MALAWI’S PROGRESS TOWARD SHARED PROSPERITY SINCE 2004 September 2018 Table of Contents 1. Economic Growth and Poverty ....................................................................................................................... 1 1.1 Recurring Shocks and Macroeconomic Performance................................................................................ 1 1.2 Trends of Poverty ...................................................................................................................................... 2 1.3 Number of People Living in Poverty ....................................................................................................... 12 2. Progress Toward Shared Prosperity .............................................................................................................. 13 2.1 Malawi has Become More Equal since 2010 .......................................................................................... 13 2.2 Decomposing Changes in Poverty to Growth and Redistribution ........................................................... 15 3. Profiling the Poor .......................................................................................................................................... 18 3.1 Where do the poor reside? ....................................................................................................................... 18 3.2 Characteristics of the Poor....................................................................................................................... 19 3.3 Agriculture Practice of the Poor Farmers ................................................................................................ 21 4. Conclusion ..................................................................................................................................................... 23 Appendix ........................................................................................................................................................... 25 1. Economic Growth and Poverty 1.1 Recurring Shocks and Macroeconomic Performance Malawi’s economic performance and rural livelihoods are highly susceptible to natural shocks. Historically, the country has been vulnerable to droughts, but, with climate change, the frequency of drought shocks is on the rise. For example, eight major drought episodes were recorded since 1980, and these droughts affected over 24 million people in total. The droughts have often occurred in El Niño years when extended dry spells that reduce agriculture production are normally registered. According to the 2015/16 Post-Disaster Needs Assessment (PDNA) report, these dry spells cause a 1 percent loss of gross domestic product (GDP) annually (PDNA 2017). Since 2004, the country has experienced multiple natural and macroeconomic shocks. The most recent ones include the 2012 macroeconomic crisis, the 2015 flooding, and the 2016 drought. The 2012 macroeconomic shocks following the devaluation of Kwacha and the severe fuel shortage in the same year contributed to weak economic performance. The 2015 flooding was one of the worst in decades, and it adversely affected 1.1 million people and destroyed infrastructure in 15 districts. In the subsequent year, Malawi experienced a major drought affecting many districts in the South and Central regions. The 2016 drought was greatly affected by strong El Niño conditions and resulted in erratic rains and the late onset of rainfall during the planting season (Malawi Economic Monitor October 2016; PDNA 2017). The major growth contractions Malawi experienced since 2004 could mostly be attributed to natural disasters, and the agriculture sector happens to carry the brunt of these shocks. The most recent drought (2016) and flooding (2015) have reduced real GDP growth to under the rate of population growth (3 percent). These shocks resulted in a contraction of the real agriculture GDP by 2.3 percent and 2 percent in 2016 and 2015, respectively. Even worse, during the 2005 flooding and drought, which made 4.2 million Malawians unable to meet their food needs (FAO 2005),1 the agriculture sector shrank by 9.3 percent. A similar decrease (4.6 percent) in agriculture GDP was observed during the 2006 drought (Figure 1). As a result of these multiple shocks, the real per capita agriculture GDP has decreased by 1.1 percent per year during 2004–2016. Despite the high population growth rate (2.9 percent), agriculture GDP grew by only 1.8 percent per year. As a result, the agriculture sector, which supports most rural livelihoods, has been shrinking in per capita terms. During both 2004–2009 and 2010–2016, which correspond to the interval between the two most recent household surveys, per capita agriculture value added has declined (Figure 1). 1 FAO. 2005. “Malawi facing serious food crisis” Food and Agriculture Organization of the United Nations, FAO Newsroom (September). 1 On the other hand, annual per capita real GDP, supported by the encouraging performance of the services and industrial sectors, grew by 2.3 percent between 2004 and 2016. The stellar performance of the industrial sector before 2010 and a reassuring growth of the services sector after 2010 have contributed to positive overall economic growth, even if the agriculture sector has shrunk. However, these two sectors are more urban focused than agriculture and might not facilitate poverty reduction in rural areas where the majority resides. Figure 1. Recurring shocks and the resulting weak and volatile economic growth GDP Agriculture Population Services Industry Population 10 20 7.4 6.8 6.6 5.9 15 5 3.6 4.3 3.5 Growth (%) 10 Growth (%) 0 -0.1 -2.0 -2.3 5 -5 -4.6 0 -10 -9.3 -5 Source: Authors’ calculation using data from World Development Indicators (WDI). 