1982. The drought even in 2003, for instance, caused a 25% reduction in the coffee production and earnings. The more recent drough in 2015 has also resulted in significant crop damage, yield decline and serious water shortage in this region (CCAFS 2016). In Indonesia, income increases mostly at temperature higher than 26 degrees Celsius mark. In Sumatra and Kalimantan and Suwalesi (other) we also observe a reduction of income between 24 and 26 and 22 and 24 degrees Celsius, respectively, but a sharp increase of income at higher temperature. Like in Vietnam, these regions have been very sensitive to the El Niùo–Southern Oscillation (ENSO). The recent drought event in 2016 in Sumatra18, while the increased incidence of cyclones and forest fires in Kalimantan and Suwalesi in the past decades, also associated with the ENSO, have severely affected the agricultural sector of these regions (UNDP 2013). Our results show that, indeed, farm income reduces significantly with changing weather conditions in both Vietnam and Indonesia (Figure 5). Non-farm income, instead, tends to increase with temperature consistently across both countries. Similarly, non-farm wage income is positively correlated with the increase of temperature, although in Vietnam we find a point of inflection when temperatures go beyond 28 degrees Celsius, where this income reduces on average. Our parametric estimations also show that weather conditions might worsen income inequality in certain areas of Vietnam and Indonesia (Figure 4, Panel B). More specifically, when average temperature increases beyond 26 degrees Celsius, the Gini coefficient in Vietnam tends to increase.This inequality enhancing effect is not observed for all the regions, but only for the the regions most vulnerable to El Nino events, such as Central Highlands, Southeast and Mekong Delta. In Indonesia, on the other hand, we find that climate variability is negatively correlated with the Gini coefficient. Thus, for temperature higher than 27 degrees Celsius, the Gini coefficient drops significantly in all regions but Sumatra, where the Gini coefficient actually increases.
18
https://coopcoffees.coop/the-impact-of-climate-change-in-sumatra/ 20
We investigated this relationship further across the main economic and demographic drivers of structural inequality to see whether the correlation between income, income inequality and weather variability differs across selected most vulnerable groups in Vietnam and Indonesia (Figures5,6,7). Figure 6 shows that, in Vietnam, the correlation between weather and household income changes significantly across different type of households. More specifically, household income decreases with contemporaneous temperature only in the case of rural, mostly agricultural households, female headed household and those households whose head is relatively young of age (<40 years old) or relatively old (>60 years). For other selected vulnerable groups, the correlation is, instead, positive. In Indonesia, on the other hand, we observe a positive correlation between household income and temperature across all the selected most vulnerable groups. Finally, the analysis of the correlation of weather and income inequality across vulnerable groups confirms that weather conditions might exacerbate existing inequalities (Figure 7). This is specifically in the case of Vietnam where for temperature higher than 26 degrees Celsius, income inequality increases for all the vulnerable groups considered in this analysis. We also find that changes in weather conditions might have an equalising effect in Indonesia at the higher end of the distribution but an inequality-enhancing effect at lower ends. Thus, the Gini coefficient reduces for temperature higher than 27 degrees Celsius, but it is stable or increases for temperature lower than 26 degrees. Figure 4: Non-parametric estimation of the correlation between average annual temperature, household income per capita (Panel A & C) and the provincial level Gini coefficient (Panel B & D) in Vietnam and Indonesia. Source: VHLSS 2002-2018 (Vietnam); IFLS 1993-2014 (Indonesia); CPC-NOAA ESRL (temperature).
21
Figure 5: Non-parametric estimation of the correlation between average weather conditions and farm, non-farm and wage income in Vietnam and Indonesia. Source: Authors' estimations using VHLSS 2002-2018 (Vietnam); IFLS 1993-2014 (Indonesia); CPC-NOAA ESRL (temperature).
22
Figure 6: Non-parametric estimation of the correlation between average weather conditions (average temperature per year in panel A, and total rainfall per year in panel B) and log of income per capita in Vietnam and Indonesia. Source: Authors' estimations using VHLSS 2002-2018 (Vietnam); IFLS 1993-2014 (Indonesia); CPC-NOAA ESRL (temperature); CHIRPS (rainfall).
Figure 7: Non-parametric estimation of the correlation between average weather conditions and Gini coefficient in Vietnam and Indonesia. Source: Authors' estimations using VHLSS 2002-2018 (Vietnam); IFLS 1993-2014 (Indonesia); CPC-NOAA ESRL (temperature)
23
IV – Accounting for non-linearity and heterogeneity of impacts The non-parametric analysis confirms that there exists a non-linear relationship between weather and income as different levels of temperature affect income and inequality in different ways. It also shows that poor, rural, agricultural, and most marginalis ed households might be bearing a bigger burden of changing weather conditions. However, these results only show the direct correlation of weather, income and inequality without considering other factors that could affect this relationship. We now present the results of the analysis which controls for, household and community level characteristics, heterogeneities of impacts and non-linearity between weather, income and inequality. 4.1. Income sensitivity to annual and seasonal weather variability Our results show that variability of contemporary temperatures has a positive effect on income in Vietnam. Namely, an increase of a single day’s temperature generates more total household income per capita in the same year (table A3). The positive effect is even more pronounced for agricultural, female headed households, and households with a young household head (Table A4). For these households, the largest increase in income occurs when temperature increase beyond the highest levels (>30 degrees Celsius). These results contradict findings of previous literature that argue that climate and weather variability should have a negative impact on income (e.g. Deryugina and Hsiang, 2017; Dell et al., 2012; and specifically for Vietnam, Narlock, 2016; Espagne and De Laubier-Longuet, 2019). There exists, however, a number of authors who challenge this claim, and argue that the observed negative association of warming climate conditions and income is, in fact, causal. These authors claim that omitted variables, such as social and political institutions, play a prominent role in mitigating or exacerbating the effect of climate and weather on income, which can in same case be negligible or positive (e.g. Gallup, Sachs, and Mellinger, 1999; Acemoglu, Johnson, and Robinson, 2002; Easterly and Levine, 2003; Nordhaus, 2006). These studies, though, mainly present evidence from the developed world where household level risk mitigating and insurance schemes, coupled with government driven defensive investments, can offset the overall negative impact of changing climate on prices and outputs in the short term (Deryugina and Hsiang, 2014). Vietnam is not a developed country, yet, and as such we would not expect that households would be able to fully offset the negative impact of climate variability on income as in wealthy economies. It is, however, plausible that changes in weather conditions might push households to leverage extra income in the short term in the form of transfers or remittances to be used as insurance for current and future climate hazards (Yang and Choi, 2007). These strategies might be especially used by those households who rely mostly on weather conditions for their livelihood, such as farming households, or the most exposed and vulnerable to climate hazards. In our sample, it is the most vulnerable groups (ethnic minorities, rural and female headed households, households with young, old, and
24
less educated heads) who receive, on average, a significantly higher amount of transfers than other households in the country, in the form of generic and disaster related allowances. To test whether receiving transfers might affect the impact of climate variability on income, we estimate our models with the addition of a variable that capture the receipt of allowances, including international and domestic remittances and other disaster related allowances. The results of this analysis show that transfers and remittances might be used as insurance by the most vulnerable and farming households (Table A5). For these households, the contemporaneous effect of climate variability on farm income is not significant. When controlling for transfers and remittances, this effect becomes negative for rural and ethnic minorities households, suggesting that, indeed, these households might be using transfers to cope with short term climatic shocks. Similar solutions might not play as well for the non-farm sectors, especially for poor, female and less educated households, who, instead, experience a significant reduction in this source of income with increase in climate variability in the same year and this impact does not change when we control for transfers. Coping capacity of Vietnamese households, however, seem to mostly focus on the short term. Our results show that the effect of changing of temperatures’ days in the previous year on income persists in the current year. This holds for the entire population, for agricultural households, as well as for the most vulnerable households, such as poor households and for ethnic minority who experience a negative impact of lagged temperature on current farm income. In Indonesia, our findings are mostly in line with the studies that postulate a negative, contemporaneous, effect of changing weather on income (such as Deryugina and Hsiang, 2001 and 2017; Dell et al., 2012). We find that the highest effect on total income is at the extreme ends of the temperature distributions, below 18 degrees Celsius and above 30 degrees Celsius (Table A6). Furthermore, agricultural households experience a significant reduction of total income with any change in temperature in the current year. Among different sources of livelihood, it is the farm income that reduces quite systematically for changes in contemporaneous weather conditions. Poor, low educated and those households who mostly rely on agriculture for their livelihood are the ones who are the most affected, especially for temperature lower than 18 degrees Celsius and higher than 24 degrees Celsius. We also find that non-farm income for rural and female headed households and wage income for household headed by older people reduce with changes in weather condition, which confirms the argument that climate impacts do not occur exclusively for agricultural activities (e.g. Dell et al., 2009; Hsiang, 2010) 19. To test whether seasonality affects the relationship between weather and income in our analysis, we also estimate the impact of seasonal climate variability on income, across type of income source and other households’ demographic characteristics. The results of this analysis confirm that the effect of climate variability is mostly positive in Vietnam, especially for agricultural households during the wet season (May to October) who
Also for Indonesia we estimate our models which control for transfers and remittances (Table A8). We find, however, that the results are consistent with the main model and therefore are not discussed here. 19
25
