Policy Research Working Paper 10210 Impact of High Inflation on Household Livelihoods in Urban South Sudan Alvin Etang Thierry Hounsa Utz Pape Poverty and Equity Global Practice October 2022 Policy Research Working Paper 10210 Abstract Using panel data, this paper analyzes the impact of high increasing unemployment among urban residents. Infla- inflation on household livelihoods in urban South Sudan. tion is exacerbating food insecurity and hunger, particularly Based on a difference-in-difference approach, inflation is for the poorest households who are more vulnerable to found to have a strong negative impact on urban poverty hunger. Inflation has also negatively affected households’ between 2015 and 2017, mainly driven by the increase perceptions of welfare. These changes in welfare are mostly of non-food prices. Food price inflation had a negative explained by the period of near hyper-inflation in 2017. and statistically significant impact on girls’ primary and Addressing high inflation must be at the center of efforts secondary school attendance, while proximity to school to reduce poverty and hunger to improve the welfare of the is very important for girls’ school attendance. Increases in people of South Sudan. food prices led to a decline in labor force participation, This paper is a product of the Poverty and Equity Global Practice. 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 aetangndip@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 Impact of High Inflation on Household Livelihoods in Urban South Sudan Alvin Etang, Thierry Hounsa and Utz Pape1 Keywords: Inflation, livelihoods, poverty, urban, household surveys, South Sudan JEL Codes: E31, I30, O18, C23, O55 1 Alvin Etang and Utz Pape are Senior Economists in the Poverty and Equity Global Practice of the World Bank. Thierry Hounsa is a consultant at the World Bank. Many thanks to Luca Parisotto and Ando Rahasimbelonirina for assistance with data work. Advice and comments from Nobuo Yoshida and Emmanuel Skoufias are also gratefully acknowledged. We also thank Pierella Paci for her guidance. Peer reviewer comments from Arden Finn and Nora Dihel are also gratefully acknowledged. The analysis in this paper is based on the publicly available data from three waves of the High Frequency South Sudan Survey, available on www.thepulseofsouthsudan.com. The findings, interpretations, and conclusions of this paper are those of the author and should not be attributed to the World Bank or its Executive Directors. The corresponding author Alvin Etang may be contacted at aetangndip@worldbank.org. 1. Introduction The Republic of South Sudan gained independence on July 9, 2011, following a peace agreement with the Republic of Sudan in 2005, which put an end to Africa’s longest running civil war. South Sudan is a small country with vast oil wealth, but its abysmal developmental outcomes reflect a history of conflict, characterized by a poorly functioning state and a lack of institutional services provision. Only two years after independence, civil war broke out in South Sudan and an unfavorable external macroeconomic environment triggered an economic crisis. In the years from 2015 to 2017, the South Sudanese economy displayed all the characteristics of a war economy, including severe output contraction, rapid currency devaluation, and soaring inflation. Oil dependency has tied the fate of the nation to the volatility of global commodity prices. Widespread fighting and large-scale displacement over several consecutive planting seasons have disrupted many households’ normal agricultural activities, resulting in increasingly large production deficits each year and widespread food insecurity. Compounding on this, falling international oil prices triggered the rapid devaluation of the local currency driven by pressures from a low domestic supply of foreign currency, exacerbated by concurrent high domestic demand for foreign currency due to the need to supplement domestic production shortages with imported food. Falling oil prices also meant a collapse of government revenues, which resorted to financing its deficit by printing money and incurring a growing stock of debt. Combined, these shocks have led to rapidly rising food prices, with the year-on-year CPI inflation reaching its peak at 549 percent in September 2016 (Figure A1). While the level of inflation almost reaches hyper-inflation, it remained – on an annual basis – still below the threshold of hyper-inflation. Inflation has been high, but variable, across all categories of goods and services (Figure A2). Non- food items experienced price increase between June 2015 and June 2017. However, food prices also increased substantially during this period. This is a concern since food inflation typically hurts the poor disproportionately, due to the higher share of food in the poor’s consumption basket. Given the already widespread poverty, such high food price inflation can be critical in the case of South Sudan. Although some poor rural households may be net producers of food (producing more than they consume), and thus less impacted by the high food price inflation, the very limited agricultural sector in South Sudan and the unusually high reliance on imported food suggest that the poor are also dependent on food imports, whose prices and availability have been severely affected by inflation. An important and inevitable question is how inflation is affecting household livelihoods in South Sudan, particularly the poor. 2 High inflation can have negative impacts on household livelihoods due to increased prices for consumed goods and services with lagging wage and social assistance increases. However, households that produce goods like food are usually less affected by high inflation as they are shielded from market prices. In fact, they can benefit from inflation if they sell products in the market. Also other characteristics like product types and market access can influence how much a household loses or benefits. For non-agricultural households, type of 2 In 2015 nearly 66 percent of the population in South Sudan was poor, based on the $1.90 2011 PPP poverty line (excluding Jonglei, Unity, Upper Nile, and Warrap due to insecurity), which is a considerable increase in poverty from an already high level of 52 percent in 2009. 2 employment, level of education and other factors can render households more resilient against shocks. The theoretical causes and impacts of hyperinflation are well known, and provided in the seminal work of Cagan (1956) and Nordhaus (1973). A more recent review and update was conducted by Fischer et al. (2002). A historic overview can be found in He (2017). Recent studies focus on the causes and policy options (for example, Acemoglu et al., 2003 and Reinhart & Savastano, 2003), and historic dimensions often in the context of Zimbabwe (for example, Coomer & Gstraunthaler, 2011). Given the dearth of micro-data in countries with high- or hyperinflation, only very few studies look at the direct welfare impacts of high- or hyperinflation. Fajardo & Dantas (2018) study the impact of hyperinflation on investment behavior in Brazil. However, they do not touch on welfare or livelihood impacts. Larochelle et al. (2014) use a small-area-based approach with an asset index and find that rural poverty also increased in Zimbabwe’s hyperinflation period. In contrast, Kurasha (2021) uses micro-data from several years before and after hyperinflation, but finds that rural poverty fell while urban poverty increased, while asset inequality dropped during the hyperinflationary period. Health indicators worsened for both urban and rural as well as access to electricity, safe drinking water, improved toilets and health care. In this paper, we assess the shorter-term impacts of high inflation on household livelihoods in urban South Sudan. Longitudinal micro-data for a representative sample of households is used to understand the changes in livelihoods between 2015 and 2017, accompanied by continuous price data collected across South Sudan. The novel data sets based on a set of innovative high-frequency surveys allow the use of a difference-in-difference approach providing a stronger identification than can currently be found in the literature. Furthermore, the paper identifies resilient households to draw conclusions for social protection programs and policies. The next section of this paper presents the data and methodology in more detail. Section 3 presents descriptive results followed by results from the identification on inflation impacts on urban livelihoods in South Sudan. Section 4 concludes with a discussion and policy recommendations. 2. Data and Methodology 2.1. Data This paper makes use of three waves of panel survey data from the High Frequency South Sudan Survey (HFSSS; Table 1). 