Policy Research Working Paper 10383 Is Climate Change Slowing the Urban Escalator out of Poverty? Evidence from Chile, Colombia, and Indonesia Shohei Nakamura Kseniya Abanokova Hai-Anh Dang Shinya Takamatsu Chunchen Pei Dilou Prospere Poverty and Equity Global Practice & Development Data Group March 2023 Policy Research Working Paper 10383 Abstract While urbanization has great potential to facilitate poverty to escape from poverty, indicating better access to economic reduction, climate shocks represent a looming threat to such opportunities in those areas. However, the climate shocks upward mobility. This paper empirically analyzes the effects offset such benefits of urban agglomerations, as extreme of climatic risks on the function of urban agglomerations rainfalls and high flood risks significantly reduce the chance to support poor households to escape from poverty. Com- of upward mobility. The findings underscore the need to bining household surveys with climatic datasets, the panel enhance resilience among the urban poor to allow them to regression analysis for Chile, Colombia, and Indonesia finds fully utilize the benefits of urban agglomerations. that households in large metropolitan areas are more likely This paper is a product of the Poverty and Equity Global Practice and the Development Data Group, Development Economics. . It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http:// www.worldbank.org/prwp. The authors may be contacted at snakamura2@worldbank.org and hdang@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Is Climate Change Slowing the Urban Escalator out of Poverty? Evidence from Chile, Colombia, and Indonesia1 Shohei Nakamura, World Bank Kseniya Abanokova, World Bank Hai-Anh Dang, World Bank Shinya Takamatsu, World Bank Chunchen Pei, Beijing Normal University Dilou Prospere, Osaka University JEL Classification: R23; O18; I32 Keywords: Migration; Urban agglomeration; Poverty; Climatic Change; Flooding 1 Corresponding author: Shohei Nakamura (snakamura2@worldbank.org). This paper was prepared as a background paper for the World Bank report Thriving: Making Cities Green, Resilient, and Inclusive in a Challenging Climate. An updated version of this paper will be forthcoming in the International Journal of Environmental Research and Public Health. We would like to thank Mark Roberts, Megha Mukim, Sailesh Tiwari, Leonardo Lucchetti, Somik Lall, Rinku Murgai, and Carlos Rodriguez Castelan for their valuable comments on an earlier draft. We are also grateful to Benny Istanto, Imam Setiawan, and Luis Quintero for sharing and processing datasets for us. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations or those of the Executive Directors of the World Bank or the governments they represent. 1. Introduction Urban areas around the world are attractive places for people looking for opportunities for a better life. About 55 percent of the world’s population lived in urban areas in 2018 and that number is likely to grow by 68 percent by 2050 (United Nations 2019). Urban agglomerations spur economic growth through productivity gains within economic sectors and structural transformation (Duranton and Puga 2004; Glaeser et al. 1992; Glaeser and Gottlieb 2009; Michaels, Rauch, and Redding 2012). At the same time, urban residents tend to be vulnerable to climatic and environmental shocks triggered by the increase in economic activities and are often pushed back to or remain trapped in poverty (Hallegatte et al. 2017). Without proper mitigation and adaptation measures against climate change, the benefits of urbanization could be negated (Mukim and Roberts 2022). In this context, this paper attempts to test the following two hypotheses. The first is that people are more likely to become or stay nonpoor in larger or denser cities, compared to smaller or less densely populated towns. The second hypothesis is that large or dense cities that are more exposed to climatic and environmental shocks offer residents a lower chance to become or stay nonpoor, compared to cities of similar size but not exposed to such shocks. By empirically testing those hypotheses, we investigate the following key question: Do climatic and environmental shocks hamper the key function of urban agglomerations as the escalator out of poverty in the developing world? Confirming this question is critically important as it underscores the need for policy interventions to help achieve inclusive and green growth through urban development. We develop an analytical approach to addressing the questions with and without panel datasets by combining panel (or synthetic panel) household surveys with climatic datasets. The synthetic panel method is a useful approach to analyzing poverty dynamics when panel household survey datasets are not available. We develop synthetic panel datasets out of repeated cross-sectional household surveys in Chile between 2011 and 2015 and Colombia between 2008 and 2010. We then examine the association between poverty changes over time and city population size as well as the heterogeneity of the association by flood risks. We also analyze another country, Indonesia, to apply a similar analytical framework. Instead of developing a synthetic panel, however, we estimate two- way fixed-effect (FE) regressions on the five waves of the Indonesia Family and Life Surveys (IFLS) to analyze the variation of probabilities of poor households escaping from poverty by urban agglomeration classifications and climatic shocks/risks. The IFLS spans from 1993 to 2014 over 298 districts and tracks the same households over time. We focus on flood as the climatic factor, by measuring the rainfall anomaly and heavy rain measured by the Standardized Precipitation Evapotranspiration Index (SPEI). The results of our analysis support the hypothesis that climatic risks could undermine the upward mobility facilitated by urban agglomerations. The synthetic panels in Chile between 2011 and 2015 and Colombia between 2008 and 2010 indicate a reduction in urban poverty rates measured by the upper-middle-income international poverty line (US$5.5 per day in 2011 purchasing power parity [PPP]). In those countries, 7.4 percent and 4.1 percent of the urban poor escaped from poverty during the periods, respectively. The analysis finds that the probabilities of households’ transition from poor to nonpoor status were positively correlated with the city population size in both countries. More importantly, such upward mobility was observed only in larger cities with low flood risk. The empirical analysis of Indonesia finds that compared to rural areas, the chance of getting out of vulnerability is higher by 7.0 percentage points in metropolitan cores based on the results with household FEs. However, heavy rainfall and high flood risk decrease the chance of upward mobility in the cores and urban peripheries of metropolitan areas. 2 Our paper contributes to the literature on the nexus between urbanization and poverty. Several studies show urban-rural gaps in productivity, wages, and amenities (see Lagakos 2020 for a review). Some studies find earnings and welfare gains from rural to urban migration in the developing world, such as rural to urban migration in Tanzania (Beegle, de Weerdt, and Dercon 2011), seasonal migration in Bangladesh (Bryan, Chowdhury, and Mobarak 2014), and the interplay of locations and migrant characteristics in determining gains in China (Combes et al. 2020). Hamory et al. (2021) analyzed panel datasets in Indonesia and Kenya, finding that a large part of the measured returns from migration came from the sorting of migrants. However, very few have analyzed the role of climate change as a hindrance to urban agglomeration as the urban escalator out of poverty. Therefore, we attempt to shed light on the role of the climatic and environmental shocks on the urban escalator out of poverty. The rest of the paper is organized as follows. Section 2 explains the framework with a description of our hypotheses. In Section 3, we elaborate on the methodology with the description of the datasets and our econometric approach. Section 4 reports the results of our econometric analysis. Finally, Section 5 concludes with the summary of findings and discussion, policy implications, and limitations. 