experience an increase in the total income and wage non-farm income (Table A 9). The only exception is for farm income in the dry season, which reduces with temperature variability when households are headed by old people (> 60 years old). For all the other incomes and demographic drivers of structural inequality, the contemporary results are either non-significant or positive (Table A10). Like in the previous estimates, we test whether the receipt of transfers affects households’ coping capacity differently during wet and dry season. We find that transfers and remittances play a mitigating role of the impact of climate variability on income especially in the dry season and especially for farming households, who may use these sources of income as insurance during periods of high climate uncertainty (Table A11). However, as in the analysis of the impact of annual climate variability, we find that these coping strategies are effective solely in the short term. Specifically, we find that during the dry season (November to April) the effect of the increase of the number of hot days (>30 degrees Celsius) in the previous year has a consistent negative effect on total household income, farm income and wage non-farm income, and that these effects are more pronounces for agricultural and poor households. During the wet season, instead, the effect of weather condition of the previous year reduces total income and wage non-farm income for rural, less educated households and those belonging to ethnic minorities (Table A10) 20. In Indonesia, we find that changes in temperature consistently reduce total and non-farm income for poor households during the dry and wet seasons (Table A12). Among household demographic characteristics, the age of the household head seems to exacerbate the impact of climate variability for households whose head is young (<40 years old) or old (>60 years old), with some differences across livelihood sources (Table A13). Some studies suggest that older people can adapt slower and less efficiently to climate and weather variations than other vulnerable groups (e.g. Maguza-Tempo, 2017; Mango et al., 2018). This lower overall coping capacity of older households is confirmed in our results, as climate variability significantly reduces the total income of these households. Disaggregating by livelihood sources, though, shows that older households might be able to cope or offset the negative impact of climate variability on farm and wage income in the short term. In Indonesia, older people are less likely to migrate to find better employment opportunities and more likely to receive transfers from young members of the households as migration, in Indonesia, is mostly young phenomenon (Sukamdi and Mujahid 2015). Like in Vietnam, it is plausible that these households might use remittances and other transfers as an insurance during periods of higher climate variability (Yang and Choi, 2007). In our sample, however, older people are not the only recipients of transfers and remittances. We find that middle aged widow, less educated, female headed households, with more dependents (children and adults in productive age) receive a significantly higher amount of nonlabour income than other households in the sample. They are more likely to be poor and live in urban areas in the Sumatra region and their sources of livelihood are primarily nonlabour and non-farm wage income. Opportunities to find better employment in more developed areas of the country tend to favour single, unmarried men (Sukamdi and Mujahid 2015). Young, low skilled, married households’ heads, with more dependents have also lower chances of improving their livelihood through migration. This is particularly true
20
These effects do not change when controlling for transfers. 26
in the case of women, who due to their traditional roles in the childcare might not be able to pursue their careers and may become recipients of remittances sent by a migrant husband or relative. We test whether the positive effect of climate variability is linked to the use of remittances and transfers as insurance and we find that our results remain mostly unchanged across different demographic groups, except for rural and female headed households (Table A14). For these households, farm income and non-farm income, respectively, reduce because of climate variability suggesting that also in Indonesia, like in Vietnam, coping strategies using transfers and remittances as insurance in periods of climate variability might be used21. Longer term impacts of climate variability are, however, not managed as well, especially for those households who are engaged in non-farm activities during the dry season and total income for agricultural households in the wet season, also suggesting that foresight in coping capacities and interventions is limited. 4.2. Income inequality sensitivity to annual and seasonal weather variability In this section we present the results of the analysis of the impact of annual and seasonal climate variability on the provincial level income Gini coefficient. Our results show that annual climate variability has a consistent inequality-increasing effect in Vietnam (Table A15). This effect is relatively more pronounced for those provinces where mostly agricultural households and ethnic minorities are located and especially at the highest end of the temperature distribution (>30 degrees Celsius). Similarly, in provinces where poor, rural, mostly agricultural, low educated households and those headed by a young head, income inequality also worsens for temperature changes below 18 degrees Celsius. Seasonal climate variability, on the other hand, has a mostly equalising effect across different groups in Vietnam (Table A16). This is exclusively observed for an increase in the number temperature days around the average levels (21-24 degrees Celsius) and during the wet season for all the groups selected for the analysis. For more remote provinces and those where households with a young head are located, instead, this equalising effect is observed for all changes in temperatures above 21 degrees Celsius22. In Indonesia, our results show inequality increases consistently for climate variability at the extreme ends of the distribution (temperature lower than 18 degrees Celsius and higher than 27 degrees Celsius). The inequality enhancing effect is larger in the case of provinces where poor households and households headed by older people are located. However, changes in temperature days around the average values (24-30 degrees Celsius) reduce provincial inequality (Table A19 & 20). Climate variability during the wet season also tends to decrease income inequality, especially for temperature above 30 degrees Celsius.
Remoteness also seems to affect the contemporaneous income response of the Indonesian households. Thus, for rural households, seasonal climate variability tends to increase their income in both the dry and wet seasons, but this effect is exclusively observed for non-farm wage activities, such as those related for example to services and other professional activities. 21
We test the role of transfers in the impact of annual and seasonal climate variability on income inequality and the results do not change (Table A17 & 18). 22
27
During the dry season, any change in temperature days reduces inequality exclusively for those provinces where mostly agricultural households are located, whereas in provinces with a lower level of education increases in temperature tend to also increase inequality in this season23.
Like for the analysis of Vietnam, these results do not change when controlling for transfers (Table A21 & 22) 23
28
V – Poverty mobility, drivers of structural inequality and climate variability in Vietnam The results presented thus far show that certain groups in the population do suffer more than others of the impact of climate variability. Despite some discrepancies across seasons, our results suggest that remoteness, type of economic activity (agriculture), poverty and ethnicity seem to exacerbate the negative effect of climate variability. Vietnam is one of the fastest growing economies in SEA and poverty reduction achievements have been extraordinary, which suggests that social mobility in the past years has been particularly high, allowing people to escape poverty through time. Nonetheless, inequality has not been reducing as steadily as poverty rates, while weather abnormalities have increased in the past decades. In this section, we use Vietnam as a case study to assess whether the relationship between inequality and climate variability affects the ability of households to escape poverty. We do so by investigating whether there is correlation between poverty mobility, climate variability and observable characteristics linked to structural inequality. For this analysis, we use a synthetic panel approach. Synthetic panels have widely been recognized as robust alternative to panel data to study poverty mobility (Dang et al., 2019). Synthetic panel analysis allows to reconstruct household panel data using pure cross-sectional data and controlling for house time-invariant characteristics24. The main advantage of this approach is to provide similar benefits of panel data when these do not exist, as in the case of Vietnam. We follow Dang and Lanjouw (2013) approach to study the relationship between mobility and inequality for climate change impacts25. Figure 8: Probability to escape and to enter poverty in Vietnam across demographic and economic groups, synthetic panel analysis of Vietnamese households 2002-2016. Source: Authors estimation using VHLSS data from 2002 to 2016.
Probability in 2014-2016
Probability in 2002-2004
Panel A: To escape poverty
50 an rb U
0
Ur ba n
al ur R
Ru ra l
ty ty rs re ci ci to ltu ni ni ec th th icu rs re re gr e o o a j h a in in ot M M pl in pl Em Em
M ot al e he rs Em ec pl to in rs ag ric ul tu M re aj or et hn ici M ty in or et hn ici ty
e al M
Fe m al e
e an al rb m U Fe
Em pl in
al ur R
Ru ra l Ur ba n
ty ty rs re ci ci to ltu ni ni ec th th icu rs re re gr e o o a j h a in in ot M M pl in pl Em Em
M ot al he e rs Em ec pl t o in rs ag ric ul tu M re aj or et hn M i c in ity or et hn ici ty
e al M
30 Fe m al e
e al m Fe
Em pl in
0
20
2
30
5
40
4
Percentage (%)
10
6
50 40
Percentage (%)
Panel B: To enter poverty 8
60
Panel B: To enter poverty 15
60
Panel A: To escape poverty
The main assumption of this approach is that that the underlying population being sampled in survey rounds 1 and 2 are identical such that their time-invariant characteristics remain the same over time. 25 A more comprehensive presentation of the synthetic panel analysis can be found in Appendix 2. 24
29
Figure 8 shows the results of the correlation between upward and downward poverty mobility, respectively, and structural inequalities for Vietnam. The data confirms our previous finding in that structural inequality’s drivers, such as type of economic activities, remoteness and ethnicity are significantly associated with a lower probability to escape poverty and increase poverty immobility. Thus, rural, agricultural, and ethnic minority households have a much lower probability to move out of poverty and a much higher likelihood to become poor across all the rounds of VHLSS. Interestingly, female headed households seem to have higher chances of upward poverty mobility than male households, although the difference is not likely to be significant. Figure 9 shows the correlation between upward poverty mobility of the past 12 and 24 months, respectively, and average annual temperature and rainfall patterns. As in our previous analysis, contemporaneous correlation between upward poverty mobility and temperature is non-linear and positive. A negative relationship is, instead, found between rainfall and the probability to escape poverty. These results are consistent for current and lagged value of weather variables (Figure 28). Figure 9: Correlation between contemporaneous (left panel) and lagged (12 months before – right panel) average weather and (conditional) probability (at household level) a poor household in the first period becomes non-poor in the second period. Source: Authors estimation using VHLSS data from 2002 to 2016
Finally, figure 10 show the interaction between climate variability and demographic characteristics linked to structural inequalities that are correlated with the probability of becoming poor. The overall findings confirm puzzling results for Vietnam, especially for temperature variability. The analysis shows that there exists a negative relationship between increase in temperature and the probability to enter poverty for female headed households, rural, agricultural, and ethnic minority households. For the latter type of households, the figure also shows that between 24- and 26-degrees downward poverty mobility is, in fact, positively associated with temperature increase. On the other hand, we find that higher rainfalls (above 1000 mm) do increase the probability to enter poverty consistently across the different groups (female headed households, rural, agricultural and ethnic minority households).
30
Overall, these results suggest that there exists a heterogeneity of poverty mobility across the sample, and that poor, rural, agricultural and ethnic minority households are more likely to become poor or escape poverty across time. Figure 10: Correlation between weather of the past 12 months and (conditional) probability (at household level) a non-poor household in the first period becomes poor in the second period, by gender of household head (top left), ethnicity (top right), economic activity (bottom left) and remoteness (bottom right) . Source: Authors’ estimation using VHLSS data from 2002 to 2016.