3 Wave 1 of the HFSSS was conducted largely before prices exploded, while waves 2 and 3 were implemented in the period of high inflation, and wave 4 was conducted when prices had escalated. We use location- and time-specific price differences to quantify the impact of high inflation on poverty and other livelihood indicators. The panel analysis in this paper is restricted to urban households, as it aims to identify factors that make households resilient. While the restriction to urban areas limits the scope of this paper, the panel analysis allows to gain better understanding of the impact of inflation. For urban areas, waves 1, 2 and 4 provide household panel data. The panel data will be used to analyze within-household dynamics in times of high 3 The High Frequency South Sudan Survey, funded by DfID, was conducted by the World Bank in collaboration with South Sudan’s National Bureau of Statistics, to monitor welfare and perceptions of citizens in all accessible areas of South Sudan. 3 inflation. The models will be applied to changes in livelihoods and determinants of the impact mainly at the household level. Since different causes affected livelihoods in this period of instability in South Sudan, the difference-in-difference approach will identify the effect of inflation on livelihoods by correlating changes in prices with changes in livelihood indicators. The data sets contain information on security, economic conditions, education, employment, access to services, and perceptions. They also include comprehensive information on assets and consumption, to allow estimation of poverty based on the Rapid Consumption Survey methodology as detailed in Pape and Mistiaen (2018). Table 1: High Frequency South Sudan Survey (HFSSS), survey dates and coverage Data collection Geographic coverage Rural/Urban coverage dates Wave 1 February 2015 - 6 out of 10 states: Western Equatoria, Covered urban and rural September 2015 Central Equatoria, Eastern Equatoria, households Northern Bahr El Ghazal, Western Bahr El Ghazal, and Lakes state. Wave 2 February 2016 – 7 out of 10 states: wave 1 + Warrap Revisited urban June 2016 state. The other three former states households interviewed in (Jonglei, Unity, and Upper Nile) could Wave 1 not be surveyed due to security concerns. Wave 3 September 2016 - 7 out of 10 states: Same as Wave 2 Covered a new cross- March 2017 section of urban and rural households Wave 4 May 2017 - 7 out of 10 states: Same as Waves 2 and Revisited urban August 2017 3. households from Waves 1 and 2 Source: High Frequency South Sudan Survey Prices and Inflation The consumption section of the household survey (HFSSS) collects information on items’ unit prices and quantities. As with all data collected from sample surveys, the household-reported prices are subject to sampling errors. Item non-response and measurement error will also lead to biased estimates (Dahlhamer et al., 2003; Garner et al., 2009). However, household-reported prices have a key strength: knowing precisely the prices paid by households who make expenditures themselves has an advantage in that it captures the parallel exchange rates, showing households’ real purchasing power. This is particularly important in the context of South Sudan with a strong parallel exchange market. Based on the strengths and weaknesses of the three price data sources, we decided to use household-reported prices because it covers the entire sample and has price information for all items consumed by the household. Thus, for our analysis, inflation is calculated based on unit price household survey data (using Laspeyres price index). In addition to using the total inflation variable, we also break it down into food price inflation and non-food price inflation to explore which of the two might be driving the results. 4 Outcomes To analyze the impact of inflation on household livelihoods, our dependent variables are household level (and individual level) outcome indicators. The variables cover a range of household social and economic indicators, which can be calculated based on the panel data (waves 1, 2 and 4). Table 2 shows the sample for the analysis compared to the initial sample. The outcome variables are selected from the following five categories: poverty, education, labor, hunger, and perceptions of welfare (Table 3). Table 2: Sample size Initial Sample Size Sample Size for the Analysis Wave 1 3550 423 Wave 2 1189 423 Wave 4 944 423 Source: High Frequency South Sudan Survey Table 3: Outcomes variables Variable Description Poverty Poor or non-poor Whether the household is poor or not based on the $1.90 2011 PPP poverty line Consumption Household consumption expenditure in real terms Education School attendance Whether children aged between 6-13 years and between 14-18 years are currently attending school Labor 4 Labor force participation rate The ratio of the active in the labor force to the total working age population (15-64 years) Employment rate A person is employed if he/she is of working age and has engaged in one form of employment activity. 5 The employment rate is the number of persons in employment as a percentage of the total labor force. Unemployment rate A person is unemployed if he/she is of working age, is not in employment during the reference period, and has been seeking employment over the past 4 weeks. The unemployment rate is the number of persons in unemployment as a percentage of the total labor force. Outside the labor force/or A person is outside the labor force (or “inactive”) if he/she is of working-age and inactivity neither employed nor unemployed, according to the preceding definitions. An inactive person is not necessarily idle, especially in the context of a developing economy. The data breaks this group down into those who are inactive because they do household work, those who are enrolled in education, those who are discouraged, etc. Hunger 4 The labor market statistics presented in this paper follow closely the international standard set as per the International Labour Organisation’s (ILO) Key Indicators of the Labour Market (KILM). There are two key reference periods: (a) the short observation period defined as 7 days, and (b) the long observation period defined as 12 months. Following ILO guidelines, statistics are reported for the short observation period unless explicitly stated. All persons aged 15-64 are defined as being of working age. 5 The five employment activities are: (i) working as an apprentice, (ii) working on the household’s farm, raising livestock, hunting or fishing, (iii) conducting paid or commissioned work, (iv) running a business of any size for oneself or for the household, (v) helping in a household business of any size. The definition further includes persons who are temporarily absent from their work due to training or working time arrangements such as overtime leave, and paid interns. Note that the definition excludes household work. 5 Hunger How often households lacked food or lacked resources to buy food at least once in the past month Perceptions of welfare Satisfaction with life The extent to which households are satisfied with life Living conditions Households views about their present and future living conditions Economic conditions Households views about the present, past and future economic situation of South Sudan. Control over life The extent to which households feel that they have control over their life Future of South Sudan Households biggest fear about the future of South Sudan Note: The labor force refers to the sum of persons in employment and in unemployment. It is the counterpart of the group of inactive persons, i.e. the labor force plus the inactive sum up to the entire working-age population (ILO, 2013). 2.2. Model Specification To estimate the impact of inflation on household livelihoods in urban areas of South Sudan, we use a difference-in-difference (double difference) approach to exploit both the time dimension and differences in the exposure to inflation. This identification will eliminate pre-inflation differences in the outcome variable and controls for anything that also changes over time and affects both groups. Hence, the assumption will be made that changes in outcomes from households in areas with high and low inflation would have been the same in the absence of the inflation shock: ̂ 1DD = ( � 0H) – ( � 1H – � 0 L) �1L – (1) More specifically, the difference-in-difference estimator β1 is computed by comparing the first- differenced values of the outcome for the high- (H) and low-inflation (L) groups. Hence, the outcome differences for the low-inflation group are differenced from the high-inflation group after taking the simple difference, which gives us the difference-in-difference estimate. The purpose of a difference-in-difference approach is to analyze whether the estimate β1 is statistically and significantly different from zero. To estimate the difference-in-difference effect, we use an ordinary least squares (OLS) regression model including the control vector: = β0 + β1 (post) + β2 (postt *inflationst )+ βX + γs +δt + (2) where is an outcome measured for the individual or household i living in Boma s at time t ; postt is a binary variable indicating time period t (pre- or post-inflation); Postt = 1 for each of waves 2 and 4 and zero otherwise (i.e. we treat waves 2 and 4 as having occurred at different times, with wave 1 being the reference period); and inflations is a continuous variable measuring the inflation rate of the Boma s. Inflation is computed as the first difference of the log price index at the Boma level. To avoid an omitted variable bias (as there are other confounding factors affecting the given outcome variables besides time-period and exposure to inflation), a control vector for household i living in Boma s at time t is introduced; γs and δt are respectively the Boma fixed 6 effects and the time fixed effects. Standard errors will be clustered at the Boma level to allow for within-cluster correlation. 