2. Framework Urban escalator out of poverty: Hypothesis 1 Larger or denser urban areas potentially provide people with ample economic opportunities to escape from poverty, through better access to higher-wage jobs, infrastructure, services, and so on. As a result, more urbanized areas tend to be characterized by higher incomes and consumption, higher productivity, better access to services, and higher human capital. Earlier studies have found that nominal wages are higher in larger or denser cities due to productivity gains from agglomeration economies in developed countries (Glaeser and Mare 2001; Melo, Graham, and Noland 2009; Puga 2010; Rosenthal and Strange 2004) and in the developing world (Chauvin et al. 2017; Combes et al. 2022; Grover, Lall, and Timmis 2021; Quintero and Roberts 2022). Rural to urban migration can be welfare enhancing, given the gaps in income and amenities in the developing world (Lagakos 2020). When both poverty and urban areas are measured in a comparable way across countries, poverty tends to be lower in dense urban areas (Combes et al. 2022) (Figure 1).2 2 On the other hand, some studies highlight urbanization without growth (Bloom, Canning, and Fink 2008; Castells- Quintana and Wenban-Smith 2020; Fay and Opal 1999; Gollin, Jedwab, and Vollrath 2016). 3 Figure 1. Subnational extreme poverty rates across countries 72 47 42 42 44 46 38 35 32 29 28 24 22 23 21 18 12 13 14 9 8 5 2 2 2 2 2 3 1 0 2 2 0 0 0 Angola Bangladesh Egypt Ethiopia Ghana Tanzania Vietnam National Urban Urban center Urban cluster Rural Source: Combes et al. 2022. Note: Poverty is measured with the international poverty line (US$1.9 in 2011 PPP). Following the Degree of Urbanization methodology (Dijkstra et al. 2021), urban centers (clusters) are defined based on spatially contiguous sets of 1 km2 grid cells for which population density of each cell ≥ 1,500 (300) people per km2 and aggregate settlement population ≥ 50,000 (5,000). Nevertheless, the contribution of urbanization to poverty reduction is not self-evident. Larger or denser cities do not necessarily help people escape from poverty in the presence of overcrowding, congestion, crime, air pollution, high cost of living, insufficient jobs for low-skilled workers, segregation, and so on. Also, cross-sectional negative correlations between density and poverty do not necessarily indicate upward mobility. Thus, it is not evident a priori that larger or denser cities are the best place for people to escape from poverty (for example, the argument favoring secondary towns in Gibson, Jiang, and Susantono (2022) and Christiaensen and Todo (2014)). Moreover, larger or denser cities accommodating a larger share of richer people does not necessarily mean that people are more likely to be nonpoor in those cities due to the sorting. Therefore, it is an empirical question whether urban agglomerations facilitate poverty reduction. The first hypothesis to be tested is people are more likely to become or stay nonpoor in larger or denser cities, compared to smaller or less densely populated towns. Climatic and environmental stressors: Hypothesis 2 Even if urban agglomerations support poverty reduction, the upward mobility of the poor could be hampered by climatic shocks and risks. Urban households may fall into poverty due to higher exposure to shocks, asset vulnerability, and lack of socioeconomic resilience (Hallegatte et al. 2017). In urban areas, poor households tend to be exposed to floods due to their residences facing high environmental hazard risks, high building density and overcrowdedness, and inadequate infrastructure. The more urban a location, the scarcer and more expensive the land becomes, pushing the poor into undesirable and risky locations within or at the peripheries of cities. As net consumers, urban households are also vulnerable to food price shocks triggered by climate anomalies. The literature indeed found urban households more vulnerable to flooding/drought shocks in several countries. For example, Baez et al. (2017) analyzed the impacts of a severe tropical storm that hit Guatemala in 2020 with the largest rainfall in the country during the last five decades. Median per capita consumption fell by 12.6 percent in urban areas, a larger drop than in rural areas, significantly increasing urban poverty. Rising food prices due to the disaster shock contributed to the loss in urban households’ consumption, while a social safety net program protected mainly rural households. 4 Upward mobility offered by agglomeration economies could be offset and hindered by climatic and environmental stressors. Therefore, we hypothesize that large or dense cities that are more exposed to climatic and environmental shocks do not offer residents a higher chance to become or stay nonpoor, compared to cities of smaller size. 3. Methodology 3.1 Data We selected Chile, Colombia, and Indonesia as the cases for this study to demonstrate the application of analytical approaches with and without panel datasets. Analyzing these countries also merits the test of the approaches in countries where poverty is measured by income (Chile and Colombia) and consumption expenditures (Indonesia). The setting of Indonesia—its rapid urbanization and heterogenous urban and climatic characteristics across subnational regions—is particularly suitable to our analysis. Highly urbanized countries like Chile and Colombia have useful density variations to explore as well. To answer our research question and verify our hypotheses, we combined household surveys with climatic datasets. For Chile and Colombia, we constructed synthetic panel datasets out of repeated cross-sectional household surveys. Flood risk is estimated as a key climate factor for each town. For Indonesia, we relied on panel household surveys (IFLS), combined with two climate indicators: SPEI and the flood risk index. Synthetic panel data for Chile and Colombia Following Dang et al. (2014) and Dang and Lanjouw (2013), we applied the synthetic panel method to the household surveys of Chile (Encuesta de Caracterización Sociooeconómica Nacional [CASEN]) 2011 and 2015 and Colombia (Gran Encuesta Integrada de Hogares [GEIH]) 2008 and 2010.3 This method essentially exploits the time-invariant variables in the cross-sectional surveys and some cohort-based assumptions about the error terms to construct the synthetic panels. The methodology is described in detail in Annex B. Recent applications and further validations of the synthetic panel methods have been implemented using household survey data from various countries in Sub-Saharan Africa, East Asia and Pacific, Europe and Central Asia, Latin America, South Asia, and the Middle East and North Africa (see Dang, Jolliffe, and Carletto (2019); Dang and Lanjouw (forthcoming); and Garcés‐Urzainqui, Lanjouw, and Rongen (2021) for recent reviews). We begin by identifying the potential time-invariant variables available in the cross-sectional surveys at two time points, which include household heads’ gender, age, level of education, and residence area (that is, urban or rural). These variables can usually be assumed to be time invariant if the underlying population remains unchanged over time. One way to test this assumption is to use a t-test for the equality of the means of the same variables in the two survey rounds. We have provided test results that consider the complex survey design in Table A1 for Chile and Table A2 for Colombia. The assumption of the equality of means is satisfied for household heads’ gender and secondary level of education in Chile and heads’ secondary and tertiary levels of education in Colombia. The assumption is satisfied for residence areas in both countries. Although the difference in heads ’ primary level of education is statistically significant in Colombia, it has similar means over time. The differences between surveys for heads’ primary and tertiary levels of education in Chile and heads’ 3 The years for the household surveys for Chile and Colombia were selected based on the availability of the subnational location information that can be matched with climatic layers on the geographic information system (GIS) platform. 