1.2 Trends of Poverty Consistent with the weak and volatile economic performance, the national poverty rate has not changed much since 2004.2 Even though national poverty rate has declined slightly by 1.8 percentage points between 2004 and 2010, it increased, although not statistically significant, by 0.8 percentage points since 2010. The trend in head count poverty has, however, been different in urban and rural areas (Figures 2a and 2b). Even if the urban poverty headcount has decreased substantially (by 8.1 percentage points) between 2004 and 2010, this progress has stalled after 2010. In the backdrop of turbulent macroeconomic conditions following the devaluation of Kwacha in 2012 and severe fuel shortage, 2 Poverty rate is defined based on a minimum per capita expenditure required to satisfy basic food and non-food needs, and those individuals below this minimum welfare level (moderate poverty line) are considered ‘poor’ (or ‘moderately poor’). In the Integrated Household Survey 2 (IHS2) (2004/05), the poverty line was MWK 37,002 per year per person. This poverty line has been adjusted in IHS3 (2010/11) and IHS4 (2016/17) to reflect changes in the cost of living. Those individuals with total expenditure below the amount needed to satisfy basic food needs are considered ‘ultra- poor’. The ultra-poverty line in IHS3 was MWK 22,956 per year per person, and this has also been adjusted to reflect changes in living cost during IHS3 and IHS4. 2 and the slowdown of industrial sector growth after 2009, by 2016, the poverty rate in urban areas has remained at its 2010 level (Figures 2a and 2b). On the other hand, rural poverty has been on the rise and more steeply so since 2010. It increased by 2.8 percentage points to 59.5 percent between 2010 and 2016. This rather disappointing development is expected in the light of the flooding and drought experienced in 2015 and 2016. As a predominantly agrarian-based economy, Malawi’s rural poverty is strongly associated with the performance of the agriculture sector, which, in turn, is dependent on rainfed farming. This high dependence on rainfed agriculture contributes to volatile agriculture performance as extended dry spells result in failure of major staple crops such as maize. The South and Central regions, which were severely affected by the recent flooding and drought, account for most of the increase in rural poverty during 2010–2016. During this period, poverty headcount in Rural Center and Rural South increased from by 4.9 and 1.9 percentage points to 53.6 percent and 65.2 percent, respectively. However, rural poverty remains the highest in the South (65.2 percent) and North (59.9 percent) regions and is on the rise in the Central region from 46.7 percent to 53.6 percent between 2004 and 2016 (Figures 2c and 2d). Figure 2. Trends in poverty rate since 2004 (a) Poverty rate (b) Change in poverty rate 2004 2010 2016 12 2004-2010 2010-2016 Change in poverty rate 8 Poverty rate 0.4 0.8 4 0.9 52.4 50.7 51.5 55.9 56.6 59.5 0 -8.1 -4 2.8 25.4 -1.8 17.3 17.7 -8 -12 National Urban Rural National Urban Rural (c) Rural poverty, by region (d) Change in rural poverty, by region 2004 2010 2016 12 2004-2010 2010-2016 Change in poverty rate 8 1.9 Poverty rate 1.9 64.4 63.3 65.2 4 56.3 59.9 59.9 0 53.6 0 46.7 48.7 -4 3.7 4.9 -1.1 -8 Rural North Rural Central Rural South -12 Rural North Rural Central Rural South Source: Authors’ calculation based on IHS2–IHS4. Note: For Figures (b) and (d), the 95 percent confidence intervals are shown by the corresponding vertical lines. 3 Growth incidence analysis indicates that there have been starkly different changes in the welfare fortune of those in the bottom and the top of the consumption distribution during 2004–2010 and 2010–2016. The analysis, which estimates the real per capita consumption growth for individuals in each consumption percentile ranking, is conducted for 2004–2010 and 2010– 2016. The results are summarized in Figure 3 and Figure A.1 in the appendix. These two periods show very different consumption growth trajectories for the poor (the bottom 10 percent and bottom 40 percent) and the non-poor (the top 60 percent). Growth has been pro-poor after 2010, compared to the less pro-poor growth trajectory observed during 2004–2010. For instance, the consumption of the bottom 10 percent and bottom 40 percent Malawians declined by 2.6 percent and 1.1 percent per year during 2004–2010. This has been reversed after 2010 when consumption of the bottom 10 percent grew by 3.1 percent and that of the bottom 40 percent grew by 1.7 percent per year. On the other hand, the consumption of the top 60 percent grew by 1.6 percent per year during 2004–2010, which was later reversed after 2010 when it declined by 1.5 percent per year (Figures 3a, 3b, and 3c). Similarly, the consumption growth in rural areas has been pro-poor since 2010 and less so during 2004–2010. The real consumption of the bottom 10 percent of rural households grew by 2.9 percent per year between 2010 and 2016, compared to a −1.6 percent annual growth for the top 60 percent. However, during 2004–2010, the consumption of the bottom 10 percent shrank by 2.9 percent while that of the top 60 percent grew by 0.3 percent per year. Like those in the lowest decile, the bottom 40 percent also saw a positive consumption growth after 2010, which compensated for the negative growth registered before (Figures 3a, 3b, and 3c). This pro-poor rural growth after 2010 is observed in all regions. Before 2010, the consumption growth of the bottom 10 percent and bottom 40 percent was negative and much lower than the consumption growth of the top 60 percent in all regions. This has changed vividly after 2010; consumption growth was robustly positive for those in the bottom of the consumption distribution and negative for those in the top of the distribution (Figures A.1 in the appendix). On the other hand, consumption growth of the urban poor has mostly remained positive since 2004. During 2004–2010, the bottom 10 percent and bottom 40 percent experienced consumption growth of 2 percent and 2.8 percent, respectively. This growth has slowed down respectively to 1.2 percent and 0 percent between 2010 and 2016. Urban growth has been pro-poor during 2010–2016: the growth of the top 60 percent has been negative, compared to a non-negative growth for those in the bottom of the consumption distribution. However, despite the positive growth in consumption of those in the lower consumption percentiles, the 2004–2010 period growth was not pro-poor (Figures 3a, 3b, and 3c). 