6 β1 is the difference-in difference estimator. Household Resilience: Triple Difference To identify factors that make households resilient to the inflation shock, we estimate the following triple difference equation where hi is a potential resilience factor. = β0 + β1 (post) + β2 (postt *inflationst) + β3 (postt *hi) + β4 (postt *inflationst *hi) + βX + γs +δt + (3) β3 is the triple difference estimator. In this triple difference setting, β1 is the diff-in-diff estimate for the reference group (h=0). It captures average differential change in y from the pre- to post- treatment period for the reference group in the treatment group relative to the change in y for the reference group (h=0) in the untreated group. The total treatment effect for both groups is β1 + β3. Conflict indicator Given the ongoing conflict in South Sudan, conflict will likely be one of the confounding factors affecting household livelihoods. We control for this by including a conflict variable in the regressions. We construct an exogenous conflict variable based on conflict event data from the Armed Conflict Location & Event Data (ACLED) 7 for the period of our study. The data set codes the exact location of all political violence incidents that were reported during this time. We use proximity to a deadly conflict event to generate a continuous conflict exposure variable (i.e. the number of fatalities). 6 Default standard errors can greatly overstate estimator precision. Instead, if the number of clusters is large, statistical inference after OLS should be based on cluster-robust standard errors. Failure to control for within-cluster error correlation can lead to very misleadingly small standard errors, and consequent misleadingly narrow confidence intervals, large t-statistics and low p-values (Cameron and Miller, 2015). 7 Information about ACLED methodology can be found at https://www.acleddata.com/. 7 3. Results Price Changes Consistent with the CPI, the HPI shows that inflation exploded between 2015 and 2017. The mean price index increased substantially from 1.31 in 2015 to 8.07 in 2016 and exploded to 30.75 in 2017 (Figure 1). The results show that non-food price inflation is drastically higher than food price inflation during this period. We use an adjusted Wald test to test the significance of the differences in the means across the three years. The test confirms that the differences are statistically significant (p<0.01), meaning that inflation has increased significantly over time. Figure 1: Recent trends in price index 60.00 40.00 20.00 0.00 Food price index Non Food price index Total price index 2015 2016 2017 Source: Authors’ calculations based on HFS market prices data The increase in prices varies across states and have been much higher in the Eastern Equatorial, Central Equatorial, and Northern Bahr El Ghazal states than in other states (Figure 2). This is mainly driven by the fact that food price inflation increased much more in these states. Since the price developments have been so different across the country, it is important to consider this significant geographic variation when analyzing the impact of inflation on livelihoods. Figure 2: Inflation by state - 2017 (Base year 2015=100) 500% 400% 300% 200% 100% 0% Northern Bahr El Western Bahr El Lakes Western Equatoria Central Equatoria Eastern Equatoria Ghazal Ghazal Food inflation Non Food inflation Total inflation Source: Authors’ calculations based on HFS market prices data 8 Poverty and Consumption We find strong evidence of the negative effect of inflation on household consumption and poverty. 8 The inflation impact is entirely driven by non-food price inflation. Poverty increases with the severity of exposure to inflation by the households. If inflation increases by 1 percent, the share of poor urban population (living below USD 1.90 per day PPP) increases by 0.353 percent. The results also show that the conflict variable is associated with slight decreases in consumption, with a marginal impact on poverty. We also observe that being employed itself does not matter for poverty reduction, probably due to the low urban unemployment rates. Households whose heads are employed in the services sector are less likely to be poor compared to those in the agriculture sector. This probably reflects higher wages for those in the services sector. Some common findings similar to other poor countries are also observed here: household head being a female, and larger household size are associated with low consumption and the likelihood of being poor. On the other hand, education and land ownership help to reduce poverty. We also run a regression that includes an interaction term between inflation and land ownership. The coefficient on the interaction term is not statistically significant for both food price inflation and non-food price inflation. This suggests that the impact of inflation on poverty and real consumption is the same for both households that do own land and those that do not. Finally, university education increases consumption and reduces poverty. The impact of inflation on real consumption is significantly less for households whose heads do have university education than those who do not (the coefficient on the interaction term is 0.985, p<0.01). No significant differences in education level exist when it comes to the inflation impact on poverty. School Attendance About 3 in 4 South Sudanese children were attending primary school in 2015. 9 The primary school attendance rate remained stable in 2016 and increased to 80 percent in 2017 (Figure A3). Secondary school attendance remained stable from 2015 and 2016 at 78 percent but increased to 81 percent in 2017 (Figure A4). Primary school attendance for boys and girls increased at about the same rate between 2015 and 2017 (Figure A5). For the older children, attendance rate for boys declined between 2015 and 2017 by 9 percentage points (Figure A6). The opposite is true for girls’ attendance, which has slightly increased during this period. Perhaps older boys are dropping out of school to join the labor force. This is plausible as children of working age can be expected to join the workforce to help the household support its livelihood during times of economic hardship (World Bank, 2017). However, there is no evidence that this is already happening at a large scale in the states covered by the survey. This is because the difference in school attendance rate of boys aged 14 to 18 between 2015 and 2017 is not significant in a statistical sense. Food price inflation had a negative and statistically significant impact on girls’ school attendance (but no effect on boys). For girls, the likelihood of attendance diminishes with a rise in food prices. The distance to the nearest school is also important for school attendance. The chances of girls attending school diminish with increases in the distance they would have to walk to the nearest school. We run a regression for girls’ attendance that includes an interaction term between inflation and distance to school. The coefficient on the interaction term is statistically significant and 8 The complete difference-in-difference estimates are presented in the appendix (Tables A1-A7). A summary of the regression results of the impact of inflation (β1 coefficients, difference-in difference) is provided in Table 4. 9 Attendance rates for children of primary school age (6-13) and secondary school age (14-18) are reported. 9 negative (-0.712; p<0.01). This means that the impact of food price inflation on school attendance is greater for girls who take more than 5 hours to walk (one way) to the nearest school from their homes compared to girls who take less than 30 minutes to do so. One explanation for this result is that when faced with an economic shock such as inflation, households become poorer (as noted above), and tend to sacrifice the education of their female children whose schools are far away from their homes as they may not be able to afford the costs related to living far away from school. In this regard, bringing schools closer to households will help to mitigate the adverse impact of inflation on girls’ school attendance. The results also suggest that school attendance increases if the household head is a woman and has secondary or university education. Designing programs to promote female education will help to improve education outcomes in general, and for girls in particular. Labor South Sudan’s economic instability led many of working age to drop out of the labor force between 2015 and 2016. During this period, the urban labor force participation rate dropped significantly from about one-half to about one-third (Figure A7). In 2015, the labor force participation rate remained relatively similar between poor and non-poor households and across expenditure quintiles. There are no significant differences between men and women in labor force participation in both years. 