5 gender in Colombia are less than five percentage points. Thus, these may not make much difference to the final estimates in practice. IFLS panel household survey data: Indonesia The IFLS includes a total of 54,000 household observations over five waves from 2,556 subdistricts in 26 provinces. Focusing on the socioeconomic and health aspects of the households, the survey was conducted for the first time in 1993 covering 13 of the total 26 provinces in the country. In a sample of 22,000 individuals from 7,224 households, the survey collected data on individual respondents and their families (households) in addition to data on communities, health, and education facilities. In 1997/98, the second wave was administered to the same respondents with a recontact rate of 94.4 percent. That survey had the particularity to account for the economic and political crisis in Indonesia. The third wave in 2000 was managed to recontact 95.3 percent of the first wave sample while the fourth and the fifth rounds conducted in 2007/08 and 2014/15 recontacted 93.6 and 90.5 percent, respectively, of the first wave sample (Strauss et al. 2016, cited in Setiawan, Tiwari, and Rizal 2018). We measured poverty based on per capita consumption expenditures with the national poverty line following Setiawan, Tiwari, and Rizal (2018). The IFLS accounts for the expenditure of the consumption of 37 food items over a seven-day recall period and various nonfood items. 4 The nominal consumption aggregate is both temporally and spatially deflated.5 In addition to poverty, we identified vulnerable people with a vulnerability line, which is set at 1.5 times the poverty line. Roberts, Sander, and Tiwari (2019) highlight the importance of classifying urban areas based on their functionality instead of mere population size in Indonesia. Following Duranton (2015), we defined the following four location categories: (1) metro core, which stands for Jakarta or a district with the highest population density for other metros; (2) urban peripheries, which are predominantly urban non-core districts; (3) other urban areas that account for single-district metro (predominantly urban with kotas) or non-metro urban (predominantly urban non-metro districts); and (4) rural areas, which encompass the rural periphery (predominantly rural non-core district) or non-metro rural areas (predominantly rural non-metro districts). Climate data: Flood risk index and SPEI To account for climatic and environmental shocks, we used two indicators: flood risk index and the SPEI. Those climatic variables are prepared at the subdistrict level. The primary climatic stressor analyzed in this study is flood risk, given its potential threat to urban livelihoods. To capture the flood risk, we used the flood depth data provided by FATHOM in 2016. The flood depth is expressed in meters and computed at 3 arc-second (approximately 90 m) resolution and has a global coverage between 56°S and 60°N. The computation is based on pluvial data with a return period of 100 years (1-in-100 flood depth).6 The 1-in-100 flood depth means a flood event that has a 1 percent probability of occurring in any given year within 100 years. We classified the areas 4 Nonfood expenditures include household amenities (for example, refrigerator, TV, and telephone); housing; assorted items such as clothing, furniture, medical, ceremonies, education (tuition, uniform, transportation, boarding); and others. Regarding the housing expenditure, the actual monthly rent paid was recorded. However, if the household owns the house, the estimated rent was imputed. 5 Temporal deflation is based on the consumer price index series; spatial deflator is calculated based on the ratio of the regional poverty lines to the national poverty line, obtained from the National Socioeconomic Survey (SUSENAS) of the corresponding wave. 6 See https://agupubs.onlinelibrary.wiley.com/doi/epdf/10.1002/2015WR016954 for more details on the computation method. 6 with the top 25 percent flood depth in each country as high flood risk areas. As this is an indicator of long-term flood risk, we have an index only for one time point. The flood risk maps for three case countries are shown in Figure 2. Taking advantage of the panel data spanning over a long duration, we additionally analyzed rainfall anomalies as a climatic factor for Indonesia. The SPEI is a multiscalar drought index (Vicente-Serrano et al. 2010). The construction requires data on temperature, precipitation, and potential evaporation. Accordingly, we processed monthly precipitation and potential evapotranspiration derived from the terraclimate data from 1958 to 2020. The SPEI data are fitted to a gamma distribution and normalized to a flexible multiple time scale such as 1, 4, 6, 12, 24, and 48 months. For the study, we considered a 12-month time scale for each year from 1993 to 2015 with two lag periods for each IFLS wave.7 Negative SPEI values represent rainfall deficit—less than median precipitation—and high potential evapotranspiration (dry) starts when the SPEI value is equal to or below −1.0. On the other hand, positive SPEI values indicate rainfall surplus—greater than median precipitation—and low potential evapotranspiration (wet) starts when the SPEI value is equal to or above 1.0. Figure 3 illustrates the SPEI values in Indonesian districts as of March 2014. 7 The computation involves the following three steps (Harari and La Ferrara 2018). We first compute the difference (D) between precipitation and evapotranspiration (PET) and accounted for the climatic water balance defined at the monthly level. The Penman-Monteith equation is used to approximate the PET (as recommended by the Food and Agriculture Organization of the United Nations (FAO) as the best method for determining reference evapotranspiration). Maximum temperature, minimum temperature, vapor pressure, precipitation accumulation, downward surface shortwave radiation, and wind speed are used as input data. The next step is the aggregation of the climatic water balance at different time scales and finally, we standardize the time series according to a gamma distribution. The SPEI is then computed as the standardized values of the gamma function. 7 Figure 2. The maps of 100-year flood risks (A) Chile (B) Colombia (C) Indonesia Source: Based on FATHOM data. Figure 3. 