4 Figure 3. Growth incidence during 2004–2016 (a) 2004–2010 in urban and rural areas (b) 2010–2016 in urban and rural areas 10 National Urban Rural 10 National Urban Rural 5 Annual growth rate (%) Annual growth rate (%) 5 0 0 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 -5 -5 -10 -10 -15 -15 Expenditure percentiles Expenditure percentiles (c) Annualized growth by location 6 National Rural Urban Annualized growth (%) 4 Pro-poor 2 The bottom 10 percent and 40 percent 0 experienced a strong consumption growth during 2010–2016 -2 Not pro-poor -4 -6 B10% B40% T60% B10% B40% T60% 2004-2010 2010-2016 National -2.6 -1.1 1.6 3.1 1.7 -1.5 Rural -2.9 -1.6 0.3 2.9 1.7 -1.6 Urban 2.0 2.8 3.8 1.2 0.0 -4.5 Source: Authors’ calculation based on IHS2–IHS4. Despite a seemingly static poverty rate since 2004, a huge seasonality in well-being is recorded. The welfare status of Malawian households deteriorates markedly during the lean season: food and non-food consumption decreases and hence poverty increases during the lean season, particularly in drought years. Analysis of the poverty rate across survey quarters of the three IHS waves indicates that there is a substantial change in welfare and poverty rate over the seasons of a given year. The IHS waves are designed to be representative of economic conditions throughout the year, and the analysis shows a huge variation in welfare between lean and harvest seasons of a given year. Maize is normally harvested in April, and generally the lean season starts in December/January. During the lean season, smallholder poor farmers run out of food stock, and the lean season might ‘arrive early’ during drought years. The first survey quarter (April–June) is just after harvest is collected and food shortage is rarely observed. In subsequent quarters, rural households start to slowly run out of food stock, and by the lean season (January– 5 March), many smallholder farmers tend to deplete their maize reserves and severe shortage kicks in. As a result, food consumption declined by 23 percent and 15 percent in the lean season, compared to the harvest season, following the 2005 and 2016 droughts, respectively. The seasonal change in the fortune of rural households is well-captured in the survey quarters of the three IHS waves (Figure 4a). The dynamics in welfare within a year is more pronounced in the drought years. For example, after the 2004/05 drought, households’ ran out of food stock and the median per capita real food consumption declined by 23 percent during the lean season: from MWK 83,400 during harvest time (April–June) to MWK 64,400 in the lean season (January– March). Similarly, a 15 percent decrease in food consumption is recorded in the lean season following the 2016 drought. On the other hand, when no major drought was recorded in 2010, the seasonal fluctuation in food consumption is less pronounced. Only a 12 percent consumption gap between harvest time and the lean period was observed in 2010, and there was no clear seasonal pattern (Figures 4a and 4b). During the lean seasons of drought years, not only did food consumption decline, but overall real consumption also declined. The total real consumption per capital declined by 23 percent (in 2005) and 17 percent (in 2016), compared to the corresponding harvest seasons. On the other hand, like food consumption, the decrease in total consumption during the lean season of a non- drought year (2010) was small (9 percent). Following a similar pattern as real consumption, the rural poverty rate increased by 14 and 19 percentage points during the lean seasons of 2016 and 2004 compared to the corresponding harvest seasons. After the 2016 drought, the poverty headcount increased from 49.6 percent in the harvest season to a whopping 63.9 percent in the lean season. Similarly, after the 2004/05 drought, the rural poverty rate increased from 50.8 percent in the harvest season to 70 percent in the lean season. This lean season poverty surge is worse in drought years compared to normal years like 2010. For example, the gap in the rural poverty rate between the harvest and lean periods of 2010 was only 6 percentage points (Figure 4c). 6 Figure 4a. Consumption level declines during lean season 4a.(1) Food consumption 4a.(2) Total consumption 2004/5 2010/11 2016/17 2004/5 2010/11 2016/17 100 150 capita (1000 Kwacha in 2016 90 Median consumption per 130 80 110 90 prices) 70 70 60 50 50 Apr-Jun Jul-Sept Oct-Dec Jan-Mar Apr-Jun Jul-Sept Oct-Dec Jan-Mar Survey quarters Survey quarters Figure 4b. Quarterly change (%) in food consumption, relative to harvest period (April–June) 10 Jul-Sept Oct-Dec Jan-Mar 4 0 harvest period (Apr-Jun) -2 consumption relative to -5 Change (%) in median -9 -9 -9 -10 -12 -12 -13 -15 -13 -10 -15 -18 -17 -21 -23 -20 -23 -30 2004/5 2010/11 2016/17 2004/5 2010/11 2016/17 Food consumption Total consumption Figure 4c. Rural poverty increases during lean seasons of drought years (2004/5 and 2015/6) 2004/5 2010/11 2016/17 Poverty rate 70.0 63.9 60.3 61.4 60.8 57.9 57.7 55.6 50.8 51.7 49.6 48.2 Apr-Jun Jul-Sept Oct-Dec Jan-Mar Survey quarters 2004/2010/2016 2005/2011/2017 Source: Authors’ calculation based on IHS2–IHS4. Note: IHS2 and IHS3 were implemented during March 2004–March 2005 and March 2010–March 2011, respectively. IHS4 was implemented between April 2016 and April 2017. Hence, the first quarter for IHS4 excludes March, and April is included in the last quarter. 