10 The urban unemployment rate was 8 percent in 2015 and 7 percent in 2016 (Figure A8). This number reduced substantially to 3 percent in 2017. It may not be very surprising to observe relatively high employment rates because like many other poor countries, South Sudan lacks social safety nets, which forces unemployed individuals to seek employment. Similar to school attendance, food price inflation has a strong impact on the labor market. Increasing food prices leads to a decrease in labor force participation and increasing unemployment. In urban areas, education level is a strong determinant of unemployment for both men and women. Hunger Between 2015 and 2016, hunger incidence deteriorated severely for households in the poorest quintile, with the likelihood of experiencing hunger ‘often’ (more than 10 times per month) increasing from 4 percent to 10 percent (Figure A10, p<0.05). Due to rising prices without compensatory income increases, especially the wage-dependent urban population lost real purchasing power. Food insecurity and hunger remain a serious issue for South Sudan. For the poorest households, the likelihood of experiencing hunger ‘sometimes’ (3-10 times per month) has been reducing from 38 percent in 2015 to 29 percent in 2016 and rising to 40 percent in 2017. This confirms that the poorest households are more vulnerable to hunger than richer households in the face of rising food prices. Richer households are much more likely to adjust their diets to cope with a lack of food, while the poorest households cope with a lack of food by going entire days without eating. This may pose serious health issues, and affect children’s education outcomes, with both short-term and long-term adverse effects on poverty. Resorting to more moderate strategies, compared with the poorest households, those in the top 4 poverty quintiles are more likely to deal 10 The 2017 labor force participation numbers are not entirely comparable with the previous years because of the changes in the questionnaire. 10 with a lack of food by reducing the number of meals or portion size, or consuming less preferred foods. Inflation increases hunger, and the combined effect of both food price inflation and non-food price inflation is very strong, with rising food prices having greater impact. While the pinch of inflation was felt by every household, the poorest ones were the worst affected. Rising food prices have led to growing food insecurity for the poorest households, for whom the incidence of hunger has increased sharply. The poorest households are in a vicious circle as they may become poorer due to the consequences of hunger, including poor health, child malnutrition and education outcomes. The finding that rapidly rising food prices is a causal source of hunger and food insecurity is consistent with findings from other poor countries (Jolliffe et al., 2016). Households whose heads have university education experience less hunger than others. The coefficients on the poverty quintiles are significant and negative and get larger in magnitude as one moves up the consumption distribution from Q2 to Q3, Q4 and largest for the richest quintile. This suggests that poverty also has a significant impact on hunger, with hunger incidence declining as consumption increases. Perception The deterioration of economic conditions in South Sudan, as indicated by continued high inflation, is well echoed by households' perceptions. In 2017, almost all (97 percent) South Sudanese residing in urban areas felt that economic conditions in their country were bad or very bad (Figure A11). This is a drastic increase compared to the previous years. This figure stood at almost two- thirds of urban South Sudanese in 2015, increasing to almost 9 in 10 (63 vs. 86 percent respectively, p<0.001). However, there seems to be a growing sense of optimism among the urban residents about the future. In 2015, nearly half of households (46 percent) believed that economic conditions will be better or much better in 3 months’ time. While people became less optimistic in 2016 (29 percent), the figure increased considerably to two-thirds in 2017 (66 percent). It should also be noted that a sizable share of residents remains pessimistic about the future, with 21 percent of households of the view that economic conditions will get worse or much worse in 3 months’ time. The deterioration of economic conditions is also well reflected in households’ perceptions of their own living conditions. In 2015, almost half (45 percent) of urban households felt that their living conditions were fairly bad or very bad (Figure A12). This figure increased significantly to 78 percent in 2016 and 80 percent in 2017. There does not seem to be much hope for many households who believed that their personal living conditions will deteriorate in the next 3 months. Between 2015 and 2016, the share of households that believe living conditions will get worse or much worse increased significantly from 25 percent to 46 percent (p<0.001). The figure decreased to 31 percent in 2017, though it is still high. Nevertheless, as with economic conditions, there seems to be growing optimism about the future for one-half of urban households. The share of households believing that living conditions will get better or much better increased from 39 percent in 2016 (10 percentage points less, compared 2015) to 53 percent in 2017. There seems to be a correlation between people having control over their lives and the extent to which they are satisfied with life. The share of urban residents who felt that they have no control over their lives increased from 26 percent in 2015 to 37 percent in 2016 but decreased substantially to 16 percent in 2017 (Figure A13). Feeling much more in control of their lives, 32 percent of households strongly agreed that they are satisfied with life (Figure A14). Note that in 2016 there 11 was a general decline in life satisfaction relative to 2015, which reflected a growing feeling among urban folks that they were powerless in the face of deteriorating political and economic conditions. Increases in inflation are associated with less satisfaction with life. Regarding satisfaction with present living conditions (ranging from 1: very good to 5: very bad), the positive coefficient means that people are less satisfied in the face of inflation. However, people’s views about future living conditions are positive, consistent with optimism noted from descriptive statistics. There is a strong feeling among urban residents who are exposed to inflation that they are powerless and have no control over their lives. Table 4: Summary of regression results for each outcome indicator and inflation variable Total Inflation Food price Non-food price inflation inflation Outcomes Poverty Poor (below USD 1.90 PPP) 0.252 0.00854 0.238* Log(real consumption) -0.829*** -0.116 -0.680*** Education Currently attending school (Girls) -0.0767 -0.136*** -0.0202 Labor Labor force participation: Active -0.138 -0.207*** -0.049 Unemployed 0.0841 0.126** 0.042 Hunger Hunger incidence 0.430*** 0.329** 0.193* Perceptions of welfare Life satisfaction -1.205* -0.178 -0.807* Present living conditions 0.480** 0.22 0.22 Future living conditions 1.789* -0.039 1.343** Control over own life -0.611** -0.055 -0.514** Present economic conditions 0.264 0.394** 0.053 Future economic conditions 1.370 -0.588 1.22* *** p<0.01, ** p<0.05, * p<0.1. Robustness We perform robust checks of the results by running various regressions. They include regressions for wave 1 to wave 2, wave 1 to wave 4, using household fixed effects, and adding the Boma CPI in the regression to pick up the effect of high prices and high poverty. We also run regressions where households are classified as high and low inflation households using the mean inflation for urban South Sudan. All households living in an area with an inflation rate higher than the mean inflation rate are classified as “high inflation” households and are therefore in the treatment group. Households with an inflation rate lower than the mean inflation rate are in the control group. The results, presented in the appendix, in Table A8, confirm the findings discussed above related to the impact of inflation on welfare in urban South Sudan. The results are generally consistent with initial specifications regardless of whether we use Boma CPI or household fixed effects or classify households as experiencing high or low inflation. Perhaps the most interesting revelation of these analyses is that the welfare impact comes from the “hyperinflation” experienced in South 12 Sudan during wave 4 in 2017 rather than inflation between waves 1 and 2 in 2016. The coefficients for the regression model for waves 1 and 2 are generally not statistically significant, while those for the regression for waves 1 and 4 are significant with the initial specifications above. This is consistent with the inflation dynamics between waves 1 and 2 and waves 1 and 4 and changes in welfare indicators (see Figure A16). 