12-month SPEI in Indonesia, March 2014 Source: Based on terraclimate data. Descriptive statistics Table 1 presents the summary statistics for Chile (Panel A) and Colombia (Panel B) based on the synthetic panel data. Households’ probability of transition from poor to nonpoor is the outcome variable. The average probability of poverty transition for Chile from 2011 to 2015 is 73.1 percent, 8 and for Colombia from 2008 to 2010 it is 16.8 percent. About 24.2 and 13.8 percent of households in Chile and Colombia, respectively, are in high flood risk areas. Table 1. Summary statistics, Chile and Colombia Count Mean SD Min Max Panel A: Chile Poor in 2011 (1 = yes, 0 = no) 44,614 0.096 0.295 0.000 1.000 Poor in 2015 (1 = yes, 0 = no) 61,433 0.039 0.193 0.000 1.000 Probability from poor to nonpoor between 2011 and 2015 36,035 0.731 0.101 0.485 0.930 Log of population size in 2015 36,035 10.887 1.334 7.432 14.434 High flood risk (1 = yes, 0 = no) 36,035 0.242 0.429 0.000 1.000 Panel B: Colombia Poor in 2008 (1 = yes, 0 = no) 188,801 0.276 0.447 0.000 1.000 Poor in 2010 (1 = yes, 0 = no) 190,344 0.241 0.428 0.000 1.000 Probability from poor to nonpoor between 2008 and 2010 119,692 0.168 0.076 0.053 0.399 Log of population size in 2010 119,692 12.745 0.923 7.988 14.546 High flood risk (1 = yes, 0 = no) 119,692 0.138 0.345 0.000 1.000 Sources: Based on CASEN 2011 and 2015 and GEIH 2008 and 2010. Note: Poverty measure is based on per capita household income, with a threshold of US$5.50 per day. The probability of changing poverty status from poor to nonpoor is estimated based on the synthetic panel approach described in Annex B). We classify the areas with top 25 percent flood depth in each country as high flood risk areas (see Section 3.1.3). SD = Standard deviation. Table 2 presents the summary statistics of key variables for Indonesia. Households’ poverty (1 = nonpoor; 0 = poor) and vulnerability (1 = neither poor nor vulnerable; 0 = poor or vulnerable) status are dummy variables used as the outcome variables for our regression analysis. Around 88 percent of household observations in our five-wave panel data are nonpoor, while 69 percent are neither poor nor vulnerable. The urban location typology—metro core, periphery urban, other urban, and the rural area—are also defined as dummy variables. Around 45 percent of household observations are from periphery urban, followed by other urban (19.2 percent), rural (18.6 percent), and metro core areas (17.4 percent). Households’ movements across locations between each of the five IFLS waves are summarized in Table A5. About 5.9 percent of households are in SPEI-dry areas, 90.7 percent in areas with SPEI-normal, and 3.4 percent in areas that experienced heavy rains. About 25 percent of households are exposed to high flood risks. Table 2. Summary statistics, Indonesia Count Mean SD Min Max Nonpoor (1 = yes, 0 = no) 47,796 0.877 0.328 0.000 1.000 Neither poor nor vulnerable (1 = yes, 0 = no) 47,796 0.690 0.462 0.000 1.000 City: Metro Core (1 = yes, 0 = no) 47,796 0.174 0.379 0.000 1.000 City: Periphery urban (1 = yes, 0 = no) 47,796 0.448 0.497 0.000 1.000 City: Other urban (1 = yes, 0 = no) 47,796 0.192 0.394 0.000 1.000 City: Rural (1 = yes, 0 = no) 47,796 0.186 0.389 0.000 1.000 SPEI: Dry (SPEI < −2) (1 = yes, 0 = no) 47,796 0.059 0.235 0.000 1.000 SPEI: Normal (1 = yes, 0 = no) 47,796 0.907 0.290 0.000 1.000 SPEI: Rainy (SPEI > 2) (1 = yes, 0 = no) 47,796 0.034 0.182 0.000 1.000 High flood risk (1 = yes, 0 = no) 47,796 0.251 0.433 0.000 1.000 Source: Based on IFLS 1993, 1997/98, 2000, 2007/8, and 2014/15. Note: Poverty is measured with the national poverty line; vulnerability is measured with the vulnerability line, which is set at 1.5 times the poverty line. We classify the areas with top 25 percent flood depth in each country as high flood risk areas (see Section 3.1.3). 9 3.2 Econometric approach Since the synthetic panel data only allow for studying the correlational relationship between climate change and poverty mobility, we examined the correlational relationship to provide support for the first hypothesis for Chile and Colombia. To test the second hypothesis for Colombia and Chile with synthetic panel datasets, we estimated the following first-difference regression model for household i in city j with the probability of transition from poor to nonpoor status between the two time points t0 and t1, or (0−1) : (0−1) = + 1 POPSIZE,0 + 2 (POPSIZE × CLMT ) + 3 CLMT + . (1) With POPSIZE,0 as the variable indicating the population size of city j at year t0 and CLMT as the 1- in-100-year flood risks at city j, is the error term. The parameter 2 indicates how the relationship between the city population size and the upward mobility (that is, the probability of poor people escaping from poverty) varies by the climatic risks. To test our first hypothesis that urban areas support upward mobility for Indonesia where the panel data are available, we estimated the following two-way FE model: = + 4 CITY + + + , (2) where stands for the poverty status (1 = nonpoor; 0 = poor) or vulnerability status (neither poor nor vulnerable = 1; 0 = poor or vulnerable) of household (i) in city (j) at year (t). Since there are no data on population size for Indonesia, we employed the variable CITY that indicates the location typology—metro core, urban periphery, other urban areas, and rural areas, with the rural areas as the reference category—for this country. and stand for the household FEs and the year FEs, respectively. With household FEs, we focused on the probability of escaping poverty among the movers. We expected the parameter 4 for multidistrict metropolitan areas to be positive based on the first hypothesis. For the second hypothesis, we analyzed the interaction of climatic conditions with the location effect of urban areas on poverty by adding to Equation (3) an interaction term between the location typology and the climate variable. = + 5 CITY + 6 (CITY × CLMT ) + 7 CLMT + + + , (3) where CLMT indicates the exposure to flood or flood risks. In the case of exposure to flood, the precipitation anomalies were measured for each IFLS wave (see Section 3.1). The parameter 6, the coefficient for the interaction term, captures the effect of the climate shocks associated with the city indicators. We estimated the panel regressions in Equations (2) and (3) as linear probability models with standard errors clustered at the enumerator areas. 4. Results 4.1 Chile and Colombia: Synthetic panel analysis The results of the synthetic panel analysis for Chile and Colombia show that the probabilities of urban residents escaping poverty are positively associated with the population size of their cities. From 2011 to 2015 in Chile, 7.4 percent of urban population (or two-thirds of the urban poor) escaped from poverty. As shown in column 1 in Table 3, the probability of the transition from poor to nonpoor is 10 positively correlated with the population size of cities. A similar correlation is observed for Colombia between 2008 and 2010 (column 1 in Table 4), where 4.1 percent of population escaped from poverty. We then estimated regression models in Equation (1) for Chile and Colombia to examine the heterogeneity in the link between the probability of poverty transition and city population size by flood risks. As shown in Column 4 in Table 3 and Table 4, the interaction term between the log of city population size and the flood risk variables is negative (−0.005 for Chile and −0.003 for Colombia), indicating that the upward mobility in large cities tends to be limited if they face high flood risks. With flat lines for high-risk areas and steep lines for low-risk areas, Figure 4 clearly shows such heterogeneity by flood risks. Table 3. First-difference model (dep var: the probability of transition from poor to nonpoor), Chile (1) (2) (3) (4) Log population 2015 0.0083*** 0.0075*** 0.0089*** (0.0004) (0.0004) (0.0004) Flood risk is high −0.0174*** −0.0136*** 0.0380*** (0.0012) (0.0011) (0.0085) Flood risk is high # log population 2015 −0.005*** (0.0008) Constant 0.641*** 0.736*** 0.654*** 0.638*** (0.0041) (0.0006) (0.0040) (0.0050) Observations 36,035 36,035 36,035 36,035 R-squared 0.012 0.006 0.016 0.016 Note: The table summarizes the estimation results of synthetic panel models in Equation (1). The dependent variable is the probability of each household’s transition from poor to nonpoor between 2011 and 2015. Standard errors in parentheses are estimated with 1,000 bootstraps. *p < 0.1, **p < 0.05, ***p < 0.01. Table 4. First-difference model (dep var: the probability of transition from poor to nonpoor), Colombia (1) (2) (3) (4) Log of population 2010 0.0029*** 0.0021*** 0.0025*** (0.0002) (0.0002) (0.0002) Flood risk is high −0.0101*** −0.009*** 0.0223*** (0.0006) (0.0006) (0.0075) Flood risk is high # Log pop. 2010 -0.0025*** (0.0006) Constant 0.132*** 0.170*** 0.143*** 0.138*** (0.0029) (0.0002) (0.0030) (0.0032) Observations 119,692 119,692 119,692 119,692 R-squared 0.001 0.002 0.003 0.003 Note: The table summarizes the estimation results of synthetic panel models in Equation (1). The dependent variable is the probability of each household’s transition from poor to nonpoor between 20 08 and 2010. Standard errors in parentheses are estimated with 1,000 bootstraps. *p < 0.1, **p < 0.05, ***p < 0.01. 