31
Conclusions The past few decades have seen an extraordinary increase of variability in climate conditions, especially in tropical countries, where each year an increased number of typhoons, floods, droughts are disrupting the life of entire populations. The ability of people to respond to these new hazards may diverge signficantly due to the existence of socioeconomic and structural inequalities that might further constrain opportunities to improve their wellbeing. The impact of climate variability might enhance these existing inequalities and force “the last mile” groups into a spiral of further poverty and social exclusion. In this paper we investigated the impact of the so-called “environment-inequality nexus” in two of the fastest growing economies in South-East Asia, Vietnam and Indonesia. In these countries both climate variability and inequality have been increasing (Le and Booth, 2013; Nguyen et al., 2015; Bui et al., 2017; Nguyen et al., 2017; Nguyen and Nguyen, 2017; Asra, 2000; Akita, 2002; Oxfam, 201726; Ananda and Pulungan, 2019). We directly analyse the effect of annual and seasonal temperature on income and income inequality across years. We do so by looking at the Vietnamese and Indonesian populations as a whole and also investigating more in-depth how these impacts change for the most vulnerable and marginalised groups. Our results show that the effect of climate variability is regressive as income decreases and income inequality increases with changing in climatic conditions in both Vietnam and Indonesia. Our results show that the “last mile” groups, poor, rural and agricultural households and minorities, bear the biggest burden of climate variability, forcing them in a spiral of poverty and social exclusion. Among different sources of income, our findings show that farm income is negatively correlated with increase in temperature especially in Indonesia, especially during dry seasons, which is in line with previous findings (e.g. Fisher et al, 2002; Hallegatte et al, 2014; First, 2019; Farbotko, 2020). Our analysis also shows that in both Vietnam and Indonesia non-farm economic sectors suffer the increase of climate instability (as in Dell et al., 2009; Hsiang, 2010; Graff Zivin and Neidell, 2014; Seppannen et al., 2013). The most affected areas are those that have been historically more vulnerable to the El Nino events, such as Sumatra, Kalimantan and Sulawesi in Indonesia and Southeast, Central Highlands and Mekang Delta in Vietnam. Country specific differences can be noted. For instance, in Vietnam we find that changes in weather conditions has a positive effect on the income of the same year. Some authors claim that climate uncertainties might push households to leverage extra income in the short term in the form of transfers or remittances which are then used as insurance for current and future climate hazards (Yang and Choi, 2007). These strategies might be especially used by those households who rely mostly on weather conditions for their livelihood, such as farming households, or the most exposed and vulnerable to climate hazards. Our analysis confirms this claim and it shows that rural and ethnic minority and
26
https://www.oxfam.org/en/indonesia-even-it/inequality-indonesia-millions-kept-poverty. 32
farming households might be using transfers to cope with short term climatic shocks. Coping capacity of Vietnamese households, however, seem to mostly focus on the short term as our results show that the effect of changing of temperatures’ days in the previous year on income persists in the current year.In Indonesia, on the other hand, we find that these coping strategies might apply to poor, younger and older, female headed and low educated households, who are the biggest receipients of transfers and remittances. Migration, in Indonesia, is mostly a young phenomenon, which tends to favour single, unmarried men (Sukamdi and Mujahid 2015). Like for Vietnam, we find that, controlling for the receipt of transfers and remittances, farm income and non-farm income, respectively, reduce because of climate variability, suggesting that this coping capacity might be effective in the short term. Longer term impacts of climate variability are, however, not managed as well, also suggesting that foresight in coping capacities and interventions is limited. Finally, our analysis confirms the highest exposure and vulnerability of ethnic minorities to changes in weather conditions in Vietnam. The marginalisation and vulnerability of ethnic minority has been extensively documented (e.g. Bruun, 2012; McElwee, 2015; Son and Kingsbury, 2020) and our study confirms the need to design ad-hoc solutions for these groups in light of increasing climate impacts. In the past decade, policies in both Vietnam and Indonesia were designed to target and support low-income groups in coping with emergencies, they often had little relevance to the needs, rights and priorities of the poorest people. The lack of participation and voice of the most marginalised in the places of power has in the recent years worked in favour of the better-off and widened the gap between those and the poorest, reducing even further their ability to face climatic challenges (Oxfam, 2017; Oxfam in Vietnam, 2015; Nguyen Tran Lam et al., 2016; Muhtadi and Warburton, 2020). Developing specific measures to help rural and agricultural households coping, adapting and mitigating the annual and season impacts of the increasing climate instability will be paramount in these countries. Specific attention should also be given to those groups, such as ethnic minorities and younger or older households, who have found to bear the biggest burden of climate variability.
33
References Acemoglu, D., Johnson, S., and Robinson, J. A. (2001). The Colonial Origins of Comparative Development: An Empirical Investigation. The American Economic Review, 91(5):1369–1401. Acemoglu, D., Johnson, S., and Robinson, J. A. (2001). Reversal Of Fortune: Geography And Institutions In The Making Of The Modern World Income Distribution. The Quarterly Journal of Economics 117 (4):1231–1294.
Barro, R. J. (2000). “Inequality and Growth in a Panel of Countries.” Journal of Economic Growth. 5 (1): 5–32. Beck, U., (2010). Remapping social inequalities in an age of climate change: for a cosmopolitan renewal of sociology. Global Networks, 10(2), pp.165-181. Booth, A., (2019). Measuring poverty and income distribution in Southeast Asia. Asian‐Pacific Economic Literature, 33(1), pp.3-20.
Akita, T., (2002). Regional income inequality in Indonesia and the initial impact of the economic crisis. Bulletin of Indonesian Economic Studies, 38(2), pp.201-222.
Bresciani, F. and Valdés, A. eds., (2007). Beyond food production: the role of agriculture in poverty reduction. Food & Agriculture Org.
Ananda, C.F. and Pulungan, A.M., (2019). Determinants of Income Inequality in Indonesia: Case Study of 33 Provinces in 2011-2016. Journal of Applied Business and Economics, 21(6).
Bruun, O., (2012). Sending the right bill to the right people: Climate change, environmental degradation, and social vulnerabilities in Central Vietnam. Weather, Climate, and Society, 4(4), pp.250-262.
Andrews, D. and Leigh, A., (2009). More inequality, less social mobility. Applied economics letters, 16(15), pp.1489-1492.
Bui, A. T., Nguyen, C. V., & Pham, T. P. (2017). Poverty among ethnic minorities: the transition process, inequality and economic growth. Applied Economics, 49(31), 3114-3128.
Asra, A., (2000). Poverty and inequality in Indonesia: estimates, decomposition and key issues. Journal of the Asia Pacific Economy, 5(1-2), pp.91-111. Auffhammer, M., Hsiang, S., Schlenker, W., and Sobel, A. (2013). Using weather data and climate model output in economic analyses of climate change. Review of Environmental Economic Policy, 7:181–198.
Burke, M., Hsiang, S. M., and Miguel, E. (2015). Global non-linear effect of temperature on economic production. Nature, 527(7577):235–239. Cameron, L., Suarez, D. C., & Rowell, W. (2019). Female Labour Force Participation in Indonesia: Why Has it Stalled?. Bulletin of Indonesian Economic Studies, 55(2), 157-192.
34
CGIAR Research Program on Climate Change, Agriculture and Food Security- Southeast Asia (CCAFS-SEA) (2016). Assessment Report: The drought crisis in the Central Highlands of Vietnam. Hanoi, Vietnam. Climate Service Center (2015). Climate Fact Sheet – Indonesia, updated version 2015. Accessed on March 2020: http://www.climate-servicecenter.de/products_and_ publications/fact_sheets/clima te_fact_sheets/index.php.en Dang, H. A., & Lanjouw, P. (2013). Measuring poverty dynamics with synthetic panels based on cross-sections. The World Bank. Dang, Hai-Anh, Dean Jolliffe, and Calogero Carletto. (2019). "Data Gaps, Data Incomparability, and Data Imputation: A Review of Poverty Measurement Methods for Data-Scarce Environments". Journal of Economic Surveys, 33(3): 757-797. Dang, H.A.H., Lokshin, M.M., Abanokova, K. and Bussolo, M., (2020). Welfare Dynamics and Inequality in the Russian Federation During 1994-2015 The European Journal of Development Research, pp.1-35 Dani, A.A. and de Haan, A. eds., (2008). Inclusive states: social policy and structural inequalities. The World Bank. Deschênes, O. and Greenstone, M., (2011). Climate change, mortality, and adaptation: Evidence from annual fluctuations in weather in the US. American Economic Journal: Applied Economics, 3(4), pp.152-85.
Dell, M., Jones, B. F., and Olken, B. A. (2009). Temperature and Income: reconciling New CrossSectional and Panel Estimates. American Economic Review, 99(2):198–204. Dell, M., Jones, B. F., and Olken, B. A. (2012). Temperature shocks and economic growth: Evidence from the last half century. American Economic Journal: Macroeconomics, 4(3):66–95. Deryugina, T. and Hsiang, S. M. (2017). The marginal product of climate. NBER Working Paper, No. 24072 Dabla-Norris, M.E., Kochhar, M.K., Suphaphiphat, M.N., Ricka, M.F. and Tsounta, E., (2015). Causes and consequences of income inequality: A global perspective. International Monetary Fund. Espagne, E. and DE LAUBIERLONGUET, N., (2019). Non-linear Impacts of Climate Change on Income and Inequality in Vietnam (No. e47825f2-00594010-8291-189c04770a51). Easterly, W. and Levine, R., (2003). Tropics, germs, and crops: how endowments influence economic development. Journal of monetary economics, 50(1), pp.3-39. Eckstein, D., Hutfils, M., & Winges, M. (2019). Global Climate Risk Index 2019: Who suffers most from extreme weather events? Weatherrelated loss evens in 2017 and 1998 to 2017. German Watch. FAO, (2016). ‘El Nino’ Event in Vietnam: Agriculture Food Security and Livelihood Needs Assessment in Response to Drought and Salt Water Intrusion.
Farbotko, C., (2020). Is it too late to prevent systemic danger to the world's poor?. Wiley Interdisciplinary Reviews: Climate Change, 11(1), p.e609. First, P.J., (2019). Global Warming of 1.5 C An IPCC Special Report on the Impacts of Global Warming of 1.5 C Above Pre-Industrial Levels and Related Global Greenhouse Gas Emission Pathways, in the Context of Strengthening the Global Response to the Threat of Climate Change. Sustainable Development, and Efforts to Eradicate Poverty. https://www. ipcc. ch/sr15/. Accessed, 1. Fischer, G., Shah, M.M. and Van Velthuizen, H.T., (2002). Climate change and agricultural vulnerability. Gallup, J.L., Sachs, J.D. and Mellinger, A.D., (1999). Geography and economic development. International regional science review, 22(2), pp.179-232. Graff Zivin, J. and Neidell, M. (2014). Temperature and the allocation of time: Implications for climate change. Journal of Labor Economics, 32(1):1–26. Hallegatte, S., Bangalore, M., Bonzanigo, L., Fay, M., Narloch, U., Rozenberg, J. and VogtSchilb, A., (2014). Climate change and poverty—an analytical framework. The World Bank. Hsiang, S. M. (2010). Temperatures and cyclones strongly associated with economic production in the caribbean and central america. Proceedings of the National Academy of sciences, 107(35):15367–15372.