7 The 2016 poverty estimates uncovered some good news in that ultra-poverty has declined by 4.4 percentage points since 2010. Most of the decrease in ultra-poverty came from rural areas where it declined from 28.1 percent to 23.8 percent. This is a very encouraging development despite the 2016 drought and the 2015 flooding preceding the implementation of IHS4. This encouraging development has reversed the increase in rural ultra-poverty experienced between 2004 and 2010. On the other hand, despite a strong decrease in ultra-poverty in urban areas between 2004 and 2010, the progress has stalled after 2010 (Figure 5). Figure 5. Ultra-poverty declined, mainly in rural areas 2004 2010 2016 Ultra-poverty rate 28.1 24.5 24.3 23.8 22.4 20.1 7.5 4.3 4.1 National Urban Rural 12 2004-2010 2010-2016 Change in ultra-poverty 8 3.8 4 2.1 rate 0 -4.4 -0.2 -4.3 -4 -3.2 -8 -12 National Urban Rural Source: Authors’ calculation based on IHS2–IHS4. What explains the decline in ultra-poverty? This is a crucial question that requires in-depth research, particularly the relative increase in consumption of the ultra-poor during drought years, while the consumption of the top 60th percentile declined. Even though establishing causality is difficult and data limitation is also an issue, preliminary analysis suggests that transfers (private and public) could be playing an important role. Before diving into the findings, it is important to describe the data challenge and its possible implications, and the national and international food security response after the 2016 drought. The IHS questionnaire asks households about their food consumption in the past week from different sources, including own production, purchase, and gifts/transfers. The latter source (gift/transfer) is not further disaggregated to identify the source: relatives/friends, government, and/or nongovernmental organizations. As a result, we cannot distinguish private transfers from public transfers that could have been implemented as part of the government’s and/or development partners’ drought response and safety net programs. This complicates the analysis by limiting our ability to attribute the increase in transfer incomes to various programs. 8 The amount of food transfer received by the ultra-poor increased by 76 percent since 2010, and this increased the share of transfers in food consumption expenditure from 11 percent in 2010 to 17 percent in 2016. This huge increase in the amount of transfer (from MWK 4,085 per capita in 2010 to MWK 7,205 per capita in 2016, both in 2016 prices) has likely contributed to reduction in ultra-poverty. The percentage increase in transfer is much smaller for those who are moderately poor (47 percent) or non-poor (14 percent). The amount of transfer is generally high for these two groups, even though it accounts for a smaller share of their consumption expenditure (Table 1). Table 1. Amount of food gift/transfer and its share in consumption increased marked in 2016 Amount of food gift Share (%) of food gift in: (MWK/person) a Total Consumption Food Consumption Ultra-poor 4,085 7 11 2010 Poor 7,230 7 10 Non-poor 10,534 4 7 Ultra-poor 7,205 11 17 2016 Poor 10,634 10 15 Non-poor 12,004 6 9 Source: Authors’ calculation based on IHS2–IHS4. Note: a. All consumption expenditures are expressed in 2016 prices. Food transfer accounts for a high share of food consumption for those in the bottom deciles of the welfare distribution, and its share has increased considerably in 2016. As presented in Figure 6a, the share of transfers in the food basket and total consumption expenditure is much higher for those in the lowest consumption deciles and continuously declines in higher deciles. Between 2010 and 2016, the share of transfer in consumption expenditure has increased markedly for those in the lowest consumption deciles. The increase is much lower for those in the top of the consumption distribution. The increase in food transfer to extremely poor households and hence the resulting reduction in ultra-poverty could be the result of concerted international response to the 2016 drought. According to the 2016/17 Food Insecurity Response Plan, 6.5 million Malawians needed food assistance during July 2016–March 2017, and 2.4 million farmers lost all their production. A total of US$233.8 million was needed for food security response (food budget). The Government of Malawi and the international community have responded decisively by contributing US$68 million and US$111.1 million, respectively. With a total of US$179.1 million contributions, only a small (US$54.8 million) shortage was recorded. In terms of the actual quantity of food, a total of 261,555 metric tons of maize/cereal was needed and 200,808 metric tons was secured (Department of Disaster Management Affairs, Office of the Vice President). 3 3 https://reliefweb.int/sites/reliefweb.int/files/resources/FIRP%20FV%20July%2013%202016.pdf 9 In the absence of a massive food security response following the 2016 drought, ultra-poverty could have increased slightly. This is better understood by conducting analysis of real consumption distribution when food transfer is excluded (or included) in the total consumption basket. The results (Figure 6b) show that when food transfers are excluded from aggregate consumption of households both in IHS4 and IHS3, there has been minimal change in real consumption per capita of those below the ultra-poverty line between 2010 and 2016. Part of the consumption density functions (of IHS3 and IHS4) before the ultra-poverty line pretty much overlap when transfer is excluded from the consumption basket. In fact, ultra-poverty could increase slightly if the transfers did not surge in response to drought. Figure 6a. Share of public and private transfer food in household consumption (by deciles of total consumption) 2010 Share in total consumption 2010 Share in food consumption 20 2016 Share in total consumption 2016 Share in food consumption 16 Share (%) 12 8 4 0 Poorest 2 3 4 5 6 7 8 9 Richest (lowest) (highest) decile decile Figure 6b. Density of consumption expenditure with and without including food transfers (i) Without excluding transfers/gifts (ii) Excluding transfers/gifts Source: Authors’ calculation based on IHS2–IHS4. 