11 Figure 4. Probability of poverty transition by city population size and flood risk (A) Chile (B) Colombia Source: Authors’ construction. Note: The predicted probabilities of becoming nonpoor are based on the results of Columns 4 in Table 3 (Chile) and Table 4 (Colombia). The error bars indicate 95% confidence intervals (CIs). 4.2. Indonesia: Panel data analysis Urban escalator out of poverty Table 5 summarizes the estimation results of the linear probability models in Equation (2) using the household’s nonpoor status as the dependent variable. Columns 1 and 2 report specifications without household FEs, whereas Columns 3 and 4 include household FEs. Year FEs are included in Columns 2 and 4. The main variable of interest is the indicator for metro core, its coefficient indicating the probability of households’ becoming nonpoor relative to those in rural areas. In the baseline specification with no FEs (Column 1), the coefficient estimate for metro core is 0.065 (90% CI = 0.019), meaning that the metro core offers a 6.6 percentage point higher probability to get out of poverty in comparison with the rural areas. Other urban areas also have a positive coefficient of 0.034 (90% CI = 0.020). By contrast, peri-urban areas show a negative coefficient, indicating that the chance of being nonpoor is lower than in rural areas. Adding year FEs in Column 2 does not change the result much. With household and year FEs (Column 4), the coefficient estimate for metro core is reduced to 0.031 (90% CI = 0.031). Other urban areas show an even smaller coefficient (0.0007, 90% CI = 0.036). 12 Table 5. Baseline linear probability models (dep var: nonpoor) (1) (2) (3) (4) City : Core 0.0659*** 0.0701*** 0.0145 0.0310 (0.0118) (0.0119) (0.0207) (0.0193) City: Periphery urban −0.0268** −0.0264** −0.0023 0.0005 (0.0125) (0.0125) (0.0165) (0.0144) City: Other urban 0.0342*** 0.0338*** 0.0054 0.0073 (0.0128) (0.0128) (0.0250) (0.0225) City: Rural (Reference) Household FE No No Yes Yes Year FE No Yes No Yes Adjusted R2 0.0117 0.0189 −0.0000 0.0058 # of observations 47,795 47,795 47,795 47,795 # of households 18,490 18,490 18,490 18,490 Note: The table summarizes the estimation results of panel regression models in Equation (2) for households in the five waves of IFLS (1993, 1997/8, 2000, 2007/8, and 2014/15). The dependent variable is a binary indicator about household’s poverty status (1=nonpoor; 0=poor). Cluster robust standard errors in parentheses. *p < 0.1, **p < 0.05, ***p < 0.01. Another series of regressions in Table 6 replace the outcome variable with an indicator about vulnerability. In Columns 2 and 4, the coefficient estimate for metro core is −0.141 (without household FEs) and −0.070 (with household FEs), respectively. That is, the probability of being neither poor nor vulnerable increases for households living in the metro core areas, regardless of with or without household FEs. In other words, people living in (or moving to) the metro core areas are less likely to become poor or vulnerable. Table 6. Baseline linear probability models (dep var: neither poor nor vulnerable) (1) (2) (3) (4) City: Core 0.136*** 0.141*** 0.0482* 0.0701** (0.0196) (0.0196) (0.0286) (0.0276) City: Periphery urban −0.0504*** −0.0503*** −0.0168 −0.0159 (0.0194) (0.0194) (0.0260) (0.0242) City: Other urban 0.0610*** 0.0602*** 0.0310 0.0325 (0.0213) (0.0212) (0.0271) (0.0254) City: Rural (reference) Household FE No No Yes Yes Year FE No Yes No Yes Adjusted R2 0.0228 0.0308 0.00019 0.0088 # of observations 47,795 47,795 47,795 47,795 # of households 18,490 18,490 18,490 18,490 Note: The table summarizes the estimation results of panel regression models in Equation (2) for households in the five waves of IFLS (1993, 1997/8, 2000, 2007/8, and 2014/15). The dependent variable is a binary indicator about household’s vulnerability status (1=not vulnerable; 0=vulnerable). Cluster robust standard errors in parentheses. *p < 0.1, **p < 0.05, ***p < 0.01. Climatic shock on the urban escalator Flood risk as a shock indicator Table 7 presents the estimates for Equation (3), showing the relation between the flood risk (as a climate shock indicator) and the location effect of urban areas on poverty. Column 4 with the interaction between location and flood risk variables, as well as household and year FEs, shows that the probability of being nonpoor decreases due to high flood risks by 7.4 percentage points (90% CI 13 = 0.066) in metro core and 5.33 percentage points (90% CI = 0.040) in periphery urban areas, respectively. That is, in metro areas, high flood risk lowers the chance of getting out of poverty in comparison with low-risk areas. Predicted probabilities in Figure 5 show high flood risk areas indicate a lower predicted probability of being nonpoor in each location indicator in comparison with the low- risk areas except for metro core without household FEs and rural areas with household FEs. Table 7. Linear probability models with flood risk (dep var: nonpoor) (1) (2) (3) (4) City: Core 0.0634*** 0.0601*** 0.0293 0.0467** (0.0123) (0.0149) (0.0194) (0.0196) City: Periphery urban −0.0274** −0.0299* −0.00034 0.0176 (0.0124) (0.0157) (0.0143) (0.0157) City: Other urban 0.0304** 0.0349** 0.0064 0.0189 (0.0127) (0.0153) (0.0227) (0.0231) City: Rural(reference) High flood risk −0.0257*** −0.0282 −0.0107 0.0317* (0.0092) (0.0192) (0.0111) (0.0189) City: Core # High flood risk 0.0329 −0.0743* (0.0253) (0.0404) City: Periphery urban # High flood risk 0.00827 −0.0533** (0.0235) (0.0244) City: Other urban # High flood risk −0.0232 −0.0322 (0.0270) (0.0259) Household FE No No Yes Yes Year FE Yes Yes Yes Yes Adjusted R2 0.0200 0.0203 0.0058 0.0061 # of observations 47,795 47,795 47,795 47,795 # of households 18,490 18,490 18,490 18,490 Note: The table summarizes the estimation results of panel regression models in Equation (3) for households in the five waves of IFLS (1993, 1997/8, 2000, 2007/8, and 2014/15). The dependent variable is a binary indicator about household’s poverty status (1=nonpoor; 0=poor). Cluster robust standard errors in parentheses. * p < 0.1, **p < 0.05, ***p < 0.01. Figure 5. Predicted probability of being nonpoor by locations (A) Without household FEs (B) With household FEs 1 0.94 0.92 0.95 Pr (nonpoor) 0.9 Pr (nonpoor) 0.9 0.88 0.85 0.86 0.84 0.8 0.82 0.75 0.8 Low High Low High Low High Low High risk risk risk risk risk risk risk risk Low High Low High Low High Low High risk risk risk risk risk risk risk risk Core Periphery Other Rural urban urban Core Periphery Other urban Rural urban Source: Authors’ construction. Note: The predicted probabilities of becoming nonpoor are based on the results of Columns 2 and 4 in Table 7. Error bars indicate 90% CI. 14 SPEI as a shock indicator Turning to SPEI, we replace the flood risk with the SPEI as the second climatic shock indicator in Table 8. We keep the nonpoor status as the dependent variable. As explained in Section 3, we divide the SPEI into three categories: SPEI-rainy (SPEI > 2.0), SPEI-normal (−2 ˃ SPEI ˃ 2), and SPEI-dry (SPEI < −2.0). The coefficient estimate for the interaction between the metro core and SPEI-rainy variables is −0.098 (90% CI = 0.058) in Column 4 with household FEs, suggesting that metro core areas that experienced heavy rains have 9.8 percentage points lower chance of getting out of poverty. That means SPEI strongly reduces the urban escalator function of metro core areas. Figure 6 shows the predicted probabilities confirming the pattern of SPEI-rainy for metro core areas. Except for other urban, areas with heavy rains show lower predicted probabilities of being nonpoor with or without household FEs. When we replaced the nonpoor indicator with the status of neither poor nor vulnerable as the outcome variable in Table 9, we made a similar conclusion. The coefficient of the interaction between metro core and SPEI-rainy (Column 4) is 0.131 (90% CI=0.036), meaning that the chance of being neither poor nor vulnerable for people moving to metro core areas that experience heavy rain decreases by 13.1 percentage points compared to those who did not face heavy rains. The result is less clear for the regressions without household FEs (Column 2). 