4. CONCLUSIONS AND POLICY RECOMMENDATIONS The high inflation in South Sudan continues to put many households under extreme financial stress, as they face increased prices without compensatory increases in income. This paper contributes to the available empirical evidence on micro-level impacts of inflation by analyzing the impact of high inflation on household livelihoods in urban South Sudan. We use panel data collected during the High Frequency South Sudan Survey (HFSSS) waves in 2015, 2016 and 2017. Increasing resilience to high inflation would allow to shift the inflation crisis management paradigm from a humanitarian to a development approach. Breaking down inflation by food price inflation and non-food price inflation reveals that the latter increased drastically more than the former during this period. Although there have been significant increases in both food and non-food prices, the observed high inflation in South Sudan is mostly driven by the escalation of non-food prices. Inflation negatively impacted various household livelihood indicators related to poverty, education, labor, hunger and perceptions of welfare. Inflation had a strong negative impact on urban poverty between 2015 and 2017, mainly driven by the increase of non-food prices. The loss of purchasing power of wages and salaries has driven many of the South Sudanese residing in urban areas into poverty. Continuous increases in inflation will only worsen the already high poverty situation. Addressing the issue of high inflation must be at the center of efforts to achieve stability the economy and reduce poverty in South Sudan. In addition, higher education has a key role for poverty reduction because the impact of inflation on consumption is significantly lower for households whose heads do have university education. Food price inflation had a negative and statistically significant impact on girls’ primary and secondary school attendance. The probability of a girl attending school diminishes as food prices increase. Proximity to school is very important for school attendance. School attendance is less likely for girls who take more than 5 hours to walk from their home to the nearest school, compared to girls who take less than 30 minutes to do so. This corroborates earlier reports that long distance to school was one of the most cited reasons for dropping out of primary and secondary school in South Sudan (EMIS, 2016). While the cost of schooling is a major constraint for school attendance of both boys and girls, it disproportionately affects girls. In the face of limited resources, parents apparently prioritize boys for schooling over girls (World Bank, 2018b). Investing in female education is very important for poverty reduction and development, especially as we also find that school attendance increases if the household head is a woman and has secondary or university education. One important policy implication from this study is that bringing schools closer to households will help to mitigate the adverse impact of inflation especially on girls’ school attendance. Investing in education, particularly in fragile contexts like South Sudan, also helps to create resilience against such economic shocks. 13 Another consequence of the observed increases in food prices is a significant decline in labor force participation and a surge in unemployment among urban people. Employment programs with a focus on poverty reduction should, therefore, consider ways to mitigate the impact of rising food prices. Inflation is exacerbating food insecurity and hunger, particularly for the poorest households who are more vulnerable to hunger. Households adopt various strategies to cope with hunger, including eating less preferred food, going entire days without eating, and selling assets. However, these coping strategies may put them at increased risk for future spells of food insecurity. The coping strategies employed by the poor, especially selling productive assets such as livestock, typically put them at an even greater disadvantage in the future (Barrett 2002). Inflation has negatively affected households’ perceptions of welfare. Urban residents who are exposed to inflation strongly feel that they are powerless and have no control over their lives. This has led to less satisfaction with life and present living conditions. A key economic priority for the Government of South Sudan should be to implement urgent macroeconomic measures to reduce high inflation. Addressing the problem of high inflation will help to curb increasing poverty, and it is crucial for progress towards achieving the first Sustainable Development Goal (SDG 1) to end poverty by 2030. In addition, for South Sudan to achieve SDG 2 (to end hunger and ensure access to food by all people, including the poor, by 2030), the issue of rising inflation must be contained very quickly as it is exacerbating hunger and food insecurity. This paper shows that inflation has had adverse effects on the livelihoods of urban households. Because our analysis focuses on urban areas, some of the results may not be directly generalized to the entire country. Even if the inflation crisis improves livelihoods of the – predominantly rural – households producing food, the rapidly increasing prices of non-food items is likely to have increased rural poverty and hunger as well. 14 REFERENCES Acemoglu, D.; Johnson, S.; Robinson, J.; Thaicharoen, Y. (2003). Institutional causes, macroeconomic symptoms: volatility, crises and growth. Journal of Monetary Economics, Volume 50, Issue 1. Barrett, C. 2002. “Food security and food assistance programs” in Handbook of Agricultural Economics. Elsevier. 2(2): 2103-2190. Cagan, Philip, 1956, “The Monetary Dynamics of Hyperinflation,” in Studies in the Quantity Theory of Money, ed. by Milton Friedman (Chicago: University of Chicago Press), pp. 25–117. Cameron, A. C. & Miller, D. L. (2015). A Practitioner’s Guide to Cluster-Robust Inference. Journal of Human Resources, University of Wisconsin Press, vol. 50(2), pages 317-372. Coomer, J.; Gstraunthaler, T. (2011). THe Hyperinflation in Zimbabwe. Quarterly Journal of Austrian Economics, Vol. 14 (3). Dahlhamer J., Dixon J., Doyle P., Eargle J., Griffin D., and McGovern P. (2003). Quality at the Item Level: Terms, Methods, and Guidelines for Cross-Survey Comparisons. Unpublished manuscript, available from the Bureau of Labor Statistics, Washington, DC. Education Management and Information systems (EMIS). (2016). Ministry of General Education and Instruction (MoGEI) Fajardo, J.; Dantas, M. (2018). Understanding the impact of severe hyperinflation experience on current household investment behavior, Journal of Behavioral and Experimental Finance, Volume 17. Fischer, Stanley, Ratna Sahay, and Carlos A. Vegh, 2002, “Modern Hyper- and High Inflations,” Journal of Economic Literature, Vol. 40 (September), pp. 837–80. Garner T., McClelland R. and Passero W. (2009). Strengths and Weaknesses of the Consumer Expenditure Survey from a BLS Perspective. Unpublished manuscript, available from the Bureau of Labor Statistics, Washington, DC. GoSS, 2011: South Sudan Development Strategy, 2011-13 – Realising freedom, equality, justice, peace and prosperity for all. Ministry of Finance and Economic Planning, Juba, August 2011. He, L. (2017). Hyperinflation: A world history. Routledge. Heltberg, R., N. Hossain and A. Reva. 2012. “Living Through Crises. How the Food, Fuel and Financial Shocks Affect the Poor.” World Bank.Washington, DC Jolliffe D., Seff I. and Fuente A. (2016). Food Insecurity and Rising Food Prices: What do we learn from experiential measures? Draft Manuscript. 15 Kurasha, F.M.N. (2021). Wealth, health, and inequality: a study of hyperinflation and households in Zimbabwe. Faculty of Commerce, School of Economics. Catherine Larochelle, Jeffrey Alwang, Nelson Taruvinga (2014). Inter-temporal Changes in Well-being During Conditions of Hyperinflation: Evidence from Zimbabwe, Journal of African Economies, Volume 23, Issue 2. Malin A (2016). Inflation and Its Influence of Households Livelihoods in Mogadishu – Somalia. European Journal of Business and Management, 8(7). Narayan, Deepa, Raj Pattel, Kai Schafft, Anne Rademacher, and Sarah Koch-Schulte. 