10 The poverty gap and poverty severity have also decreased in 2016, and this shows that the consumption of the poor has increased and inequality among the poor has also decreased. These poverty indices respectively capture how far the poor are from the poverty line and inequality among the poor (see Box 1 for details). In 2010, the average consumption of the poor was 18.9 percent less than the moderate poverty line, but this consumption shortfall decreased by 2.1 percentage points to 16.8 percent in 2016. This has reversed the increase in poverty gap observed between 2004 and 2010. In rural areas, the poverty gap increased by 2.2 percentage points during 2004–2010, but decreased by 1.7 percentage points after 2010. In urban areas, the poverty gap decreased by 2.2 percentage points between 2004 and 2010, but has not changed since 2010. A similar decrease in poverty severity is observed, mainly in rural areas (Table 2). The poverty gap and poverty severity of the ultra-poor declined continuously since Box 1. Poverty indices 2004. The consumption shortfall of the To measure poverty, the class of poverty measures ultra-poor declined from 11.8 percent to proposed by Foster, Greer, and Thorbecke (FGT) are used (Foster et al. 1984). In addition to poverty headcount 4.7 percent of the ultra-poverty line index, the FGT provides poverty gap and severity indices. between 2004 and 2016. This, along with the This family of poverty indices can be summarized by the following equation: decline in ultra-poverty headcount since 1 − 2010, is quite a reassuring development in = ∑ =1 ( ) ∗ ( < ) protecting the extremely poor individuals where  is a non-negative parameter that takes value 0, 1, despite the natural disasters experienced in or 2; is the poverty line; denotes consumption of individual ; and is the total number of individuals in the recent years. The favorable change in ultra- population. ( < ) is an indicator function which is poverty gap is observed both in rural areas equal to one when individual ′ consumption is below the and urban centers. In rural areas, the poverty poverty line, and zero when the consumption is above the poverty line. gap declined from 12.8 percent to 5.6 percent The poverty headcount index (=0) is the percentage of of the ultra-poverty line between 2004 and population whose consumption is below the poverty line. 2016. Similarly, it declined from 3.9 percent This simple and easy-to-interpret index is the most widely used poverty measure. However, it has some limitations in to 0.8 percent in urban centers during the that it does not capture how close/far the poor are from the same period. Severity of ultra-poverty also poverty line and the distribution of consumption among declined in both urban and rural areas since the poor. The two other poverty indices address these limitations. The poverty gap (=1), which is the average 2004 (Table 2). consumption shortfall of the poor relative to the poverty line. Finally, the poverty severity or poverty gap squared (=2) accounts for the inequality among the poor. In the poverty severity index, larger poverty gaps carry higher weight (Haughton and Khandker 2009). 11 Table 2. Trends of poverty headcount and poverty gap National Rural Poverty levels Difference Poverty levels Difference 2004–2010 2010–2016 2004–2010 2010–2016 2004 2010 2016 2004 2010 2016 Moderate Poverty Headcount 52.4 50.7 51.5 −1.8** 0.9 55.9 56.6 59.5 0.8 2.8*** Poverty gap 17.8 18.9 16.8 1.1*** −2.1*** 19.2 21.4 19.7 2.2*** −1.7*** Poverty severity 8 9.3 7.4 1.3*** −1.9*** 8.6 10.6 8.7 1.9*** −1.9*** Ultra-Poverty Headcount 22.4 24.5 20.1 2.1*** −4.4*** 24.3 28.1 23.8 3.8*** −4.3*** Poverty gap 11.8 7 4.7 −4.8*** −2.4*** 12.8 8 5.6 −4.7*** −2.5*** Poverty severity 6.4 2.8 1.6 −3.6*** −1.2*** 7 3.3 2 −3.7*** −1.3*** Urban Poverty levels Difference 2004 2010 2016 2004–2010 2010–2016 Moderate Poverty Headcount 25.4 17.3 17.7 −8.1*** 0.4 Poverty gap 7.1 4.8 4.5 −2.2*** −0.3 Poverty severity 2.8 2 1.6 −0.8** −0.4* Ultra-Poverty Headcount 7.5 4.3 4.1 −3.2*** −0.2 Poverty gap 3.9 1.3 0.8 −2.6*** −0.5** Poverty severity 2 0.5 0.2 −1.5*** −0.3*** Source: Authors’ calculation based on IHS2–IHS4. Note: *** indicates significant at 1%, ** significant at 5% and * significant at 10%. 1.3 Number of People Living in Poverty Since 2004, the number of Malawians below the national poverty line has increased by 2 million. The number of poor people could increase or decrease depending on the rate at which the total population is increasing and rate of change in poverty. Even if Malawi has experienced a slight decrease in poverty rate between 2004 and 2010, the population grew faster than the pace of poverty reduction, and hence the number of poor people increased from 6.4 million to 7.1 million.4 This has further increased to 8.4 million by 2016. As poverty is mainly a rural phenomenon in Malawi, almost all the increase in poor population is in rural areas. The number of poor Malawians 4 The number of poor people is estimated using population implied by IHS data and the poverty rate. 12 residing in rural areas increased from 6 million in 2004 to 6.8 million in 2010, before increasing sharply to 7.9 million in 2016. In urban areas, it increased only slightly from 0.4 million in 2004 to 0.5 million in 2016. The number of people living in ultra-poverty has declined from 3.5 million in 2010 to 3.3 million in 2016. This decrease in the number of ultra-poor individuals is the result of a massive reduction in the ultra-poverty rate, which has more than offset the population growth. The decrease in ultra-poor population, which has occurred exclusively in rural areas, is an encouraging development and has partially reversed deteriorating ultra-poverty situation observed during 2004–2010. In urban areas, there has been no change in the number of people in ultra-poverty (Figure 7). Figure 7. Number of people in ultra-poverty declined in 2016 2004 2010 2016 Number of people in ultra- 3.5 3.4 3.3 poverty (million) 3.1 2.7 2.6 0.1 0.1 0.1 National Rural Urban Source: Authors’ calculation based on IHS2–IHS4. 