15 Table 8. Linear probability models with SPEI (dep var: nonpoor) (1) (2) (3) (4) City: Core 0.0676*** 0.0670*** 0.0272 0.0273 (0.0115) (0.0112) (0.0197) (0.0197) City: Periphery urban −0.0263** −0.0268** −0.0025 −0.0032 (0.0124) (0.0118) (0.0151) (0.0152) City: Other urban 0.0323** 0.0284** 0.0047 0.0036 (0.0125) (0.0123) (0.0233) (0.0233) City: Rural(reference) SPEI: Dry 0.0220** 0.0358* 0.0256** 0.0185 (0.0104) (0.0212) (0.0109) (0.0200) SPEI: Normal (reference) SPEI: Rainy −0.0817* −0.0388** −0.0445 −0.0406** (0.0460) (0.0155) (0.0310) (0.0193) City: Core # SPEI: dry −0.0135 0.0090 (0.0215) (0.0198) City: Core # SPEI: rainy −0.00019 −0.0985*** (0.0619) (0.0356) City: Periphery urban # SPEI: dry −0.0323 0.0039 (0.0259) (0.0252) City: Periphery urban # SPEI: rainy 0.0495 0.0078 (0.0496) (0.0363) City: Other urban # SPEI: dry 0.0141 0.0193 (0.0226) (0.0257) City: Other urban # SPEI: rainy 0.128*** 0.0657** (0.0473) (0.0317) Household FE No No Yes Yes Year FE Yes Yes Yes Yes Adjusted R2 0.0195 0.0199 0.0065 0.0068 # of observations 47,795 47,795 47,795 47,795 # of households 18,490 18,490 18,490 18,490 Note: The table summarizes the estimation results of panel regression models in Equation (3) for households in the five waves of IFLS (1993, 1997/8, 2000, 2007/8, and 2014/15). The dependent variable is a binary indicator about household’s poverty status (1=nonpoor; 0=poor). Cluster robust standard errors in parentheses. * p < 0.1, **p < 0.05, ***p < 0.01. Figure 6. Predicted probability of being nonpoor by locations (A) Without household FEs (B) With household FEs 1 1 0.95 Pr (nonpoor) Pr (nonpoor) 0.9 0.9 0.85 0.8 0.8 0.75 0.7 0.7 Normal Normal Normal Normal Dry Dry Dry Dry Rainy Rainy Rainy Rainy Normal Normal Normal Normal Dry Dry Dry Dry Rainy Rainy Rainy Rainy Core Periphery Other urban Rural Core Periphery Other urban Rural urban urban Source: Authors’ construction. Note: The predicted probabilities of becoming nonpoor are based on the results of Columns 2 and 4 in Table 5. Error bars indicate 90% CI. 16 Table 9. Linear probability model, regression with SPEI (dep var: neither poor nor vulnerable) (1) (2) (3) (4) City: Core 0.137*** 0.134*** 0.0646** 0.0637** (0.0193) (0.0193) (0.0284) (0.0283) City: Periphery urban −0.0504*** −0.0502*** −0.0207 −0.0199 (0.0192) (0.0190) (0.0256) (0.0255) City: Other urban 0.0581*** 0.0556*** 0.0285 0.0281 (0.0210) (0.0214) (0.0265) (0.0264) City: Rural (reference) SPEI: Dry 0.0145 0.0235 0.0273* 0.00452 (0.0181) (0.0316) (0.0155) (0.0271) SPEI: Normal (reference) SPEI: Rainy −0.0610** −0.0900** −0.0598*** −0.0229 (0.0243) (0.0419) (0.0186) (0.0204) City: Core # SPEI: Dry 0.0195 0.0433 (0.0348) (0.0293) City: Core # SPEI: Rainy 0.0266 −0.131*** (0.0690) (0.0362) City: Periphery urban # SPEI: Dry −0.0369 0.0150 (0.0407) (0.0354) City: Periphery urban # SPEI: Rainy 0.0359 −0.0434 (0.0495) (0.0314) City: Other urban # SPEI: Dry 0.0132 0.0373 (0.0384) (0.0423) City: Other urban # SPEI: Rainy 0.0558 0.0541 (0.0768) (0.0370) Household FE No No Yes Yes Year FE Yes Yes Yes Yes Adjusted R2 0.0312 0.0313 0.0097 0.0099 # of observations 47,795 47,795 47,795 47,795 # of households 18,490 18,490 18,490 18,490 Note: The table summarizes the estimation results of panel regression models in Equation (3) for households in the five waves of IFLS (1993, 1997/8, 2000, 2007/8, and 2014/15). The dependent variable is a binary indicator about household’s vulnerability status (1=not vulnerable; 0=vulnerable). Cluster robust standard errors in parentheses. *p < 0.1, **p < 0.05, ***p < 0.01. 5. Discussion and conclusion This paper examines the effects of the climatic and environmental shocks on a key function of urban agglomerations to facilitate poverty reduction. Our study also showcases different empirical approaches depending on the availability of panel datasets. We constructed synthetic panel datasets for Colombia and Chile from repeated cross-sectional household surveys and examined the association between poverty changes over time and the city population size as well as the heterogeneity of such association by flood risks. By estimating two-way FE models with five waves of IFLS spanning from 1993 to 2015, we analyzed the probabilities of households escaping poverty in different locations—metro core, urban periphery, other urban areas, and rural areas—and flooding risks in Indonesia. The results from the three case countries show similar patterns. The synthetic panel analyses for Colombia and Chile indicate a reduction in urban poverty rates measured by the upper-middle-income international poverty line (US$5.5 per day in 2011 PPP). The analysis finds that the probabilities of households’ transition from poor to nonpoor status were positively correlated with the city population size in both countries. More importantly, such upward mobility was observed only in larger cities with low flood risk. The results of our two-way FE regression analyses for Indonesia suggest that dense metropolitan areas have provided good opportunities for those migrants to escape poverty in 17 Indonesia. However, high flood risk appears to have reduced such upward mobility in large metropolitan areas. There are several potential reasons. First, heavy rainfalls and flooding cost urban residents to repair and replace their damaged assets, such as dwellings. Neighborhoods with high building density and poor infrastructure could augment the damage and thereby the recovery costs. Second, flooding may lower the productivity and outputs of workers by damaging productive assets, reducing the time allocated for work, and constraining commuting. The findings suggest the importance of reducing flood risks to promote poverty reduction through migration to large metro areas. Upgrading high-density informal settlements would be an effective approach for adaptation. In the Indonesian urbanization context, it would also be important to invest in the peripheries of metropolitan areas, as they have been receiving a large influx of migration. It is essential to reduce congestion forces due to the increased migration and better connect peripheries to the cores as the latter provide more poverty-reducing opportunities. Finally, we clarify some limitations of our study. First, although we employed two-way FE regression models for Indonesia, we could not distinguish the sorting of migrants from the location effects. We would need a stronger identification strategy, such as a natural experimental design. If those with high capability to escape from poverty tend to move to cities, our estimation of location effects might be overestimated. Second, we focused on heavy rainfall and flooding as the climatic variable, though many other climatic and environmental stressors also potentially undermine the benefits of urban agglomerations. Finally, this is a case study of the three countries; thus, we cannot automatically generalize the findings to other contexts. 