2002. “Voices of the Poor. Can Anyone Hear Us?” Oxford: Oxford University Press. Nordhaus, W.D. (1973). The Effect of Inflation on the Distribution of Economic Welfare. Journal of Money, Credit and Banking, Vol. 5 (1). Pape, Utz Johann; Mistiaen, Johan A.. (2018). Household expenditure and poverty measures in 60 minutes: a new approach with results from Mogadishu (English). Policy Research working paper; no. WPS 8430. Washington, D.C. : World Bank Group. Pape, U., Parisotto, L. (2019). Estimating Poverty in a Fragile Context - The High Frequency Survey in South Sudan. Policy Research Working Paper, 8722. World Bank. Reinhart, C., & A. Savastano, M. (2003). The Realities of Modern Hyperinflation, Finance & Development, 40(002). Republic of South Sudan Inclusive Growth Country Economic Memorandum, World Bank (2017). Rozanski Alan, James A. Blumenthal and Jay Kaplan. 1999. “Impact of Psychological Factors on the Pathogenesis of Cardiovascular Disease and Implications for Therapy.” Journal of the American Heart Association. Daniel J Brotman, Sherita H Golden, Ilan S Wittstein. 2007. The Cardiovascular Toll of Stress. The Lancet: Vol. 370: 1089–100. SSCCSE (2009). Southern Sudan - National Baseline Household Survey. Southern Sudan Centre for Census, Statistics and Evaluation, Juba. WFP (2011). Annual Needs and Livelihoods Analysis (ANLA) – Southern Sudan. ANLA Technical Working Group, World Food Program (WFP) Juba, January 2011. World Bank Briefing Note (2015). The Fiscal Impact of Declining Oil Prices on South Sudan. World Bank (January 2015). World Bank (2016). HFS Market Surveys in South Sudan (2012 to 2016) World Bank. 2017. South Sudan Economic Update: Taming the tides of high inflation. Policy options for South Sudan. https://openknowledge.worldbank.org/handle/10986/28560. World Bank Macro Poverty Outlook (2018). South Sudan Macro Poverty Outlook. World Bank (Spring 2018). 16 World Bank South Sudan Education Note (2018b). Proposed Engagement Strategy and Nest Steps. Draft April 30, 2018. 17 APPENDIX Figure A1: Trends in CPI inflation, year-on-year 600 500 400 Percent 300 200 100 0 May June July Aug Sep Oct Nov Dec Jan Feb Mar Apr May June July Aug Sep Oct Nov Dec Jan Feb 2015 2015 2015 2015 2015 2015 2015 2015 2016 2016 2016 2016 2016 2016 2016 2016 2016 2016 2016 2016 2017 2017 Source: South Sudan National Bureau of Statistics 11 Figure A2: High inflation in all categories of goods between June 2015 and June 2017 Communication Clothing and footwear Restaurants and hotels Health Miscellaneous goods and services Food and Nonalcholic beverages Furnishings, household equipment Housing, Water,Electricity, Gas and other fuel Alcoholic beverages and Tobacco Recreation and culture Transport Education 0 500 1000 1500 2000 2500 Percent Source: Authors’ calculations based on CPI data collected by South Sudan National Bureau of Statistics 11 http://www.ssnbs.org/cpi/ 18 Figure A3: School attendance, children aged 6-13, by poverty Figure A4: School attendance, children aged 14-18, by poverty 2017 2017 Total Total 2016 2016 2015 2015 2017 2017 Richest Richest 2016 2016 2015 2015 2017 2017 2016 2016 Q4 Q4 2015 2015 2017 2017 2016 2016 Q3 Q3 2015 2015 2017 2017 2016 2016 Q2 Q2 2015 2015 2017 2017 Poorest Poorest 2016 2016 2015 2015 2017 2017 Poor Poor Non Non 2016 2016 2015 2015 2017 Poor 2017 Poor 2016 2016 2015 2015 0 20 40 60 80 100 0 20 40 60 80 100 Percent Percent Attending Not Attending Attending Not Attending Source: Authors’ calculations based on HFSSS 2015, 2016 and 2017 Figure A5: School attendance, children aged 6-13, by gender Figure A6: School attendance, children aged 14-18, by gender 2017 2017 Total Total 2016 2016 2015 2015 2017 2017 Girls Girls 2016 2016 2015 2015 2017 2017 Boys Boys 2016 2016 2015 2015 0 20 40 60 80 100 0 20 40 60 80 100 Percent Percent Attending Not Attending Attending Not Attending Source: Authors’ calculations based on HFSSS 2015, 2016 and 2017 19 Figure A7: Labor force participation rate 2015 2016 Total Total Men Men Women Women Richest Richest Q4 Q4 Q3 Q3 Q2 Q2 Poorest Poorest Non Poor Non Poor Poor Poor 0 20 40 60 80 100 0 20 40 60 80 100 Percent Percent In the labor force / active In the labor force / active Outside the labor force / inactive Outside the labor force / inactive Source: Authors’ calculations based on HFSSS 2015 and 2016 Figure A8: Employment and enrollment status 2015 2016 Total Total Men Men Women Women Richest Richest Q4 Q4 Q3 Q3 Q2 Q2 Poorest Poorest Non Poor Non Poor Poor Poor 0 20 40 60 80 100 0 20 40 60 80 100 Percent Percent Employed Employed and Enrolled Unemployed Employed Employed and Enrolled Unemployed 20 2017 Total Men Women Richest Q4 Q3 Q2 Poorest Non Poor Poor 0 20 40 60 80 100 Percent Employed Employed and Enrolled Unemployed Source: Authors’ calculations based on HFSSS 2015, 2016 and 2017 Figure A9: Employment by type 2015 Total Men Women Richest Q4 Q3 Q2 Poorest Non Poor Poor 0 20 40 60 80 100 Percent 2016 Total Men Women Richest Q4 Q3 Q2 Poorest Non Poor Poor 0 20 40 60 80 100 Percent 21 2017 Total Men Women Richest Q4 Q3 Q2 Poorest Non Poor Poor 0 20 40 60 80 100 Percent Salaried labour or labour paid in kind Run a non-farm business Help in any kind of non-farm business Apprenticeship Farming or hunting or fishing at own account Source: Authors’ calculations based on HFSSS 2015, 2016 and 2017 Figure A10: Hunger incidence over the past 4 weeks 100 80 60 40 20 0 Poorest Richest Poorest Richest Poorest Richest Q2 Q3 Q4 Q2 Q3 Q4 Q2 Q3 Q4 2015 2016 2017 Never Rarely (1-2 times) Sometimes (3-10 times) Often (more than 10 times) Source: Authors’ calculations based on HFSSS 2015, 2016 and 2017 22 Figure A11: Perception of economic conditions economic conditions(3m 2017 economic Future onths) 2016 Very good/Much better 2015 Fairly good/Better 2017 Neither good nor bad/Same conditions Present 2016 Fairly bad/Worse 2015 Very bad/Much worse 0 20 40 60 80 100 Percent Source: Authors’ calculations based on HFSSS 2015, 2016 and 2017 Figure A12: Perception of living conditions 2017 Present living conditions (3 Future living months) 2016 Very good/Much better 2015 Fairly good/Better 2017 conditions Neither good nor bad/The same 2016 Fairly bad/Worse 2015 Very bad/Much worse 0 20 40 60 80 100 Percent Source: Authors’ calculations based on HFSSS 2015, 2016 and 2017 Figure A13: Feeling in control over own life 100 80 Percent 60 40 20 0 2015 2016 2017 No Control Some Control A Great Deal of Control Complete Control Source: Authors’ calculations based on HFSSS 2015, 2016 and 2017 Figure A14: Satisfaction with life 23 100 Percent 80 60 40 20 0 Strongly Disagree Slightly Neither Slightly Agree Strongly disagree disagree agree nor agree Agree disgree 2015 2016 2017 Source: Authors’ calculations based on HFSSS 2015, 2016 and 2017 Figure A15: Fear for the future of South Sudan Biggest fear, future of South Sudan 100 Other reasons 80 Bad economy Percent 60 Health and disease problems 40 20 Poverty 0 Lack of economic opportunities 2015 2016 2017 Source: Authors’ calculations based on HFSSS 2015, 2016 and 2017 Table A1: Regression results for poverty and consumption (1) (2) (3) (4) Poor Poor (below (below USD 1.90 USD 1.90 Log(real Log(real VARIABLES PPP) PPP) consumption) consumption) Survey year: 2016 -0.286 0.031 -0.524 -0.934 (0.268) (0.259) (0.449) (0.611) Survey year: 2017 -0.051 0.260 -2.541*** -2.916*** (0.228) (0.221) (0.344) (0.455) Inflation*Post 0.252 0.060 -0.829*** -0.575* (0.165) (0.161) (0.257) (0.338) Conflict 0.001 0.002 -0.005** -0.006** (0.001) (0.001) (0.002) (0.003) Female household head 0.149*** 0.203*** -0.030 -0.135* (0.051) (0.065) (0.050) (0.080) Education level of household head_Primary -0.077 -0.081 0.066 0.120 24 (0.088) (0.103) (0.069) (0.068) Education level of household head_Secondary -0.061 -0.067 0.137 0.186 (0.123) (0.123) (0.090) (0.090) Education level of household head_University -0.408*** -0.416*** 0.619*** 0.596*** (0.120) (0.129) (0.148) (0.182) Education level of household_Other -0.194*** -0.273*** 0.387*** 0.569*** (0.097) (0.084) (0.060) (0.053) Land ownership -0.152** -0.132** 0.263** 0.256*** (0.103) (0.092) (0.068) (0.058) Household head age -0.003 -0.002 0.036 0.025 (0.029) (0.019) (0.020) (0.008) Household head age-squared 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) Household size 0.077*** 0.091*** -0.175*** -0.182*** (0.022) (0.024) (0.013) (0.013) Household size-squared -0.001*** -0.002*** 0.004*** 0.004*** (0.001) (0.001) (0.000) (0.001) Household head Unemployed -0.003 -0.055 (0.145) (0.134) Household head employment: Manufacturing 0.201 -0.125 (0.123) (0.137) Household head employment: Services -0.137* 0.183** (0.087) (0.077) Household head employment: Education -0.012 -0.112 (0.169) (0.157) Household head employment: Defense/Security -0.125 0.213** (0.091) (0.098) Household head employment: Public -0.323* 0.249 Administration (0.165) (0.159) Constant -0.131 -0.247 1.897** 2.211*** (0.466) (0.274) (0.706) (0.498) Observations 703 673 703 673 R-squared 0.296 0.353 0.860 0.868 *** p<0.01, ** p<0.05, * p<0.1. Notes: Robust standard errors in parentheses. Reference is 2015 for survey year; male for gender of household head; no education for household head educational level; employed for household head employment status; and agriculture for household head’s sector of employment. 