35
Karmalkar, C., M. N. McSweeney, and G. Lizcano. "UNDP Climate Change Country Profiles Indonesia." (2016). Accessed on May 2020: http://www.geog.ox.ac.uk/resea rch/climate/projects/undpcp/UNDP_reports/Indonesia/ Kerry, B., Pickett, K.E. and Wilkinson, R., (2012). The Spirit Level: Why Greater Equality Makes Societies Stronger. CHILD POVERTY AND INEQUALITY NEW PERSPECTIVES, p.205. Kompas, T., Breusch, T., & Ward, M. B. (2017). Language, mixed communes, and infrastructure: Sources of inequality and ethnic minorities in Vietnam. World Development, 96, 145-162. Lazear, E., and S. Rosen (1981). “Rank-Order Tournaments as Optimum Labor Contracts.” Journal of Political Economy 89 (5): 841–64. Le, H. and Booth, A. (2013). Inequality in Vietnamese Urban-Rural Living Standards, 1993-2006. Review of Income and Wealth, 60(4), 862-886. Leichenko, R. and Silva, J.A., (2014). Climate change and poverty: vulnerability, impacts, and alleviation strategies. Wiley Interdisciplinary Reviews: Climate Change, 5(4), pp.539-556. Leitmann, J., Brown, T., Roesad, K., Bojo, J., & DeRidder, K. (2009). Investing in a more sustainable Indonesia: Country environmental analysis. CEA Series, East Asia and Pacific Region–Washington, DC: World Bank.–114 p.
Maguza-Tembo, F., Mangison, J., Edris, A.K. and Kenamu, E., (2017). Determinants of adoption of multiple climate change adaptation strategies in Southern Malawi: An ordered probit analysis. Journal of Development and Agricultural Economics, 9(1), pp.1-7. Mango, N., Makate, C., Tamene, L., Mponela, P. and Ndengu, G., (2018). Adoption of small-scale irrigation farming as a climatesmart agriculture practice and its influence on household income in the Chinyanja Triangle, Southern Africa. Land, 7(2), p.49. McCaig, B., Benjamin, D., & Brandt, L. (2015). Growth with equity: Income inequality in Vietnam, 2002–12. Retrieved from https://drive.google.com/fi le/d/0B5Kjg1b9s7JRZk95SmZzcm JJLWs/view McElwee, P., (2015). From conservation and development to climate change: anthropological engagements with REDD+ in Vietnam. Climate Cultures: Anthropological Perspectives on Climate Change, pp.82-106. MDRI & OXFAM, (2020), Multidimensional inequality in Vietnam. Research Report. Mendelsohn, Robert, William D. Nordhaus, and Daigee Shaw. (1994). The Impact of Global Warming on Agriculture: A Ricardian Analysis." The American Economic Review :753{771.
Narloch, U., (2016). The varying income effects of weather variation: Initial insights from rural Vietnam. The World Bank. Nguyen, C., & Nguyen, L. (2017). Intra-generational and intergenerational mobility in Vietnam. MPRA Paper 80083, University Library of Munich, Germany. Nguyen-Hoa, H. Kompas, T., Breusch, T., & Ward, M. B. (2017). Language, mixed communes, and infrastructure: Sources of inequality and ethnic minorities in Vietnam. World Development, 96, 145-162. Nguyen, V.C, Phung, D.T and Westbrook, D. (2015). Do the poorest ethnic minorities benefit from a large-scale poverty reduction program? Evidence from Vietnam. The Quarterly Review of Economics and Finance, vol. 56, pp.3-14. Nguyen Tran Lam et al. (2016). Migration and Health - The Case Study among Migrants in Vietnam. Paper presented at the Fourth 21st Century Academic Forum Conference at Harvard University, USA. 20-22 March 2016. Nordhaus, W.D., (2006). Geography and macroeconomics: New data and new findings. Proceedings of the National Academy of Sciences, 103(10), pp.3510-3517.
Oxfam in Vietnam (2015). Understanding Farmers’ Cooperation and Linkages in Agricultural Production to Promote Farmers’ Rights, Voices and Choices. Findings and Policy Recommendations. Accessed at https://cngcdn.oxfam.org/vietnam.oxfam.o rg/s3fspublic/file_attachments/Oxfam _Bao-cao-GROW-ve-Lien-ketnong-dan_Eng.pdf Pimhidzai, O., (2018). Climbing the ladder: poverty reduction and shared prosperity in Vietnam. Washington, DC: World Bank Group. Seppanen, O., J., F. W., and David, F. (2003). Control of temperature for health and productivity in offices. Lawrence Berkeley National Laboratory. Schlenker, W. and Roberts, M. J. (2009). Nonlinear temperature effects indicate severe damages to US crop yields under climate change. Proceedings of the National Academy of sciences, 106(37):15594–15598. Skoufias, E., Vinha, K., and Conroy, H. (2013). The impacts of climate variability on welfare in rural Mexico. Population and Environment, 34(3). Stiglitz, J. (2012). The Price of Inequality: How Today's Divided Society Endangers Our Future. New York: W.W. Norton. Son, H. and Kingsbury, A., (2020). Community adaptation and climate change in the Northern Mountainous Region of Vietnam: A case study of ethnic minority people in Bac Kan Province. Asian Geographer, 37(1), pp.33-51.
Muhtadi, B. and Warburton, E., (2020). Inequality and democratic support in Indonesia. Pacific Affairs, 93(1), pp.31-58.
36
Sukamdi and Mujahid, G. (2015). UNFPA Indonesia Monograph Series: No.3, Internal Migration in Indonesia. UNFPA Indonesia Monograph Series. [online] Jakarta: UNFPA. Available at: http://indonesia.unfpa.org/en/ publications/monographseries-no-3-internalmigration-indonesia Suryahadi, A. and Hadiwidjaja, G., (2011), May. The role of agriculture in poverty reduction in Indonesia. In Centre for Strategic Economic Studies– Australian Centre for International Agricultural Research International Workshop on the Role of Agriculture in Poverty Reduction, Melbourne. Taniguchi, K., & Tuwo, A. (2014). New Evidence on the Gender Wage Gap in Indonesia. Asian Development Bank Economics Working Paper Series (408).
Tuyen, T.Q. (2016). Income sources and inequality among ethnic minorities in the Northwest region, Vietnam. Environment, Development and Sustainability 18, 1239–1254 UN (2018). Inequality in Asia and the Pacific in the Era of the 2030 Agenda for Sustainable Development. UN, New York, https://doi.org/10.18356/1f52de a6-en Welch, J.R., Vincent, J.R., Auffhammer, M., Moya, P.F., Dobermann, A., Dawe, D., (2010). Rice yields in tropical/subtropical Asia exhibit large but opposing sensitivities to minimum and maximum temperatures. Proc. Natl. Acad. Sci. 107, 14562–14567. World Bank (2014). Indonesia Risk Profile. Accessed on May 2020: http://sdwebx.worldbank.org/cli mateportalb/home.cfm?page= country_profile&CCode=IDN
37
Wilkinson, R.G. and Pickett, K.E., (2009). Income inequality and social dysfunction. Annual review of sociology, 35, pp.493-511. Yang, D., Choi, H., (2007). Are remittances insurance? Evidence from rainfall shocks in the Philippines. World Bank Economic Review. 21, 219–248. Weni Lidya, Sukma & Kadir, Kadir, (2019). "Decomposition of the Gender Wage Gap in Indonesia: Analysis from Sakernas Data," MPRA Paper 94930, University Library of Munich, Germany. World Bank (2014): Indonesia Risk Profile. http://sdwebx.worldbank.org/ climateportalb/home.cfm?pag e=country_profile&CCode=IDN 2 https://en.actualitix.com/countr y/wld/co2-emissions-percapita.
Appendix - 1 Sources of data The socio-economic data used in this analysis were drawn from the Indonesian Family Life Survey (IFLS) and the Vietnam Household Living Standard Survey (VHLSS). The IFLS (RAND) is a publicly available longitudinal survey that is representative of about 83% of the Indonesia population at regional level, with a total sample of 30,000 people located in 13 out of the 27 provinces in the country. It is one of the richest datasets available in Indonesia and it has been collected since 1993, covering information at individual, household and community level for several topics. Currently, five rounds have been collected: 1993/94, 1997, 2000, 2007, 2014/15. For our main analysis, we selected only those households that have been interviewed in all the 5 rounds. Therefore, our final sample includes a balanced panel of 4909 households (HHs) across each IFLS round, for a total of 24,545 HHs. The Vietnam Household Living Standard Survey (VHLSS) is a provincial, regional and national representative survey, collected every two years since the beginning of the 1990s. The latest round dates to 2018. To ensure comparability and consistency across surveys we selected a subsample of datasets for this analysis, starting from 2002. This is because questionnaires and sampling design were substantially different before the 2002 round and dropping previous rounds ensures that we are not adding unnecessary biases in the analysis. Our final sample is composed of a total of 99,723 households/repeated crosssection observations, across 9 VHLSS rounds. Climate data was retrieved using different sources. Climate Hazards Group InfraRed Precipitation with Station (CHIRPS) data are an excellent source of rainfall frequency data, which are collected at a very high resolution (5 km) from 1981 up to now, but less so for temperature, for which data were collected for a limited amount of time (from 2010 onwards). For temperature data, we, therefore, opted for Climate Prediction Center (CPC) of the NOAA ESRL Physical Sciences Division (PSD), which provides temperature frequency data since 1979 to 201827. We used minimum and maximum daily temperature and amount of daily rain in mm. Our final sample of climate data included on average 3,285 observations for each of the commune (Vietnam) and sub-district (Indonesia).