2. Progress Toward Shared Prosperity The growth incidence analysis result presented earlier has shown that consumption growth has been pro-poor after 2010, especially in rural areas. This section further investigates the progress toward shared prosperity by analyzing growth and redistribution since 2004. 2.1 Malawi has Become More Equal since 2010 Since 2010, Malawi has made tremendous advancement toward reducing inequality, mostly in rural areas. Despite the unfavorable trend in consumption distribution Malawi experienced between 2004 and 2010, the country has managed to reverse most of this increase in inequality after 2010. Figure 8 presents the kernel density of consumption distribution during 2004–2016, and the Gini coefficients for both rural and urban areas. Inequality, as measured by the Gini coefficient, decreased from 0.45 to 0.42 during 2010– 2016, and the decrease was much greater in rural areas. Before it was reversed in 2010, 13 inequality has been on the rise as the Gini coefficient increased from 0.39 to 0.45 between 2004 and 2010, and the density of consumption for 2010 became flatter than the density in 2004. This increase in inequality was felt more in rural areas where the Gini coefficient increased from 0.34 to 0.38. The decrease in inequality after 2010 is an encouraging development, and it has reversed the increase in inequality observed in rural areas during 2004–2010: the Gini coefficient has decreased from 0.38 in 2010 to 0.32 in 2016. On the other hand, inequality has continued to increase in urban areas even after 2010, but the increase has been modest but steady (Figure 8). 5 Figure 8. Malawi has made progress toward reducing inequality after 2010 (a) Density of consumption per capita 2004–2016 (b) Inequality has decreased in rural areas 2004 2010 2016 Gini coefficient 0.48 0.49 0.50 0.45 0.42 0.39 0.38 0.34 0.32 National Urban Rural Source: Authors’ calculation based on IHS2–IHS4. International comparison shows that Malawi’s inequality level is moderately high but decreasing since 2010. Compared to other neighboring eastern and southern African countries, inequality was moderately high in 2010 but it declined since then. Except Tanzania and Uganda, Malawi’s inequality has become lower than other neighboring countries based on the latest available figures. However, given that the income level is much lower in Malawi compared to these countries, the inequality is considered to be high (Figure 9). Figure 9. International comparison of inequality 57 Gini coefficient 54 45 46 42 43 43 40 41 39 38 2004 2010 2016 2008 2007 2011 2004 2015 2011 2005 2012 Malawi Mozambique Tanzania Zambia Zimbabwe Uganda Source: WDI. 5 For consumption density in rural areas and urban centers in 2010 and 2016, see Figure A.2 in the appendix. 14 Another way of understanding the recent changes in inequality and shared prosperity within Malawi is by investigating the trends of consumption share of those in the bottom of the consumption distribution. Figure 10 presents trends of the consumption share, relative to total consumption in the country, of the bottom 20 percent and the bottom 40 percent since 2004. The consumption share of the bottom 40 percent of the population, as a percentage of the total consumption, declined from 17.8 percent in 2004 to 15 percent in 2010 before increasing to 17.1 percent in 2016. Similarly, the share of the bottom 20 percent increased from 5.6 percent in 2010 to 6.8 percent in 2016. In rural areas, the recent (post-2010) progress in shared prosperity— that is, a 3.2 percentage points increase in the consumption share of the bottom 40 percent—is much higher in that it reversed the 1.9 percentage point decline in the consumption share observed during 2004–2010. A similar improvement in the consumption share of the bottom 20 percent in rural areas is observed after 2010. In urban areas, there has been a moderate increase in the consumption share of those in the bottom of the distribution since 2010 (Figure 10). However, as expected, urban centers have generally been more unequal than rural areas. The consumption share of the bottom 40 percent in urban has been much lower than those in rural areas: 14.6 percent and 21 percent in 2016, respectively (Figure 10). Figure 10. Consumption share of those in the bottom of the distribution increased in 2016 30 2004 2010 2016 Consumption share (%) 20 in total consumption 19.7 21.0 17.8 17.1 17.8 10 15.0 13.6 13.0 14.6 7.0 5.6 6.8 5.1 4.8 5.5 7.8 6.7 8.4 0 Bottom 20% Bottom 40% Bottom 20% Bottom 40% Bottom 20% Bottom 40% National Rural Urban Source: Authors’ calculation based on IHS2–IHS4. 2.2 Decomposing Changes in Poverty to Growth and Redistribution The change in poverty since 2004 could be decomposed into growth or redistribution components as suggested in Datt and Ravallion (1992).6 The growth component is the change in poverty due to change in average consumption, holding the inequality (Lorenz curve) constant. The redistribution component is the result of the change in the Lorenz curve while keeping the average consumption constant. In this section, we decompose the changes in poverty into these two components by first discussing the growth in mean consumption and shifts in the Lorenz curve. 6 Datt, G. and M. Ravallion. 1992. “Growth and Redistribution Components of Changes in Poverty Measures: A Decomposition with Applications to Brazil and India in the 1980s.” FullText, Journal of Development Economics 38 (2): 275–295. 