25 Table A2: Regression results for poverty and consumption, interacting inflation with household head university education (1) (2) (3) (4) Poor (below Poor (below Log(real Log(real VARIABLES USD 1.90 PPP) USD 1.90 PPP) consumption) consumption) Survey year: 2016 -0.246 -0.117 0.086 0.028 (0.260) (0.216) (0.365) (0.295) Survey year: 2017 0.000 0.147 -2.043*** -2.133*** (0.220) (0.192) (0.325) (0.279) Inflation*Post 0.232 0.160 -1.185*** -1.162*** (0.153) (0.136) (0.216) (0.184) inflation*Household head_University education *Post 0.036 -0.216 0.815** 1.210*** (0.247) (0.230) (0.305) (0.268) Household head_University education*Post -0.095 0.184 -1.193*** -1.523*** (0.247) (0.218) (0.305) (0.253) Conflict 0.001 0.002 -0.004*** -0.005** (0.001) (0.001) (0.001) (0.002) Female household head 0.172*** 0.233*** -0.085* -0.192** (0.038) (0.070) (0.048) (0.092) Household head University education -0.332 -0.286 0.572* 0.374 (0.245) (0.184) (0.333) (0.241) Land Ownership -0.153** -0.127** 0.229** 0.214** (0.069) (0.057) (0.092) (0.083) Household head age -0.003 4.90e-05 0.033 0.021 (0.019) (0.008) (0.028) (0.018) Household head age-squared 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) Household size 0.077*** 0.090*** -0.168*** -0.177*** (0.013) (0.012) (0.021) (0.021) Household size-squared -0.001*** -0.002*** 0.004*** 0.004*** (0.000) (0.000) (0.001) (0.001) Household head Unemployed -0.0165 -0.0460 (0.150) (0.140) Household head employment: Manufacturing 0.173 -0.028 (0.114) (0.124) Household head employment: Services -0.148** 0.207** (0.073) (0.078) Household head employment: Education -0.028 -0.045 (0.162) (0.139) Household head employment: Defense/Security -0.124 0.203** (0.096) (0.091) Household head employment: Public Administration -0.352** 0.355*** (0.158) (0.123) Constant -0.193 -0.344 2.007*** 2.367*** (0.454) (0.266) (0.684) (0.476) Observations 703 673 703 673 R-squared 0.293 0.353 0.863 0.876 *** p<0.01, ** p<0.05, * p<0.1. Notes: Robust standard errors in parentheses. 26 Table A3: Regression results for currently attending school, boys and girls (1) (2) (3) Currently Currently Currently Attending Attending Attending VARIABLES School School School Survey year: 2016 0.067 0.0739** 0.003 (0.125) (0.032) (0.127) Survey year: 2017 0.005 0.099 -0.041 (0.106) (0.071) (0.078) Inflation*Post -0.035 (0.071) Inflation*distance to school more than 5 hours*Post 0.0664* (0.035) Food Inflation*Post -0.0865** (0.038) Food inflation* distance to school more than 5 hours*Post 0.186** (0.085) Non-food Inflation*Post 0.001 (0.053) Non-food inflation* distance to school more than 5 hours*Post 0.0447* (0.024) Distance to school: More than 5 hours (Ref: Less than 30 minutes) -0.276*** -0.279*** -0.272*** (0.020) (0.018) (0.020) Gender of school child: Girl -0.116*** -0.115*** -0.116*** (0.019) (0.019) (0.019) Female household head 0.031 0.029 0.031 (0.037) (0.037) (0.037) Education level of household head_Primary -0.036 -0.035 -0.036 (0.103) (0.103) (0.104) Education level of household head_Secondary 0.111** 0.108** 0.110** (0.045) (0.044) (0.045) Education level of household head_University 0.117* 0.120** 0.116* (0.059) (0.060) (0.058) Education level of household_Other -0.504*** -0.478*** -0.507*** (0.046) (0.043) (0.049) Constant 0.480*** 0.482*** 0.481*** (0.055) (0.055) (0.054) Observations 6435 6435 6435 R-squared 0.060 0.062 0.060 *** p<0.01, ** p<0.05, * p<0.1. Notes: Robust standard errors in parentheses. 27 Table A4: Regression results for currently attending school, girls only (1) (2) (3) Currently Currently Currently Attending VARIABLES Attending School Attending School School Survey year: 2016 0.150 0.123*** 0.063 (0.150) (0.037) (0.151) Survey year: 2017 0.043 0.162** -0.032 (0.124) (0.075) (0.090) Inflation*Post -0.077 (0.086) Inflation*distance to school more than 5 hours*Post -0.202*** (0.030) Food Inflation*Post -0.136*** (0.041) Food inflation* distance to school more than 5 hours*Post -0.745*** (0.063) Non-food Inflation*Post -0.020 (0.063) Non-food inflation* distance to school more than 5 hours*Post -0.159*** (0.020) Conflict 0.000 0.000 0.000 (0.000) (0.000) (0.000) Distance to school: More than 5 hours (Ref: < 30 minutes) 0.103*** 0.0992*** 0.110*** (0.034) (0.032) (0.036) Female household head -0.0511** -0.0538** -0.0513** (0.023) (0.023) (0.023) Education level of household head_Primary -0.075 -0.074 -0.076 (0.086) (0.085) (0.086) Education level of household head_Secondary 0.080 0.076 0.078 (0.062) (0.060) (0.062) Education level of household head_University 0.086 0.088 0.082 (0.058) (0.059) (0.058) Education level of household_Other -0.431*** -0.392*** -0.437*** (0.052) (0.047) (0.056) Constant 0.423*** 0.426*** 0.424*** (0.042) (0.041) (0.041) Observations 3,344 3,344 3,344 R-squared 0.064 0.066 0.063 *** p<0.01, ** p<0.05, * p<0.1. Notes: Robust standard errors in parentheses. 28 Table A5: Regression results for labor indicators (1) (2) (3) Labor Force Labor Force participation (Ref: participation Unemployed VARIABLES Inactive) (Ref: Inactive) (Ref: employed) Survey year: 2016 0.250 0.174*** -0.101** (0.163) (0.052) (0.042) Survey year: 2017 0.440*** 0.590*** -0.250** (0.122) (0.110) (0.097) Inflation*Post -0.138 (0.094) Food Inflation*Post -0.207*** 0.126** (0.070) (0.051) Conflict 0.000 -0.001 0.000744** (0.001) (0.001) (0.000) Respondent is a woman 0.000 -0.016 0.0474** (0.035) (0.033) (0.020) Education level of household head_Primary 0.022 0.039 0.029 (0.036) (0.031) (0.020) Education level of household head_Secondary -0.060 -0.054 0.0783* (0.068) (0.064) (0.043) Education level of household head_University -0.036 -0.057 0.0803** (0.052) (0.056) (0.035) Education level of household_Other -0.165** -0.0793** 0.015 (0.075) (0.033) (0.015) Age 0.0103*** 0.0685*** -0.005 (0.001) (0.007) (0.003) Age-squared -0.000842*** 0.000 (0.000) (0.000) Constant 0.152** -0.695*** 0.138** (0.067) (0.114) (0.061) Observations 3,838 3,838 2,011 R-squared 0.204 0.277 0.154 *** p<0.01, ** p<0.05, * p<0.1. Notes: Robust standard errors in parentheses. 29 Table A6: Regression results for hunger (1) (2) (3) VARIABLES Hunger Hunger Hunger Survey year: 2016 -0.822*** -0.322* -0.525** Survey year: 2017 (0.255) (0.187) (0.198) -0.726*** -0.718*** -0.403*** Inflation*Post (0.199) (0.255) (0.148) 0.430*** Food price inflation*Post (0.145) 0.329** Non-food price inflation*Post (0.143) 0.193* Conflict (0.108) 0.00362** 0.00563*** 0.00424** Female household head (0.002) (0.002) (0.002) (0.036) (0.020) (0.042) Education level of household head_Primary (0.072) (0.074) (0.073) -0.151* -0.146* -0.152** Education level of household head_Secondary (0.077) (0.075) (0.075) -0.209* (0.195) -0.213* Education level of household head_University (0.121) (0.119) (0.123) -0.437*** -0.449** -0.450*** Education level of household_Other (0.154) (0.169) (0.162) 0.011 (0.025) 0.034 Land ownership (0.122) (0.127) (0.123) -0.264** -0.235** -0.253** Household head Unemployed (0.104) (0.109) (0.107) 0.246 0.222 0.251 Household size (0.182) (0.182) (0.191) (0.008) (0.012) (0.010) Household size-squared (0.021) (0.022) (0.022) 0.000 0.000 0.000 2nd welfare quintile (Ref=1st quintile) (0.000) (0.000) (0.000) 0.036 0.040 0.030 3rd welfare quintile (0.151) (0.147) (0.146) (0.182) (0.164) (0.186) 4th welfare quintile (0.153) (0.155) (0.152) -0.388*** -0.407*** -0.408*** 5th welfare quintile (Richest) (0.082) (0.079) (0.082) -0.507*** -0.532*** -0.515*** Constant (0.091) (0.083) (0.090) 2.499*** 2.488*** 2.513*** Observations (0.118) (0.115) (0.115) R-squared 702 702 702 *** p<0.01, ** p<0.05, * p<0.1. Notes: Robust standard errors in parentheses. 30 Table A7: Regression results for perceptions of welfare VARIABLES (1) (2) (3) (4) (5) (6) Life Present living Future living Present Future Control satisfaction conditions conditions economic economic over own conditions conditions life Survey year: 2016 1.176 0.038 -3.906** 0.244 -3.203* 0.886 (1.058) (0.399) (1.755) (0.405) (1.870) (0.530) Survey year: 2017 1.045 0.373 -2.569* 0.711*** -1.143 1.078** (0.791) (0.338) (1.427) (0.257) (1.646) (0.412) Inflation*Post -1.205* 0.480** 1.789* 0.264 1.370 -0.611** (0.612) (0.239) (0.941) (0.208) (0.976) (0.282) Conflict -0.0177** 0.00874** 0.004 0.001 0.000 -0.002 (0.008) (0.004) (0.004) (0.002) (0.003) (0.002) 2nd welfare quintile (Ref=1st quintile) 0.526** -0.212 -0.102 0.022 0.115 -0.108 (0.237) (0.159) (0.215) (0.127) (0.300) (0.144) 3rd welfare quintile 0.504* -0.269 -0.257 -0.228 -0.198 -0.00252 (0.291) (0.191) (0.231) (0.142) (0.283) (0.113) 4th welfare quintile 0.714** -0.328 0.0364 -0.0757 0.142 -0.074 (0.324) (0.209) (0.334) (0.133) (0.310) (0.100) 5th welfare quintile (Richest) 0.656** -0.496*** -0.601** -0.0836 -0.336 0.00515 (0.322) (0.177) (0.248) -0.137 -0.337 -0.136 Constant 3.283*** 3.328*** 3.679*** 3.814*** 3.209*** 2.157*** (0.210) (0.124) (0.273) (0.0792) (0.381) (0.127) Observations 1,221 1,210 851 1,146 849 1,162 R-squared 0.276 0.292 0.255 0.234 0.285 0.226 *** p<0.01, ** p<0.05, * p<0.1. Notes: Robust standard errors in parentheses. 