The CPC datasets is slightly coarser than the CHIRPS data. Its resolution is 50 km. To extract commune (in the case of Vietnam) and sub-district (in the case of Indonesia) level temperature data, we have developed a downscaling process by mean of elevation data at 5 km. 27
38
Selected summary statistics
Table A1: Summary statistics of household level explanatory variables - Vietnam, VHLSS 2002 - 2018 Variable
Description
2002
2004
2006
2008
2010
2012
2014
2016
2018
Sex head
1=male; 0=female
0.79
0.78
0.78
0.78
0.78
0.78
0.77
0.78
0.78
Age
Age of head of household in years
48
49
49
50
48
49
51
51
51
Ethnic minority
1= if household is member of an ethnic minority; 0 otherwise
0.12
0.13
0.13
0.13
0.14
0.15
0.14
0.16
0.16
Children
Proportion of < 15 years old, % of total household members
29
27
24
23
24
23
23
24
24
Female members
Proportion of female in the household, % of total household members
50
50
51
51
51
51
51
51
51
HH size
Number of household members
5.09
5.01
4.86
4.76
4.48
4.47
4.44
4.44
4.39
Urban
1=urban; 0=rural
0.23
0.25
0.26
0.27
0.30
0.30
0.34
0.32
0.31
HH head with education or lower than primary education, %
32
30
28
26
26
25
24
24
22
HH head with primary education, %
24
24
24
25
25
25
25
24
24
Lowersecondary
HH head with lower-secondary education, %
26
24
26
26
24
25
26
26
26
Uppersecondary
HH head with upper secondary education, %
8
7
7
8
7
8
9
9
9
Technical degree
HH head with technical degree, %
6
10
10
12
11
11
10
9
10
Postsecondary
HH head with post-secondary education, %
4
4
5
5
7
6
7
8
8
8532
8503
9354
9354
9342
9320
9077
Education of household head < Primary
Primary
N
27720
8521
39
Table A2: Summary statistics of household level explanatory variables - Indonesia, IFLS 1993-2014
Variable
Description
1993
1997
2000
2007
2014
Sex head
1=male; 0=female
0.86
0.83
0.84
0.81
0.79
Age
45
47
49
52
55
Children
Age of head of household in years Proportion of < 15 years old, % of total household members
31
30
28
21
18
Female members
Proportion of female in the household, % of total household members
52
53
52
53
54
HH size
Number of household members
4
4
4
4
4
Urban
1=urban; 0=rural
0
0
0
0
0
Proportion of households’ heads with no education or lower than primary education, %
23
22
17
15
13
Primary
Proportion of households’ heads with primary education, %
54
54
53
52
48
Lower- and upper-secondary
Proportion of households’ heads with lower-secondary education, %
23
24
25
27
31
Post-secondary
Proportion of households’ heads with post-secondary education, %
0.0
0.0
4.8
6.2
7.3
4909
4909
4909
4909
4909
Education of household head < Primary
N
40
Figure A1: Distribution of income across different livelihood sources and quintiles in Vietnamese regions. Source: Authors estimations using VHLSS (2002-2018)
41
Figure A2: Gini coefficient in Vietnamese regions. Source: Authors estimations using VHLSS (2002-2018)
42
Figure A3: Distribution of income across different sources of livelihoods and quintiles in regions in Indonesia. Source: Authors estimations using IFLS (1003-2014)
43
Figure A3: Gini coefficient in Indonesian regions. Source: Authors estimations using IFLS (1993-2014)
44
Figure A4: Distribution of days per year in temperature (panel A) and rainfall bins (panel B) between 2002 2018 across Vietnamese communes. Source: Authors' estimation using CPC-NOAA ESRL (temperature) and CHIRPS (rainfall) data
45
Figure A5: Distribution of days per year in temperature (panel A) and rainfall bins (panel B) between 1993 and 2014 across Indonesian sub-districts. Source: Authors' estimation using CPC-NOAA ESRL (temperature) and CHIRPS (rainfall) data
46
Table A3: Estimated effect of climate variability on household income per capita (total, farm, non-farm and wage income in Vietnam. Selected results. (VHLSS 2002-2018).
47
Table A4: Estimated effect of climate variability on household income per capita of selected most marginalised groups in the population in Vietnam. Selected results. (VHLSS 2002-2018).
48
Table A5: Summary table of the changes in the coefficient of interest when controlling for transfers received by the households in Vietnam. Source: Authors' calculation using VHLSS data (2002-2018) 28
The CPC datasets is slightly coarser than the CHIRPS data. Its resolution is 50 km. To extract commune (in the case of Vietnam) and sub-district (in the case of Indonesia) level temperature data, we have developed a downscaling process by mean of elevation data at 5 km. 28
49
Table A6: Estimated effect of climate variability on household income per capita (total, farm, non-farm and wage income) in Indonesia. Selected results. (IFLS 1993-2014)
50
Table A7: Estimated effect of climate variability on household income per capita of selected most marginalised groupsin the population in Indonesia. Selected results. (IFLS 1993-2014).
51
Table A3: Summary table of the changes in the coefficient of interest when controlling for transfers (dummy variable – Panel A, and log real value in Panel B) received by the households in Indonesia. Source: Authors' calculation using IFLS data (1993-2014)
52
Table A9: Estimated effect of seasonal climate variability on household income per capita (total, farm, non-farm and wage income) in Vietnam. Selected results. (VHLSS 2002-2018).
53
Table A10: Estimated effect of seasonal climate variability on household income per capita of selected most marginalised groups in the population in Vietnam. Selected results. (VHLSS 2002-2018).
54
55
Table A11: Summary table of the changes in the coefficient of interest when controlling for transfers (dummy in Panel A and log of real value in Panel B) received by the households in Vietnam. Source: Authors' calculation using VHLSS data (2002-2018)
56
Table A12: Estimated effect of seasonal climate variability on household income per capita (total, farm, non-farm and wage income) in Indonesia. Selected results. (IFLS 1993-2014).
57
Table A13: Estimated effect of seasonal climate variability on household income per capita of selected most marginalised groups in the population in Indonesia. Selected results. (IFLS 1993-2014).
58
Table A14: Summary table of the changes in the coefficient of interest when controlling for transfers (dummy in Panel A and log of real value in Panel B) received by the households in Indonesia. Source: Authors' calculation using IFLS data (1993-2014)
59
Table A15: Estimated effect of climate variability on the income Gini coefficient estimated at provincial level in Vietnam
60
Table A16: Estimated effect of climate variability on the income Gini coefficient estimated at provincial level in Vietnam, seasonal variables
61
Table A17: Summary table of the changes in the coefficient of interest when controlling for transfers (dummy in Panel A and log of real value in Panel B) received by the households in Vietnam. Source: Authors' calculation using VHLSS data (2002-2018) Panel A: Transfer dummy Annual
Gini coefficient
Transfer variable
All Mostly agricultural Rural Young head Old head Low education Female head Ethnic minority Poor
no difference no difference no difference no difference no difference no difference no difference no difference no difference
not significant not significant not significant not significant not significant not significant not significant not significant not significant
Panel B: Log real value of transfers received Annual
Gini coefficient
Transfer variable
All Mostly agricultural Rural Young head Old head Low education Female head Ethnic minority Poor
no difference no difference no difference no difference no difference no difference no difference no difference no difference
not significant not significant not significant not significant not significant not significant not significant not significant not significant
Table A18: Summary table of the changes in the coefficient of interest when controlling for transfers (dummy in Panel A and log of real value in Panel B) received by the households in Vietnam. Source: Authors' calculation using VHLSS data (2002-2018) Panel A: Transfer dummy Seasonal
Gini coefficient Wet
Dry
Transfer variable
All
no difference
no difference
not significant
Mostly agricultural Rural
no difference no difference
no difference no difference
not significant not significant
Young head Old head
no difference no difference
no difference no difference
not significant not significant
Low education Female head
no difference no difference
no difference no difference
not significant not significant
Ethnic minority Poor
no difference no difference
no difference no difference
not significant not significant
Panel B: Log real value of transfers received Seasonal
Gini coefficient Wet
Dry
Transfer variable
All
no difference
no difference
not significant
Mostly agricultural Rural
no difference no difference
no difference no difference
not significant not significant
Young head Old head
no difference no difference
no difference no difference
not significant not significant
Low education Female head
no difference no difference
no difference no difference
not significant not significant
Ethnic minority Poor
no difference no difference
no difference no difference
not significant not significant
62
Table A19: Estimated effect of climate variability on the income Gini coefficient estimated at provincial level in Indonesia
All
Poor
Rural
Agricultura l HH
Female HH head
(1)
(2)
(3)
(4)
(5)
Gini
Gini
Gini
Gini
Gini
coefficient
coefficient
coefficient
coefficient
0.0989***
0.1159***
0.1044***
(0.009)
(0.016)
(0.012)
-0.0000
0.0002
(0.000)
Head
Young HH head
Old HH head
(7)
(8)
(9)
Gini
Gini
Gini
coefficient
coefficient
coefficient
coefficient
0.0851***
0.0765***
0.0775***
0.0543***
0.0807***
(0.012)
(0.008)
(0.011)
(0.015)
(0.009)
-0.0001
-0.0002
0.0001
-0.0008**
0.0002
-0.0001
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
-0.0005***
-0.0005*
-0.0006**
-0.0008**
-0.0005**
-0.0009***
-0.0006**
-0.0005*
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
tembin27_30_lag
-0.0007***
-0.0008***
-0.0008***
-0.0009***
-0.0008***
-0.0010***
-0.0008***
-0.0007**
tembin30_40_lag
(0.000) 0.0258***
(0.000) 0.0227***
(0.000) 0.0195**
(0.000) 0.0186*
(0.000) 0.0149
(0.000) 0.0167
(0.000) 0.0195***
(0.000) 0.0201*
o.tembin15_18_lag2
(0.007) -
(0.009) -
(0.010) -
(0.010) -
(0.010) -
(0.011) -
(0.008) -
(0.011) -