15 Consumption growth pattern has been very different during the pre-2010 and post-2010 periods, particularly in urban areas. Between 2004 and 2010, the national average per capita consumption grew by 2.8 percent. However, in the subsequent period (2010–2016), it shrank by 1.5 percent. The reversal in growth trajectory is more vivid in urban areas: the growth rates in consumption per capita were 19.9 percent and −16.1 percent during 2004–2010 and 2010–2016 periods, respectively. Rural areas, on the other hand, experienced a continuous decline in real per capita consumption since 2004 (Table 3). Table 3. Growth (%) in average consumption per capita during 2004–2016 (1) (2) (3) Period National Urban Rural 2004-2010 2.8*** 19.9*** -3.0*** (0.9) (2.8) (0.9) N 23,551 3,673 19,878 2010-2016 -1.5* -16.1*** -2.0** (0.9) (2.3) (0.8) N 24,718 4,505 20,213 Source: Authors’ calculation based on IHS2–IHS4. Note: *** indicates significant at 1%, ** significant at 5% and * significant at 10%. There has been a slight improvement in welfare distribution between 2010 and 2016, especially in rural areas. This is shown in the Lorenz curves, which depict the percentage shares of total consumption in Malawi captured by those individuals in the corresponding bottom percentiles of the distribution (Figure 11). The Lorenz curve for 2016 is closer to the equitable distribution line than the 2010 Lorenz curve. The recent (2010–2016) progress toward equity has partially reversed the inequality increase observed between 2004 and 2010. This is mainly true in rural areas where the inequality declined below its 2004 level. In urban areas, the change in distribution of consumption has been less pronounced (Figure A.3). 16 Figure 11. Lorenz curves, 2004–2016 (a) National (b) Rural 100 100 2004 2004 Commulative consumption (%) 2010 2010 Commulative consumption (%) 80 2016 80 2016 Equitable distribution Equitable distribution 60 60 40 40 Most 20 Most 20 unequal unequal 0 0 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Proportion of the population (%) Proportion of the population (%) Source: Authors’ calculation based on IHS2–IHS4. Note: The ‘equitable distribution’ line is where each percentile of the cumulative consumption is captured by the corresponding population percentile. In other words, total consumption is equally shared among all individuals in the country if the Lorenz curve overlaps with this line. During 2004–2010, economic growth, particularly in urban areas, has been the driving force for poverty reduction while regressive distribution has hampered poverty reduction. The role of growth and redistribution in poverty reduction has been very different before and after 2010. Figure 12 presents the growth decomposition. During 2004–2010, growth has contributed to a significant reduction in urban poverty rate (9.7 percentage points) and hence, national poverty rate (7.3 percentage points). This is consistent with the high consumption growth observed in urban areas during this period (Table 2). On the other hand, a regressive distribution has resulted in an increase in poverty during this period by 5.6 percentage points (Figure 12a). After 2010, consumption growth has been negative, and the lack of growth has increased poverty rate by 2.8 percentage points nationally. In both rural and urban areas, the negative growth has increased poverty rate by 5.3 and 4.5 percentage points, respectively. On the other hand, a more progressive distribution observed during 2010–2016 period has contributed to a 1.9 percentage point reduction in poverty (Figure 12a). The decline in the ultra-poverty rate since 2010 is also mostly a result of a more equitable distribution. During 2010–2016, a favorable distribution reduced the ultra-poverty rate by 6.5 percentage points, and the role of redistribution is more pronounced in rural areas where it reduced ultra-poverty by 8.6 percentage points. However, the lack of growth is responsible for increases in ultra-poverty both in urban and rural areas. On the other hand, before 2010, unfavorable redistribution increased ultra-poverty while growth was helping with reduction in the ultra-poverty rate, particularly in urban areas (Figure 12b). 17 Figure 12. Decomposition of changes in poverty (a) Moderate poverty (b) Ultra-poverty 12 12 Growth Distribution Growth Distribution Changes in ultra-poverty (%) 8 8 8.0 Changes in poverty (%) 4 4 4.7 5.6 5.3 4.5 4.3 1.8 1.6 2.8 1.5 2.0 2.4 0 -1.0 -1.9 0 -0.9 -2.5 -4.1 -2.6 -4 -4.7 -7.3 -4 -5.9 -6.5 -8 -9.7 -8.6 -8 -12 -12 Rural Rural Urban Urban National National Urban Urban Rural Rural National National 2004-2010 2010-2016 2004-2010 2010-2016 Source: Authors’ calculation based on IHS2–IHS4. 3. Profiling the Poor 3.1 Where do the poor reside? Poverty has been and continues to be a rural phenomenon in Malawi. The great majority of the poor (93 percent) reside in rural areas by 2016, and the poverty rate is much lower in urban areas and so is the share of the poor in urban centers. Since 2004, there has been little change in the distribution of the poor across rural and urban areas (Figure 13). Almost half of the poor population resides in the South region, while a small share of the poor population is in the North region. The South, which has a high poverty rate and a large population, accounts for 49 percent of the poor in Malawi by 2016. Its share in the total poor population has declined from 52 percent in 2004, mainly due to the increase in poverty rate in the Center region. The share of the poor population residing in the Center region has increased from 36 percent in 2004 to 42 percent in 2016. On the other hand, the North region that has a small population size accounts for only a tiny share of the total poor population in Malawi (Figure 13). 