31 Table A8: Robustness checks regression results Robustness check 1: Adding the initial level of CPI at the Boma Level to the initial specification Total Inflation Food price inflation Non-food price inflation Outcomes Poverty Poor (below USD 1.90 PPP) 0.353** 0.0332 0.322*** Log(real consumption) -0.833*** -0.173 -0.685*** Education Currently attending school (Girls) -0.0243 -0.130*** 0.0149 Labor Labor force participation: Active -0.126 -0.206*** -0.0309 Unemployed 0.0200 0.0877* 0.0107 Hunger Hunger incidence 0.509*** 0.325** 0.243** Perceptions of welfare Life satisfaction -1.218* -0.180 -0.807* Present living conditions 0.479* 0.218 0.225 Future living conditions 1.779* -0.0515 1.349** Control over own life -0.600** -0.0495 -0.516** Present economic conditions 0.272 0.399** 0.0531 Future economic conditions 1.367 -0.593 1.227* *** p<0.01, ** p<0.05, * p<0.1. Robustness check 2: Household Fixed effects with robust standard errors Total Inflation Food price inflation Non-food price inflation Outcomes Poverty Poor (below USD 1.90 PPP) 0.130* 0.0592 0.117* Log(real consumption) -0.521*** -0.243** -0.397*** Education Currently attending school (Girls) -0.114** -0.139*** -0.0557 Labor Labor force participation: Active -0.178*** -0.208*** -0.0774*** Unemployed -0.0296 0.0620*** -0.0168 Hunger Hunger incidence 0.495*** 0.493*** 0.209* Perceptions of welfare Life satisfaction -0.944* -0.287 -0.564 Present living conditions 0.308 0.315 0.0691 Future living conditions 1.472** -0.276 1.157** Control over own life -0.647*** 0.0601 -0.579*** Present economic conditions 0.319** 0.469*** 0.0801 Future economic conditions 0.813 -0.792* 0.802 *** p<0.01, ** p<0.05, * p<0.1. 32 Robustness check 3: Initial Specification for only waves 1 and 2 Total Inflation Food price inflation Non-food price inflation Outcomes Poverty Poor (below USD 1.90 PPP) 0.115 -0.0840 0.0623 Log(real consumption) -0.860* -0.141 -0.193 Education Currently attending school (Girls) 0.0971 -0.130* 0.0688 Labor Labor force participation: Active 0.0374 -0.112 0.137* Unemployed 0.0390 -0.0883 0.115 Hunger Hunger incidence -0.0397 0.570 -0.339 Perceptions of welfare Life satisfaction -0.538 2.496*** -0.437 Present living conditions 0.507 -0.224 0.0303 Future living conditions 1.367* 0.263 0.561 Control over own life -0.0384 0.128 -0.304 Present economic conditions -0.276 0.279 -0.227 Future economic conditions 1.339* -0.267 0.928** *** p<0.01, ** p<0.05, * p<0.1. 33 Robustness check 4: Initial Specification for only waves 1 and 4 Total Inflation Food price inflation Non-food price inflation Outcomes Poverty Poor (below USD 1.90 PPP) 0.596*** 0.354** 0.521*** Log(real consumption) -1.117*** -0.724*** -0.971*** Education Currently attending school (Girls) 0.0503 -0.0373 0.0646 Labor Labor force participation: Active -0.185 -0.368*** -0.0806 Unemployed -0.0123 0.117** -0.0384 Hunger Hunger incidence 0.525** 0.239 0.451*** Perceptions of welfare Life satisfaction -1.593** -1.957** -1.041** Present living conditions 0.575** 0.539** 0.416* Future living conditions 1.366 -0.0110 1.198 Control over own life -0.904* -0.310 -0.795* Present economic conditions 0.334 0.364 0.240 Future economic conditions 1.325 -0.497 1.256 *** p<0.01, ** p<0.05, * p<0.1. 34 Robustness check 5: Initial Specification for only waves 1 and 4 with inflation computed with wave 1 prices Total Inflation Food price inflation Non-food price inflation Outcomes Poverty Poor (below USD 1.90 PPP) 0.645*** 0.649*** 0.186 Log(real consumption) -1.684*** -1.509*** -0.618** Education Currently attending school (Girls) 0.116 0.0678 0.0516 Labor Labor force participation: Active -0.000360 -0.318*** 0.186** Unemployed -0.0266 -0.0890 0.0175 Hunger Hunger incidence 0.0485 0.332 -0.0918 Perceptions of welfare Life satisfaction -0.466 -1.007 0.229 Present living conditions 0.349 0.493* 0.0594 Future living conditions 1.665 2.951** 0.638 Control over own life -0.949* -0.783 -0.531* Present economic conditions 0.587** 0.344 0.364** Future economic conditions 1.121 2.153 0.500 *** p<0.01, ** p<0.05, * p<0.1. 35 Robustness check 6: Adding the initial level of CPI at the Boma Level for only waves 1 and 2 Total Inflation Food price inflation Non-food price inflation Outcomes Poverty Poor (below USD 1.90 PPP) 0.114 -0.0846 0.0618 Log(real consumption) -0.861* -0.143 -0.190 Education Currently attending school (Girls) 0.105 -0.124* 0.0658 Labor Labor force participation: Active 0.0447 -0.110 0.141* Unemployed 0.0400 -0.0860 0.118 Hunger Hunger incidence -0.0352 0.580 -0.342 Perceptions of welfare Life satisfaction -0.542 2.557*** -0.454 Present living conditions 0.489 -0.254 0.0395 Future living conditions 1.296 0.172 0.595 Control over own life -0.0220 0.146 -0.307 Present economic conditions -0.257 0.302 -0.223 Future economic conditions 1.345* -0.251 0.893* *** p<0.01, ** p<0.05, * p<0.1. 36 Robustness check 7: Adding the initial level of CPI at the Boma Level for only waves 1 and 4 Total Inflation Food price inflation Non-food price inflation Outcomes Poverty Poor (below USD 1.90 PPP) 0.599*** 0.358** 0.521*** Log(real consumption) -1.118*** -0.725*** -0.972*** Education Currently attending school (Girls) 0.0539 -0.0290 0.0660 Labor Labor force participation: Active -0.184 -0.367*** -0.0807 Unemployed -0.0122 0.117** -0.0379 Hunger Hunger incidence 0.524** 0.235 0.456*** Perceptions of welfare Life satisfaction -1.592** -1.946** -1.030** Present living conditions 0.575** 0.541** 0.408* Future living conditions 1.364 -0.000547 1.194 Control over own life -0.905* -0.324 -0.791* Present economic conditions 0.333 0.356 0.241 Future economic conditions 1.322 -0.484 1.240 *** p<0.01, ** p<0.05, * p<0.1. 37 Robustness check 8: High inflation households for waves 1 and 2 (dummy variable) Total Inflation Food price inflation Non-food price inflation Outcomes Poverty Poor (below USD 1.90 PPP) 0.0505 -0.0507 -0.0439 Log(real consumption) -0.0947 -0.229 0.111 Education Currently attending school (Girls) 0.0801 -0.136** 0.131** Labor Labor force participation: Active 0.0520 -0.101** 0.120** Unemployed -0.0219 0.0222 -0.0258 Hunger Hunger incidence -0.136 0.654*** -0.421 Perceptions of welfare Life satisfaction -0.666 1.358*** -0.605 Present living conditions 0.303 -0.413** 0.274 Future living conditions 0.457 0.212 0.369 Control over own life -0.0858 -0.0888 -0.303 Present economic conditions -0.152 0.144 -0.0478 Future economic conditions 0.417 -0.130 0.594 *** p<0.01, ** p<0.05, * p<0.1. 38 Robustness check 9: High inflation households for waves 1 and 4 (dummy variable) Total Inflation Food price inflation Non-food price inflation Outcomes Poverty Poor (below USD 1.90 PPP) 0.275** 0.454*** 0.0824 Log(real consumption) -0.787*** -0.957*** -0.399* Education Currently attending school (Girls) 0.0668* 0.0500 0.0250 Labor Labor force participation: Active 0.0187 -0.173** 0.104 Unemployed -0.00484 -0.0517 0.0238 Hunger Hunger incidence -0.0107 0.265 -0.0883 Perceptions of welfare Life satisfaction -0.0178 -0.632 0.0778 Present living conditions 0.102 0.303* 0.0946 Future living conditions 0.542 1.929*** 0.0101 Control over own life -0.356 -0.669** -0.286 Present economic conditions 0.296** 0.227 0.333** Future economic conditions 0.0931 1.576 -0.302 *** p<0.01, ** p<0.05, * p<0.1. 39 Figure A16: Robustness checks regression results Total (global) inflation and changes in real Food Inflation and Changes in real consumption consumption (Waves 1, 2, 4) (Waves 1, 2, 4) 2 2.5 2 1.5 1.5 1 1 .5 .5 0 0 -6 -4 -2 0 2 -6 -4 -2 0 2 Changes in real Consumption Changes in real Consumption Global inflation Fitted values Food inflation Fitted values Non-food Inflation and changes in real consumption (Waves 1, 2, 4) 4 3 2 1 0 -6 -4 -2 0 2 Changes in real Consumption Non food inflation Fitted values 40 Total (global) and changes in real consumption Food inflation and changes in real consumption (Waves 1, 2) (Waves 1, 2) 2 2.5 1.5 2 1.5 1 1 .5 .5 0 0 -6 -4 -2 0 2 -6 -4 -2 0 2 Changes in real Consumption Changes in real Consumption Food inflation Fitted values Global inflation Fitted values Non-Food inflation and changes in real consumption (Waves 1, 2) 4 3 2 1 0 -6 -4 -2 0 2 Changes in real Consumption Non food inflation Fitted values 41 Total (global) and changes in real consumption Food inflation and changes in real consumption (Waves 1, 4) (Waves 1, 4) 3 4 2 3 2 1 1 0 -8 -6 -4 -2 0 0 Changes in real Consumption from wave 1 to wave 4 -8 -6 -4 -2 0 Changes in real Consumption from wave 1 to wave 4 Food inflation based on wave 1 price Fitted values Global inflation based on wave 1 price Fitted values Non-Food inflation and changes in real consumption (Waves 1, 4) 5 4 3 2 1 0 -8 -6 -4 -2 0 Changes in real Consumption from wave 1 to wave 4 Non Food inflation based on wave 1 price Fitted values 42