tembin21_24_lag2
-0.0056
-0.0042
-0.0108
-0.0103
-0.0065
-0.0157**
-0.0053
-0.0062
(0.004)
(0.006)
(0.011)
(0.008)
(0.005)
(0.008)
(0.005)
(0.009)
tembin24_27_lag2
-0.0053
-0.0040
-0.0105
-0.0099
-0.0060
-0.0154**
-0.0050
-0.0060
tembin27_30_lag2
(0.004) -0.0052
(0.006) -0.0038
(0.010) -0.0104
(0.008) -0.0099
(0.005) -0.0059
(0.008) -0.0154**
(0.005) -0.0048
(0.009) -0.0059
tembin30_40_lag2
(0.004) -0.0146***
(0.006) -0.0077
(0.010) -0.0281**
(0.008) -0.0249***
(0.005) -0.0151**
(0.008) -0.0270***
(0.005) -0.0166**
(0.009) -0.0148
Indonesia
tembin15_18_lag tembin21_24_lag tembin24_27_lag
with low education
(0.005)
(0.008)
(0.011)
(0.009)
(0.007)
(0.009)
(0.007)
(0.010)
Rain current and lagged
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
HH & year FE Sub-district clustered SE
Yes Yes
Yes Yes
Yes Yes
Yes Yes
Yes Yes
Yes Yes
Yes Yes
Yes Yes
Controls
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Constant
2.9312**
2.6903
5.0231
4.6843*
3.2026
6.7410**
2.8081
3.3319
(1.387)
(2.048)
(3.816)
(2.794)
(2.021)
(2.822)
(1.892)
(3.291)
Observations R-squared
24,544 0.589
12,244 0.525
13,468 0.602
8,024 0.615
4,292 0.565
4,494 0.618
6,572 0.450
6,157 0.544
Number of hhid_n
4,909
4,766
3,027
3,454
1,841
1,760
3,016
2,740
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
63
Table A20: Estimated effect of climate variability on the income Gini coefficient estimated at provincial level in Indonesia, seasonal variables
All
Poor
Rural
Agricultura l HH
Female HH head
(1)
(2)
(3)
(4)
(5)
Gini coefficient
Gini coefficient
Gini coefficient
Gini coefficient
Gini coefficient
tembin15_18_y_lag_dry
-0.0178 (0.030)
-0.0315 (0.064)
0.0155 (0.031)
-0.0731* (0.041)
tembin21_24_y_lag_dry
0.0012 (0.004)
0.0044 (0.007)
0.0043 (0.004)
tembin24_27_y_lag_dry
0.0011
0.0047
(0.004)
(0.007)
Young HH head
Old head
(7)
(8)
(9)
Gini coefficient
Gini coefficient
Gini coefficient
-0.0423 (0.042)
0.0287 (0.043)
-0.0317 (0.048)
0.0320 (0.048)
-0.0179** (0.009)
0.0009 (0.006)
0.0086** (0.004)
-0.0004 (0.005)
-0.0011 (0.007)
0.0041
-0.0182**
0.0014
0.0093***
0.0010
-0.0008
(0.004)
(0.009)
(0.006)
(0.003)
(0.006)
-0.0001
(0.007)
0.0035
0.0030
-0.0193**
0.0003
0.0081***
0.0000
(0.004)
-0.0018
(0.007)
(0.004)
(0.009)
(0.006)
(0.003)
(0.006)
(0.007)
-
-0.1082***
-
-
-
Indonesia
tembin27_30_y_lag_dry tembin30_40_y_lag_dry
-0.1428*** (0.021)
tembin15_18_y_lag_wet
Head with low education
HH
(0.027)
0.1152***
0.1182**
0.1329***
0.1306***
0.0801**
0.1118***
0.0411
0.1248***
(0.025)
(0.051)
(0.020)
(0.021)
(0.034)
(0.027)
(0.074)
(0.020)
0.0012
0.0056**
0.0023
0.0040
0.0008
0.0052
-0.0024
0.0012
(0.002)
(0.003)
(0.003)
(0.003)
(0.004)
(0.004)
(0.003)
(0.003)
tembin24_27_y_lag_wet
0.0002 (0.002)
0.0042 (0.003)
0.0014 (0.003)
0.0033 (0.003)
-0.0007 (0.004)
0.0033 (0.004)
-0.0040 (0.003)
0.0007 (0.003)
tembin27_30_y_lag_wet
0.0003 (0.002)
0.0041 (0.003)
0.0013 (0.003)
0.0032 (0.003)
-0.0006 (0.004)
0.0030 (0.004)
-0.0040 (0.003)
0.0007 (0.003)
tembin30_40_y_lag_wet
-0.0289*** (0.008)
-0.0190* (0.010)
-0.0288*** (0.010)
-0.0228** (0.009)
-0.0298*** (0.009)
-0.0378*** (0.008)
-0.0268*** (0.010)
-0.0251** (0.010)
o.tembin15_18_y_lag2_dry
-
-
-
-
-
-
-
-
tembin21_24_y_lag2_dry
-0.0032 (0.004)
-0.0062 (0.006)
-0.0066* (0.003)
0.0123 (0.008)
-0.0042 (0.005)
-0.0124*** (0.002)
-0.0034 (0.005)
-0.0011 (0.006)
tembin24_27_y_lag2_dry
-0.0019 (0.004)
-0.0036 (0.006)
-0.0054 (0.003)
0.0138 (0.009)
-0.0031 (0.004)
-0.0107*** (0.002)
-0.0010 (0.005)
0.0000 (0.006)
tembin27_30_y_lag2_dry
-0.0011 (0.004)
-0.0023 (0.006)
-0.0045 (0.003)
0.0145* (0.009)
-0.0026 (0.004)
-0.0097*** (0.002)
0.0002 (0.005)
0.0007 (0.006)
tembin30_40_y_lag2_dry
0.0003 (0.007)
0.0123 (0.012)
0.0038 (0.011)
0.0263* (0.014)
0.0019 (0.009)
-0.0166** (0.008)
0.0140 (0.010)
0.0091 (0.010)
o.tembin15_18_y_lag2_wet
-
-
-
-
-
-
-
-
tembin21_24_y_lag2_wet
0.0003
0.0050**
0.0019
0.0040
-0.0007
0.0008
0.0055**
0.0034
tembin24_27_y_lag2_wet
(0.001) 0.0002
(0.002) 0.0036
(0.002) 0.0016
(0.003) 0.0034
(0.003) -0.0005
(0.002) 0.0001
(0.002) 0.0045*
(0.003) 0.0025
tembin27_30_y_lag2_wet
(0.001) 0.0004
(0.002) 0.0034
(0.002) 0.0018
(0.003) 0.0036
(0.003) -0.0003
(0.002) 0.0001
(0.002) 0.0047**
(0.003) 0.0025
(0.001)
(0.002)
(0.002)
(0.003)
(0.003)
(0.002)
(0.002)
(0.003)
0.0014
0.0050
-0.0119**
-0.0030
-0.0101
-0.0027
0.0050
0.0000
(0.005)
(0.007)
(0.005)
(0.006)
(0.007)
(0.006)
(0.006)
(0.007)
Rain and rain lagged
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Commune & year FE Region, commune
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Constant
0.7143* (0.413)
-1.3896 (1.030)
-0.0177 (0.825)
0.2860 (1.058)
1.0661 (1.319)
0.5863 (1.130)
0.1359 (0.784)
0.2135 (1.152)
Observations R-squared
24,544 0.622
12,244 0.570
13,468 0.641
8,024 0.664
4,292 0.599
4,494 0.658
6,572 0.506
6,157 0.560
Number of hhid_n
4,909
4,766
3,027
3,454
1,841
1,760
3,016
2,740
tembin21_24_y_lag_wet
tembin30_40_y_lag2_wet
and
year clustered SE Controls
Robust standard errors in parentheses *** p<0.01, * p<0.05, * p<0.1
64
Table A21: Summary table of the changes in the coefficient of interest when controlling for transfers (dummy in Panel A and log of real value in Panel B) received by the households in Indonesia. Source: Authors' calculation using IFLS data (1993-2014) Panel A: Transfer dummy Annual
Gini coefficient
Transfer variable
All
no difference
Mostly agricultural
no difference
Rural
no difference
Young head
no difference
Old head
no difference
Low education
no difference
Female head
no difference
Poor
no difference
positive positive positive positive not significant positive positive positive
Panel B: Log real value of transfer received Annual
Gini coefficient
Transfer variable
All
no difference
not significant
Mostly agricultural
no difference
not significant
Rural
no difference
not significant
Young head
no difference
not significant
Old head
no difference
not significant
Low education
no difference
not significant
Female head
no difference
not significant
Poor
no difference
not significant
Table A22; Summary table of the changes in the coefficient of interest when controlling for transfers (dummy in Panel A and log of real value in Panel B) received by the households in Indonesia. Source: Authors' calculation using VHLSS data (1993-2014) Panel A: Transfer dummy Seasonal
Gini coefficient Wet
Dry
Transfer variable
All
no difference
no difference
Mostly agricultural
no difference
no difference
Rural
no difference
no difference
Young head
no difference
no difference
Old head
no difference
no difference
Low education
no difference
no difference
Female head
no difference
no difference
positive positive positive positive positive positive positive
Panel B: Log real value of transfer received Seasonal
Gini coefficient Wet
Dry
Transfer variable
All
no difference
no difference
Mostly agricultural
no difference
no difference
Rural
no difference
no difference
Young head
no difference
no difference
Old head
no difference
no difference
Low education
no difference
no difference
Female head
no difference
no difference
Poor
no difference
no difference
positive positive positive positive not significant not significant positive positive
65
Appendix - 2 Synthetic panels for Vietnam, Household level data
Summary for Tables B 4.1-B4.4 1. Tables B4.1-B4.4 report estimates of mobility measured by changes in the proportion of households that move across poverty status. The tables show point estimates and standard errors of true panel mobility, together with lower and upper bound predictions of poverty mobility for the specification showed in Table B2. The model overall works well: true panel estimates lie within the estimated bounds and most of the point estimates lie within the 95 percent confidence interval of the true poverty rate. 2. For example, Table B4.2 suggests that 6.5 and 6.2 percent of households escape poverty in Vietnam between 2010-2012 and 2014-2016 respectively, while the actual panel dataset suggests that actual upward mobility is 9.7 and 7.0 percent. 3. Similarly, true downward mobility (Table B4.3) is 4.2 and 1.5 percent in Vietnam between 20102012 and 2014-2016 respectively, while the model predicts that 2.8 and 1.4 entered poverty between first and second rounds of the survey. Summary for Figures B1-B2 4. We plot the proposed point estimates of poverty transition rates for sub-groups of the population in Vietnam categorized by ethnicity (i.e., ethnic minority groups), by gender and employment sector of household heads (i.e., employed in agriculture or in other sectors), and residence areas (i.e., urban or rural households) in Figures B1 and B2. 5. These graphs indicate that female-headed are most likely to experience substantial transitions in and out of poverty in all periods compared to male-headed households. For example, Figure B1 suggests 48 percent of the population with female household heads move out of poverty in the past 2 years in 2014-2016, while 41 percent of population with male heads have similar transition. 6. Ethnic minority groups have a higher probability of moving into poverty than ethnic majority groups in all periods. For example, Figure B2 suggests 37 percent of the minor ethnicity population move in poverty in the past 2 years in 2008-2010, while only 9 percent of population with major ethnicity have similar transition. 7. On the other hand, households living in an urban area appear to be better off than those living in rural area: only 4 percent of urban population moved into poverty between 2008-2010 compared to 15 percent of rural households (Figure B2). Summary for Figures B3-B4 8. Figure B3 and B4 show the results of locally weighted regressions of (conditional) probability of upward mobility on the mean temperature (Panel A) and rainfall (Panel B) of the past 12 month (Figure B3) and of the past 12-24 month (Figure B4) from the second survey round. 9. Although, Figure B3 and Figure B4 suggest that higher temperatures are positively correlated with upward mobility (Panel A), the effect of temperature on mobility seems to be quite non-linear at the tails of the distributions. 10. On the other hand, the effect of rainfall on upward mobility (Panel B) is negative and quite linear in Figure B4 but has a non-linear shape in Figure B3 with negative effect starting from 2000mm.