18 References Baez, J. E., L. Lucchetti, M. E. Genoni, and M. Salazar. 2017. “Gone with the Storm: Rainfall Shocks and Household Wellbeing in Guatemala.” The Journal of Development Studies 53 (8): 1253– 1271. Beegle, K., J. De Weerdt, and S. Dercon. 2011. “Migration and Economic Mobility in Tanzania: Evidence from a Tracking Survey.” Review of Economics and Statistics 93 (3): 1010–1033. Bloom, D. E., D. Canning, and G. Fink. 2008. “Urbanization and the Wealth of Nations.” Science 319 (5864): 772–775. Bryan, G., S. Chowdhury, and A. Mobarak. 2014. “Underinvestment in a Profitable Technology: The Case of Seasonal Migration in Bangladesh.” Econometrica 82 (5): 1671–1748. Castells-Quintana, D., and H. Wenban-Smith. 2020. “Population Dynamics, Urbanisation without Growth, and the Rise of Megacities.” The Journal of Development Studies 56 (9): 1663–1682. Chauvin, J., E. Glaeser, Y. Ma, and K. Tobio. 2017. “What Is Different about Urbanization in Rich and Poor Countries? Cities in Brazil, China, India and the United States.” Journal Of Urban Economics 98: 17–49. Christiaensen, L., and Y. Todo. 2014. “Poverty Reduction during the Rural-Urban Transformation: The Role of the Missing Middle.” World Development 63: 43–58. Combes, P-P., S. Demurger, S. Li, and J. Wang. 2020. “Unequal Migration and Urbanisation Gains in China.” Journal of Development Economics 142. Combes, P-P., S. Nakamura, M. Roberts, and B. Stewart. 2022. “Estimating Urban Poverty Consistently Across Countries.” World Bank Poverty and Equity Notes 48. Dang, H-A., P. Lanjouw, J. Luoto, and D. McKensize. 2014. “Using Repeated Cross-Sections to Explore Movements into and out of Poverty.” Journal of Development Economics 107: 112– 128. Dang, Hai-Anh, and Peter Lanjouw. Forthcoming. “Regression-Based Imputation for Poverty Measurement in Data Scarce Settings.” In Handbook of Research on Measuring Poverty and Deprivation, edited by Jacques Silber. Edward Elgar Press. Dang, Hai-Anh, and Peter Lanjouw. 2013. Measuring Poverty Dynamics with Synthetic Panels Based on Cross-Sections. 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 3: 757–797. Dijkstra, L., Florczyk, A. J., Kemper, T., Melchiorri, M., Pesaresi, M., Shiavina, M. 2021. Applying the degree of urbanisation to the globe: A new harmonised definition reveals a different picture of global urbanisation. J. Urban Econ. 125. Duranton, G. 2015. “A Proposal to Delineate Metropolitan Areas in Colombia.” Desarrollo y Sociedad 15: 223–64. 19 Duranton, G., and D. Puga. 2004. “Micro-foundations of Urban Agglomeration Economies.” Handbook of Regional and Urban Economics 4 (4): 2063–2117. Fay, M., and C. Opal. 1999. Urbanization without Growth: A Not-So-Uncommon Phenomenon. Oxford: World Bank. Garcés‐Urzainqui, D., P. Lanjouw, and G. Rongen. 2021. “Constructing Synthetic Panels for the Purpose of Studying Poverty Dynamics: A Primer.” Review of Development Economics 25 (4): 1803–1815. Gibson, J., Y. Jiang, and B. Susantono. 2022. “Revisiting the Role of Secondary Towns: Effects of Different Types of Urban Growth on Poverty in Indonesia.” University of Waikato Working Paper in Economics 5/22. Glaeser, E. L., and J. D. Gottlieb. 2009. “The Wealth of Cities: Agglomeration Economies and Spatial Equilibrium in the United States.” Journal of Economic Literature 47 (4): 983–1028. Glaeser, E. L., H. D. Kallal, J. A. Scheinkman, and A. Shleifer. 1992. “Growth in Cities.” Journal of Political Economy 100 (6): 1126–1152. Glaeser, E., and D. Maré. 2001. “Cities and Skills.” Journal of Labor Economics 19 (2). Gollin, D., R. Jedwab, and D. Vollrath. 2016. “Urbanization with and without industrialization.” Journal of Economic Growth 21 (1): 35–70. Grover, A., S. V. Lall, and J. Timmis. 2021. “Agglomeration Economies in Developing Countries: A Meta-Analysis.” World Bank Policy Research Working Paper No. 9730. Hallegatte, S., A. Vogt-Schilb, M. Bangalore, and J. Rozenberg. 2017. Unbreakable: Building the Resilience of the Poor in the Face of Natural Disasters. Washington, DC: World Bank. Hamory, J., M. Kleemans, N. Y. Li, and E. Miguel. 2021. “Reevaluating Agricultural Productivity Gaps with Longitudinal Microdata.” Journal of European Economic Association 19 (3): 1522–1555. Harari, M., and E. L. Ferrara. 2018. “Conflict, Climate, and Cells: A Disaggregated Analysis.” The Review of Economics and Statistics 100 (4): 594–608 Lagakos, D. 2020. “Urban-Rural Gaps in the Developing World: Does Internal Migration Offer Opportunities?” Journal of Economic Perspectives 34 (3): 174–192. Melo, P. C., D. J. Graham, and R. B. Noland. 2009. “A Meta-Analysis of Estimates of Urban Agglomeration Economies.” Regional Science and Urban Economics 39 (3): 332–342. Michaels, G., F. Rauch, and S. J. Redding. 2012. “Urbanization and Structural Transformation.” The Quarterly Journal of Economics 127 (2): 535–586. Mukim, M., and M. Roberts. 2022. Thriving: Making Cities Green, Resilient, and Inclusive in a Challenging Climate. Washington, DC: World Bank. Puga, D. 2010. “The Magnitude and Causes of Agglomeration Economies.” Journal of Regional Science 50 (1): 203–219 20 Quintero, L. E., and M. Roberts. 2022. “Cities and Productivity: Evidence from 16 Latin American and Caribbean Countries.” Johns Hopkins Carey Business School Research Paper No. 22–13. Roberts, M., F. G. Sander, and S. Tiwari. 2019. Time to Act: Realizing Indonesia’s Urban Potential. Washington, DC: World Bank. Rosenthal, S. S., and W. C. Strange. 2004. “Evidence on the Nature and Sources of Agglomeration Economies.” In Handbook of Regional and Urban Economics (Vol. 4), edited by J. V. Henderson and J. F. Thisse, 2119–2171. Amsterdam: Elsevier. Setiawan, I., S. Tiwari, and H. Rizal. 2018. Economic and Social Mobility in Urbanizing Indonesia. Background paper for Roberts, Sander, and Tiwari (2019). United Nations, Department of Economic and Social Affairs, Population Division. 2019. World Urbanization Prospects: The 2018 Revision (ST/ESA/SER.A/420). New York: United Nations. Vicente-Serrano, S.M., S. Beguera, and J. I. Lopez-Moreno. 2010. “A Multiscalar Drought Index Sensitive to Global Warming: The Standardized Precipitation Evapotranspiration Index.” Journal of Climate 23: 1696–1718. 21 Annex A: Additional tables Table A1. Summary statistics, Chile (2011–2015) 2011 2015 Difference 5.875 6.177 0.302*** The logarithm of per capita income (0.039) (0.034) (0.016) 42.940 45.607 2.667 Head’s age (0.111) (0.121) (0.153) 0.342 0.340 −0.002 Head is female (0.008) (0.005) (0.009) 0.133 0.118 −0.015*** Head does not complete primary school (0.007) (0.006) (0.004) 0.300 0.260 −0.039*** Head’s highest education level is primary (0.010) (0.008) (0.007) 0.397 0.410 0.013 Head’s highest education level is secondary (0.013) (0.010) (0.008) 0.170 0.211 0.041*** Head’s highest education level is tertiary (0.018) (0.016) (0.008) 0.879 0.875 −0.004 Urban area (0.012) (0.012) (0.004) Note: Standard errors are in parentheses, and the differences are estimated, considering the complex survey design. *p < 0.1, **p < 0.05, ***p < 0.01. Population weights are applied. We do not test for the difference in the distributions of the age variable in the two survey rounds since it is a deterministic variable. Household heads ’ ages are restricted to between 25 and 55 for the first survey round and between 29 and 59 for the second survey round. Table A2. Summary statistics, Colombia (2008–2010) 2008 2010 Difference 5.341 5.422 0.081*** The logarithm of per capita income (0.052) (0.044) (0.020) 41.073 42.317 1.243 Head’s age (0.104) (0.075) (0.070) 0.261 0.285 0.023*** Head is female (0.008) (0.008) (0.004) 0.225 0.239 0.014* Head does not complete primary school (0.016) (0.017) (0.008) 0.385 0.372 −0.013*** Head’s highest education level is primary (0.004) (0.005) (0.005) 0.288 0.293 0.004 Head’s highest education level is secondary (0.010) (0.011) (0.005) 0.101 0.097 −0.005 Head’s highest education level is tertiary (0.007) (0.007) (0.003) 0.814 0.815 0.002 Urban area (0.031) (0.028) (0.011) Note: Standard errors are in parentheses, and the differences are estimated, considering the complex survey design. *p < 0.1, **p < 0.05, ***p < 0.01. Population weights are applied. We do not test for the difference in the distributions of the age variable in the two survey rounds since it is a deterministic variable. 