66
Summary for Figures B5-B8 11. The same strategy as in Figure B4 was applied to various demographic groups. There is unequal effect of rising temperature and rainfall on downward mobility of households by ethnicity, gender and occupation of household head and by locality. 12. Figure B5 indicates that male-headed households have the higher the sensitivity to warm and moderately hot days: rising the temperature from 20C to 26C is associated with lower downward mobility in this group than in female-headed households (Panel A). 13. Figure B6 shows weak positive effect of rising temperature on downward mobility for major ethnicities and weak negative effect of warming on minor ethnicities (Panel A). 14. Figures B7 and B8 show that poverty status of households with heads employed in agriculture and living in rural areas are more sensitive to rising the temperature (Panel A) as well as to rising rainfall (Panel B). Summary for Figures B9-B13 15. We also estimated the same correlations than before but instead of using mean temperature, we use deviations from (monthly) mean temperature over the past 3, 5, 10, 15 and 20 years. 16. Figures B9 and B10 are based on data in 2002, 2004, 2006 and 2008. We find that positive temperature shock has negative effect on probability of becoming poor (Figure B9) and positive effect on upward mobility (Figure B10). Relationship is linear and holds regardless of the period we used as a base. 17. Figures B11 and B12 are based on data 2008, 2010, 2012, 2014 and 2016. We again find the same pattern as in Figures B9 and B10, but the effect of positive temperature shock on mobility is more non-linear. 18. The main findings in Figures B13 and B14 are not changed by using all available data (2002-2016).
67
Table B1. Poverty Rates, Vietnam
VHLSS 2002 2004 2006 2008 2010 2012 2014 2016
Poverty headcount ratio At $1.90 a day (2011 PPP) At national poverty lines
28.8 19.5 13.5 11.7 20.7 17.2 13.5 9.8
38 26.5 19.5 14.8 4.2 2.8 2.7 2
20.7 17.2 13.5 9.8
Source: VHLSS, Poverty & Equity Databank and PovcalNet
Table B2. Estimated parameters of household consumption per capita (first stage regression: OLS model for second year based on characteristics of household heads),VHLSS 2004 0.010*** (0.00) Female 0.093*** (0.02) Education level (reference level - < primary) Primary 0.157*** (0.02) Lower secondary 0.228*** (0.02) Upper-secondary 0.373*** (0.03) Technical degree 0.502*** (0.02) Post-secondary 0.808*** (0.03) Ethnic minorities -0.452*** (0.02) Type of location Urban 0.457*** (0.02) _cons 8.504*** (0.04) 0.438 Adjusted R2 Number of observations 6 784 Age of head
2006 0.010*** (0.00) 0.084*** (0.01)
2008 0.009*** (0.00) 0.087*** (0.02)
2010 0.009*** (0.00) 0.121*** (0.02)
2012 0.010*** (0.00) 0.082*** (0.02)
2014 0.010*** (0.00) 0.076*** (0.02)
2016 0.012*** (0.00) 0.100*** (0.02)
0.132*** (0.02) 0.182*** (0.02) 0.361*** (0.03) 0.504*** (0.02) 0.805*** (0.03) -0.414*** (0.02)
0.168*** (0.02) 0.241*** (0.02) 0.411*** (0.03) 0.514*** (0.02) 0.848*** (0.03) -0.418*** (0.02)
0.202*** (0.02) 0.294*** (0.02) 0.440*** (0.03) 0.567*** (0.02) 0.922*** (0.03) -0.516*** (0.02)
0.220*** (0.02) 0.301*** (0.02) 0.448*** (0.03) 0.550*** (0.02) 0.892*** (0.03) -0.482*** (0.02)
0.200*** (0.02) 0.269*** (0.02) 0.455*** (0.03) 0.559*** (0.03) 0.851*** (0.03) -0.571*** (0.02)
0.173*** (0.02) 0.267*** (0.02) 0.420*** (0.03) 0.538*** (0.03) 0.820*** (0.03) -0.571*** (0.02)
0.417*** (0.02) 8.723*** (0.04) 0.435
0.338*** (0.02) 8.763*** (0.04) 0.403
0.292*** (0.02) 9.255*** (0.04) 0.426
0.250*** (0.02) 9.294*** (0.04) 0.410
0.211*** (0.02) 9.400*** (0.04) 0.428
0.218*** (0.02) 9.491*** (0.04) 0.431
6 788
6 669
6 829
6 665
6 500
6 347
Note: *** p<0.01, ** p<0.05, * p<0.1; Standard errors clustered at psu are in parentheses. Household heads' ages are restricted to between 25 and 55 for the first survey round and between 27 and 57 for the second survey round. All estimates are obtained using cross sectional (representative) data for each year
68
Table B3. Comparison of estimates Dang et al.(2019) 2012-2014
Period Model
Dang et al.(2014) 2006-2008 Specification 2
Our estimates 2006-2008 2012-2014
First Stage Year R_squared N
2008 0.465 1,335
2008 0.403 6,669
2014 0.428 6,500
10.1
8.7
11.6
4.5
3.8
4.2
4.6
3
2.4
80.9
84.5
81.8
3,701
6,669
6,500
Poverty Status Poor, Poor Poor, Non-poor Non-poor, Poor Non-poor, Non-poor N
10.80 (0.3) 5.90 (0.1) 4.00 (0.1) 79.30 (0.4) 3,519
References: Dang, Hai-Anh, Dean Jolliffe, and Calogero Carletto. "Data Gaps, Data Incomparability, and Data Imputation: A Review of Poverty Measurement Methods for Data-Scarce Environments." Journal of Economic Surveys, (2019) Dang, Hai-Anh, Peter Lanjouw, Jill Luoto, and David McKenzie. "Using repeated cross-sections to explore movements into and out of poverty." Journal of Development Economics 107 (2014): 112-128
Table B4. Estimated rho from cross-sectional data, VHLSS partial rho
Period
simple rho
round 1
round 2
2002-2004 2004-2006 2006-2008 2008-2010 2010-2012 2012-2014 2014-2016
0.96 0.95 0.94 0.92 0.93 0.95 0.93
0.92 0.87 0.85 0.87 0.86 0.88 0.85
0.83 0.87 0.86 0.79 0.85 0.89 0.84
69
Poverty dynamics from synthetic panel data for Vietnam Table B4.1 Poverty dynamics (Poor in Year 1 and Poor in Year 2) Period 2002-2004
Non-parametric estimates Lower bound Upper bound 19.8 8.9
Point Estimates 16.1
Number of hhs 6,784
2004-2006
13.4
5.4
11.3
6,788
2006-2008
11.9
3.7
8.7
6,669
2008-2010
12.1
5.8
10.2
6,829
2010-2012
16.9
7.6
13.8
6,665
2012-2014
13.7
5.9
11.6
6,500
2014-2016
10.6
5.1
8.6
6,347
Truth
(se)
. . 10.8 (0.8) 7.7 (0.7) . . 12.7 (0.8) 10.6 (0.8) 8.0 (0.7)
Note: Poverty rates in percent are based on the household consumption per capita and predictions obtained using data in the second survey rounds. All numbers are weighted using population weights and adjusted for complex survey design. Number of replications for non-parametric estimates is 500. Household heads' ages are restricted to between 25 and 55 for the first survey round and between 27 and 57 for the second survey round. Consumption levels are deflated using CPI to obtain real levels in 2018 prices.
Table B4.2 Poverty dynamics (Poor in Year 1 and Non-Poor in Year 2) Period
Non-parametric estimates Lower bound Upper bound
Point Estimates
Number of hhs
Truth . . 10.3 (0.7) 5.6 (0.5) . . 9.7 (0.6) 6.8 (0.5) 7.0 (0.6)
2002-2004
7.5
16.7
9.3
6,784
2004-2006
7.1
13.2
7.6
6,788
2006-2008
1.3
8.3
3.8
6,669
2008-2010
0.0
5.7
1.7
6,829
2010-2012
4.2
12.3
6.5
6,665
2012-2014
1.8
9.4
4.2
6,500
2014-2016
4.7
9.4
6.2
6,347
(se)
Note: Poverty rates in percent are based on the household consumption per capita and predictions obtained using data in the second survey rounds. All numbers are weighted using population weights and adjusted for complex survey design. Number of replications for non-parametric estimates is 500. Household heads' ages are restricted to between 25 and 55 for the first survey round and between 27 and 57 for the second survey round. Consumption levels are deflated using CPI to obtain real levels in 2018 prices.
70
Table B4.3 Poverty dynamics (Non-Poor in Year 1 and Poor in Year 2) Non-parametric estimates Lower bound Upper bound
Point Estimates
Number of hhs
Truth . . 2.7 (0.4) 4.1 (0.4) . . 4.2 (0.4) 4.5 (0.5) 1.5 (0.3)
2002-2004
0.0
10.9
3.0
6,784
2004-2006
0.0
8.0
1.6
6,788
2006-2008
0.1
8.3
3.0
6,669
2008-2010
8.6
14.9
10.1
6,829
2010-2012
0.0
9.3
2.8
6,665
2012-2014
0.1
7.9
2.4
6,500
2014-2016
0.0
5.5
1.4
6,347
(se)
Note: Poverty rates in percent are based on the household consumption per capita and predictions obtained using data in the second survey rounds. All numbers are weighted using population weights and adjusted for complex survey design. Number of replications for non-parametric estimates is 500. Household heads' ages are restricted to between 25 and 55 for the first survey round and between 27 and 57 for the second survey round. Consumption levels are deflated using CPI to obtain real levels in 2018 prices.
Table B4.4 Poverty dynamics (Non-Poor in Year 1 and Non-Poor in Year 2) Period 2002-2004
Non-parametric estimates Upper Lower bound bound 72.7 63.5
Point Estimates
Number of hhs
71.6
6,784
2004-2006
79.6
73.4
79.5
6,788
2006-2008
86.6
79.6
84.5
6,669
2008-2010
79.3
73.6
78.0
6,829
2010-2012
78.9
70.7
76.8
6,665
2012-2014
84.4
76.8
81.8
6,500
2014-2016
84.6
80.0
83.8
6,347
Truth
(se)
. . 76.3 (1.1) 82.6 (1.0) . . 73.4 (1.1) 78.2 (1.1) 83.5 (1.0)
Note: Poverty rates in percent are based on the household consumption per capita and predictions obtained using data in the second survey rounds. All numbers are weighted using population weights and adjusted for complex survey design. Number of replications for non-parametric estimates is 500. Household heads' ages are restricted to between 25 and 55 for the first survey round and between 27 and 57 for the second survey round. Consumption levels are deflated using CPI to obtain real levels in 2018 prices.