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. 22 Table A3. Estimated OLS model of household income per capita in 2015, Chile Coef/SE 0.014*** Head’s age (0.00) −0.139*** = 1 if the head is female (0.01) Education level (reference - if head does not complete primary school) 0.129*** = 1 if the head’s highest education level is primary (0.01) 0.398*** = 1 if the head’s highest education level is secondary (0.02) 1.183*** = 1 if the head’s highest education level is tertiary (0.07) 0.016 = 1 if the area of residence is urban (0.02) 5.141*** Constant (0.04) Adjusted R2 0.294 Number of observations 45,954 Note: Standard errors clustered at primary sampling units are in parentheses. *p < 0.1, **p < 0.05, ***p < 0.01. The dependent variable is the logarithm of household income per capita. Household heads’ ages are restricted to between 29 and 59. Table A4. Estimated OLS model of household income per capita in 2010, Colombia Coef/SE 0.019*** Head's age (0.00) −0.127*** = 1 if the head is female (0.02) Education level (reference - if head does not complete primary school) 0.319*** = 1 if the head’s highest education level is primary (0.02) 0.767*** = 1 if the head’s highest education level is secondary (0.02) 1.704*** = 1 if the head’s highest education level is tertiary (0.05) 0.358*** = 1 if the area of residence is urban (0.06) 3.915*** Constant (0.05) Adjusted R2 0.348 Number of observations 133,483 Note: Standard errors clustered at primary sampling units are in parentheses. *p < 0.1, **p < 0.05, ***p < 0.01. The dependent variable is the logarithm of household income per capita. Household heads ’ ages are restricted to between 30 and 60. 23 Table A5. Residential movement across IFLS waves, Indonesia (percentage of household) Wave 2 Metro core Periphery U Other U Rural Metro core 95.2 0.7 3.5 0.6 Periphery U 0.1 99.4 0.2 0.3 Wave 1 Other U 0.3 0.3 99.3 0.2 Rural 0.0 0.7 0.3 99.0 Wave 3 Metro core Periphery U Other U Rural Metro core 96.2 1.6 2.0 0.3 Periphery U 0.2 95.7 1.2 3.0 Wave 2 Other U 0.6 4.0 95.0 0.3 Rural 0.1 13.9 2.9 83.2 Wave 4 Metro core Periphery U Other U Rural Metro core 91.9 2.2 4.9 1.1 Periphery U 0.3 93.1 2.2 4.4 Wave 3 Other U 1.6 3.3 91.0 4.2 Rural 0.4 8.9 0.8 89.9 Wave 5 Metro core Periphery U Other U Rural Metro core 91.8 2.5 4.5 1.2 Periphery U 0.3 98.6 0.4 0.7 Wave 4 Other U 2.4 1.5 95.5 0.7 Rural 0.7 2.0 0.7 96.7 24 Annex B: Synthetic panel method This annex provides a brief summary of the synthetic panel method based on Dang et al. (2014) and Dang and Lanjouw (2013). Let be a vector of household characteristics observed in survey round j (j = 1 or 2) that are also observed in the other survey round for household i, i = 1,…, N. These household characteristics can include time-invariant variables such as ethnicity, religion, language, place of birth, and parental education as well as other time-varying household characteristics if retrospective questions about the first round values of such characteristics are asked in the second round survey. To reduce spurious changes due to changes in household composition over time, we usually restrict the estimation samples to household heads in a certain age range, for example, 25–55, in the first cross-section and adjust this age range accordingly in the second cross-section. This restriction also helps ensure certain variables such as heads’ education attainment remains relatively stable over time (assuming most heads have completed schooling).8 This age range is usually used in traditional pseudo-panel analysis but can vary depending on the cultural and economic factors in each specific setting. Population weights are then employed to provide estimates that represent the whole population. Then, let represent household consumption or income in survey round j, j = 1 or 2. The linear projection of household consumption (or income) on household characteristics for each survey round is given by ′ = + . (B.1) Let be the poverty line in period j. We are interested in knowing the unconditional measures of poverty mobility such as (1 < 1 2 > 2 ). (B.2) which represents the percentage of households that are poor in the first survey round (year) but nonpoor in the second survey round or the conditional measures such as (2 > 2 | 1 < 1 ), (B.3) which represents the percentage of poor households in the first round that escape poverty in the second round. If true panel data are available, we can straightforwardly estimate the quantities in B.2 and B.3, but in the absence of such data, we can use synthetic panels to study mobility. To operationalize the framework, we make two standard assumptions. First, we assume 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. More specifically, coupled with Equation (B.1), this implies the conditional distribution of expenditure in a given period is identical whether it is conditional on the given household characteristics in period 1 or period 2 (that is, 1 = 2 implies 1 |1 and 1 |2 have identical distributions). Second, we assume that i1 and i2 have a bivariate normal distribution with positive correlation coefficient ρ and standard deviations 1 and σ2 , respectively. Quantity (B.2) can be estimated by 8 While household heads may still increase their education achievement in theory, this rarely happens in practice. 25 ′ 1 −1 ′ 2 −2 2 2 (1 < 1 2 > 2 ) = Φ2 ( ,− , −), (B.4) 1 2 where 2 (. ) stands for the bivariate normal cumulative distribution function and 2 (. ) stands for the bivariate normal probability density function. In Equation (B.4), the estimated parameters obtained from data in both survey rounds are applied to data from the second survey round (x2) (or the base year) for prediction, but we can use data from the first survey round as the base year as well. It is then ′ 1 −1 straightforward to estimate quantity (B.3) by dividing quantity (B.2) by  ( 2 ) , where 1 (. )stands for the univariate normal cumulative distribution function. In Equation (B.4), the parameters and are estimated from Equation (B.1), and ρ can be estimated using an approximation of the correlation of the cohort-aggregated household consumption between the two surveys (12 ). In particular, given an approximation of 12 , where c indexes the cohorts constructed from the household survey data, the partial correlation coefficient ρ can be estimated by ′ ( ) 1 2 √(1 )(2 )−1 2 = . (B.5) 1 2 The standard errors of estimates based on the synthetic panels can in fact be even smaller than those of the true (or design-based) rate if there is a good model fit (or the sample size in the target survey is significantly larger than that in the base survey; see Dang and Lanjouw (2013) for more discussion). Equation (B.4) can be extended to incorporate the case of extreme poverty. For example, we can estimate the percentage of extremely poor households in the first period that escape extreme poverty but still remain moderately poor in the second period (joint probability) as 1 −1 ′2 2 −2 ′2 (1 < 1 2 < 2 < 2 ) = 2 ( , , ) − 1 2 1 −1 ′2 2 −2 ′2 2 ( , , ), (B.6) 1 2 where 1 and 2 stand for the extreme poverty lines in period 1 and period 2, respectively. More detailed derivations are provided in the cited studies. 26