71
Figure B1. Profiling of the Poor Population that Escaped Poverty (conditional probabilities), Vietnam 2002-2004
2006-2008
e al M
e al m Fe
in pl
he ot
rs to ec rs in pl
a
re ltu icu gr
or aj M
ty ici hn et
r ino M
ty ici hn et
l ra Ru
n ba Ur
60 40 20 e al M
e al m Fe
in pl
Em
Em
0
0
0
20
20
40
40
60
60
2004-2006
he ot
rs to ec rs Em
Em
in pl
a
re ltu icu gr
or aj M
ty ici hn et
r ino M
ty ici hn et
l ra Ru
n ba Ur
in pl
l ra Ru
n ba Ur
Em
60 40
he ot
in pl
a
ity ity nic nic th th e e r or aj ino M M
re ltu icu gr
l ra Ru
l ra Ru
n ba Ur
l ra Ru
n ba Ur
Em
e al M
e al m Fe
in pl
Em
20
Em
in pl
rs to ec rs
n ba Ur
60 Em
0
e al M
r ino M
l ra Ru
20 Em
he ot
ty ity re ici ltu nic hn th icu et e r r r o ag aj ino in M M pl
rs to ec rs
2014-2016
e al m Fe
or aj M
ty ici hn et
0 e al M
e al m Fe
in pl
Em
a
ty ici hn et
40
40 20 0 he ot
in pl
re ltu icu gr
2012-2014
60
60 40 20
in pl
ty ity re ici ltu nic hn th icu et e r r r o ag aj ino in M M pl
rs to ec rs
rs to ec rs
2010-2012
0
e al M
he ot
Em
2008-2010
e al m Fe
e al M
e al m Fe
n ba Ur
Em
72
he ot
ty ity re ici ltu nic hn th icu et e r r r o ag aj ino in M M pl
rs to ec rs Em
ot he r se cto rs
M al e
ag ric ul tu re M aj or et hn ici M ty ino re th nic ity
se cto rs
M al e
73 se cto rs
M al e
se cto rs
M al e
Fe m al e
Ur ba n
Ru ra l
Ur ba n
et hn ici ty
et hn ici ty
Ur ba n
M ino r
M aj or
ag ric ul tu re
ot he r pl in
Em pl in Em
se cto rs
M al e
Fe m al e
Ur ba n
Ru ra l
et hn ici ty et hn ici ty
ag ric ul tu re M aj or et hn ici ty M ino re th nic ity
pl in
ot he r
M ino r
M aj or
Ru ra l
2012-2014
Ru ra l
et hn ici ty
et hn ici ty
10 20 30 40
2010-2012
M ino r
0
Em
M al e se cto rs
ag ric ul tu re
ot he r
pl in
Em pl in
Em
Em pl in
Fe m al e
0
0
0
10 20 30 40
10 20 30 40
10 20 30 40
2004-2006
M aj or
ag ric ul tu re
ot he r pl in
pl in
Fe m al e
Ur ba n
Ru ra l
10 20 30 40
10 20 30 40
2008-2010
Em
se cto rs
M al e
ag ric ul tu re M aj or et hn ici ty M ino re th nic ity
pl in
ot he r
0
0
2002-2004
Em
Em
pl in
Fe m al e
Ur ba n
Em
ot he r
pl in
pl in
Em
Em
Fe m al e
Ur ba n
10 20 30 40
Ru ra l
0
2014-2016
Ru ra l
ag ric ul tu re M aj or et hn ici M ty ino re th nic ity
pl in
pl in
Em
Em
Fe m al e
Figure B2. Profiling of the Non-Poor Population that Enter Poverty (conditional probabilities), Vietnam 2006-2008
Weather and Poverty Figure B3. Correlation between average weather of the past 12 months and (conditional) probability (at household level) a poor household in the first period becomes non-poor in the second period
Panel B: Rainfall
.2
.25
.3
.35
.4
.45
Panel A: Temperature
18
20
22
24
26
28
Mean temperature of the past 12 months
74
1000
2000
3000
4000
Total rainfall of the past 12 months
5000
Figure B4. Correlation between average weather of the past 12-24 months and (conditional) probability (at household level) a poor household in the first period becomes non-poor in the second period Panel A: Temperature
.2
.3
.4
.5
.6
Panel B: Rainfall
20
22
24
26
28
0
1000
Mean temperature of the past 12-24 months
2000
3000
4000
Total rainfall of the past 12-24 months
75
5000
Figure B3b. Correlation between average weather of the past 12 months and (conditional) probability (at household level) a poor household in the first period becomes non-poor in the second period Panel B: Rainfall
0
.2
.4
.6
Panel A: Temperature
18
20
22
24
26
28
Mean temperature of the past 12 months
76
1000
2000
3000
4000
Total rainfall of the past 12 months
5000
Figure B4b. Correlation between average weather of the past 24 months and (conditional) probability (at household level) a poor household in the first period becomes nonpoor in the second period
Panel A: Temperature
0
.2
.4
.6
.8
Panel B: Rainfall
20
22
24
26
28
0
Mean temperature of the past 12-24 months
1000
2000
3000
4000
Total rainfall of the past 12-24 months
77
5000
Poverty and Weather by Demographic Groups Figure B5. Correlation between weather of the past 12-24 months and (conditional) probability (at household level) a non-poor household in the first period becomes poor in the second period, by gender of household head
Panel A: Temperature
0
.05
.1
.15
.2
Panel B: Rainfall
20
22
24
26
28
Mean temperature of the past 12-24 months
Female
Male
78
0
1000
2000
3000
4000
5000
Total rainfall of the past 12-24 months
Female
Male
Figure B6. Correlation between weather of the past 12-24 months and (conditional) probability (at household level) a non-poor household in the first period becomes poor in the second period, by ethnicity of household head Panel A: Temperature
0
.1
.2
.3
Panel B: Rainfall
20
22
24
26
28
0
Mean temperature of the past 12-24 months
Major
1000
2000
3000
4000
Total rainfall of the past 12-24 months
Minor
Major
79
Minor
5000
Figure B7. Correlation between weather of the past 12-24 months and (conditional) probability (at household level) a non-poor household in the first period becomes poor in the second period, by employment sector of household head
Panel B: Rainfall
0
.05
.1
.15
.2
.25
Panel A: Temperature
20
22
24
26
28
Mean temperature of the past 12-24 months
Other sectors
Agriculture
80
0
1000
2000
3000
4000
5000
Total rainfall of the past 12-24 months
Other sectors
Agriculture
Figure B8. Correlation between weather of the past 12-24 months and (conditional) probability (at household level) a non-poor household in the first period becomes poor in the second period, by locality
Panel B: Rainfall
0
.05
.1
.15
.2
.25
Panel A: Temperature
20
22
24
26
28
0
Mean temperature of the past 12-24 months
Rural
1000
2000
3000
4000
5000
Total rainfall of the past 12-24 months
Urban
Rural
81
Urban
Temperature shock (Using data 2002-2004, 2004-2006, 2006-2008)
.02
.03
.04
.05
.06
Figure B9. Correlation between temperature shock and (conditional) probability (at household level) a non-poor household in the first period becomes poor in the second period
-2
-1
0
1
2
Deviation of the temperature in first round from the previous mean over the period
over the past 3 years over the past 10 years over the past 20 years
Note: at disaggregated level by months
82
over the past 5 years over the past 15 years
.35
.4
.45
.5
.55
Figure B10. Correlation between temperature shock and (conditional) probability (at household level) a poor household in the first period becomes non-poor in the second period
-2
-1
0
1
2
Deviation of the temperature in first round from the previous mean over the period
over the past 3 years over the past 10 years over the past 20 years
Note: at disaggregated level by months
83
over the past 5 years over the past 15 years
Temperature shock (Using data 2008-2010, 2010-2012, 2012-2014, 2014-2016)
-.02
0
.02
.04
.06
.08
Figure B11. Correlation between temperature shock and (conditional) probability (at household level) a non-poor household in the first period becomes poor in the second period
-2
-1
0
1
2
3
Deviation of the temperature in first round from the previous mean over the period
over the past 3 years over the past 10 years over the past 20 years
Note: at disaggregated level by months
84
over the past 5 years over the past 15 years
.35
.4
.45
.5
.55
.6
Figure 12. Correlation between temperature shock and (conditional) probability (at household level) a poor household in the first period becomes non-poor in the second period
-2
-1
0
1
2
3
Deviation of the temperature in first round from the previous mean over the period
over the past 3 years over the past 10 years over the past 20 years
85
over the past 5 years over the past 15 years
Temperature shock (Using data 2002-2004, 2004-2006, 2006-2008, 2008-2010, 2010-2012, 2012-2014, 2014-2016)
0
.02
.04
.06
.08
Figure B13. Correlation between temperature shock and (conditional) probability (at household level) a non-poor household in the first period becomes poor in the second period
-2
-1
0
1
2
3
Deviation of the temperature in first round from the previous mean over the period
over the past 3 years over the past 10 years over the past 20 years
86
over the past 5 years over the past 15 years
.35
.4
.45
.5
.55
.6
Figure B14. Correlation between temperature shock and (conditional) probability (at household level) a poor household in the first period becomes non-poor in the second period
-2
-1
0
1
2
3
Deviation of the temperature in first round from the previous mean over the period
over the past 3 years over the past 10 years over the past 20 years
87
over the past 5 years over the past 15 years
Agence française de développement 5, rue Roland Barthes 75012 Paris l France www.afd.fr
What is AFD ? The Agence Française de Développement (AFD) Group is a public entity which finances, supports and expedites transitions toward a more just and sustainable world. As a French overseas aid platform for sustainable development and investment, we and our partners create shared solutions, with and for the people of the global South. Active in more than 4,000 projects in the French overseas departments and some 115 countries, our teams strive to promote health, education and gender equality, and are working to protect our common resources – peace, education, health, biodiversity and a stable climate. It’s our way of honoring the commitment France and the French people have made to fulfill the Sustainable Development Goals. Towards a world in common. Publication Director Rémy Rioux Editor-in-Chief Thomas Melonio Legal deposit 3rd quarter 2020 ISSN 2492 - 2846 © AFD Graphic design MeMo, Juliegilles, D. Cazeils Layout AFD Printed by the AFD reprography service To browse our publication: https://www.afd.fr/en/ressources-accueil