18 Figure 13. Poverty continues to be a rural phenomenon and concentrated in South region 2004 2010 2016 Proportion of the poor by 95 95 93 location of residence 52 49 49 42 36 37 12 14 7 9 5 5 Urban Rural North Center South Source: Authors’ calculation based on IHS2–IHS4. The poor are also concentrated in some districts, and in fact 15 of the 31 districts covered in IHS account for three-fourth of the poor population. The poor population is concentrated in Non-city Lilongwe and Mangochi Districts which respectively account for 9 percent and 8 percent of the total poor population. The top 15 districts in concentration of poor population (from Non- city Lilongwe to Balaka in Figure A.4 account for 74 percent of the poor population in Malawi by 2016. Mzuzu City, Zomba City, and Likoma District had the lowest share of the total poor population (Figure A.4). However, the difference in the distribution of the poor population across districts stems mainly from differences in population size among districts, not exclusively due to differences in poverty rate. As shown in Figure A.5 and Map A.1, poverty rate difference among rural districts is much smaller and those districts with high share of the poor population such as Non-city Lilongwe and Mangochi do not have particularly high poverty rates. 3.2 Characteristics of the Poor It is not surprising that poor Malawian households have low human capital, which hinders their access to decent labor income. These households tend to have large families due to large number of children compared to non-poor households. This creates a high (child) dependence ratio and limits the amount of income the households earn. In addition, the heads and other adult members of poor households have lower education and skills than that of non-poor households. In 2016, only 79 percent of the poor households were headed by literate individuals, about 48 percent of the heads cannot read in Chichewa and even more (75 percent) cannot read in English.7 The literacy rate is much higher (91 percent) among heads of the non-poor households. In addition, the average years of education for the head of a poor household is lower by about three years compared to that of the head of a non-poor household. Similarly, the spouse of the household head and other adult members of the poor households have lower years of education than that of the non-poor 7 A person is defined as literate if he/she has ever been to school. 19 households. Table 4 presents the differences in the profile of poor and non-poor households. The relative human capital position of the poor has not changed much since 2010. A salient feature of poverty in Malawi is that most of the poor engage in agriculture, particularly crop production. By 2016, 91 percent of the poor households participate in crop production—this is 20 percentage points higher than the share of non-poor households involved in crop production. Larger shares (44 percent) of adult members of poor households are also employed in agricultural activities compared to adults in non-poor households (37 percent).8 In the same way, the share of household heads engaged in agricultural activities is higher in poor than non-poor households (Table 4). Engagement in nonfarm activities and wage employment are associated with low poverty rate, while casual employment/ganyu is prevalent among poor households. Even though participation in nonfarm activities is not widespread in Malawi, a relatively high share (14 percent) of adult members in non-poor households work in family-owned nonfarm enterprises, compared to only 6 percent for those in poor households. Some 20 percent and 9 percent of heads of non- poor and poor households are also employed in own nonfarm enterprises, respectively (Table 4). The poor reside far from some basic services such as road, markets, and Agricultural Development and Marketing Corporation (ADMARC) centers, and this limits their market access. A typical poor household is located 4 km farther from a road than a non-poor household. The poor are also far away from the ADMARC centers and daily markets, which limits their market access and possibly reduces returns from agriculture. Remoteness of the location, averaging 37 km from a population center of 20,000 people, also exacerbates the poor’s lack access to market (Table 4). About 87 percent of the poor have reported facing food insecurity in the 12 months leading up to the 2016/17 survey, and this is after a marked increase compared to the 61 percent of households in months leading up to 2010/11. The level of food insecurity for non-poor households is much lower: 36 percent in 2010/11 and 59 percent in 2016/17. As a result of the 2015 flooding and 2016 drought, reported food insecurity increased for both poor and non-poor households. However, the problem is exacerbated more among poor households. Food insecurity in a short recall period of one week is also higher among the poor than the non-poor (Table 4). Table 4. Profile of the poor in Malawi 2010/11 2016/17 Non- Differ- Non- Differ- Poor poor ence Poor poor ence Household characteristics Household size 6.10 5.20 0.90*** 5.60 4.70 0.93*** 8 This is based on employment in the previous week, and hence it might be affected by seasonality of agricultural activities. 20 Number of children 3.30 2.40 0.93*** 3.00 2.00 0.98*** Number of elderly members 0.15 0.12 0.03*** 0.14 0.14 0.00 Number of working age members 2.70 2.70 -0.05** 2.50 2.60 -0.04* Dependence ratio 1.56 1.10 0.46*** 1.50 1.03 0.47*** Child dependence ratio 1.50 1.05 0.45*** 1.43 0.96 0.47*** Old age dependence ratio 0.15 0.12 0.03*** 0.14 0.16 -0.02** Female head 0.22 0.17 0.05*** 0.29 0.22 0.07*** Head's age (in years) 43.30 41.60 1.71*** 43.50 42.80 0.69*** Head is literate 0.69 0.87 -0.19*** 0.79 0.91 -0.12*** Head can read in Chichewa 0.55 0.78 -0.23*** 0.62 0.82 -0.20*** Head can read in English 0.23 0.50 -0.27*** 0.25 0.53 -0.28*** Head's years of education 4.10 6.90 -2.83*** 4.60 7.40 -2.77*** Spouse's years of education 3.00 5.60 -2.59*** 3.90 6.40 -2.51*** Average years of education of adult members 4.10 6.60 -2.48*** 4.80 7.20 -2.36*** Economic activity Household engaged in crop production 0.94 0.77 0.17*** 0.91 0.70 0.21*** Household engaged in livestock rearing 0.49 0.50 -0.01 0.39 0.43 -0.05*** Share of household member that are employed a 0.71 0.68 0.03*** 0.63 0.63 0.01 Share of household members engaged in own agriculture 0.63 0.48 0.15*** 0.44 0.37 0.07*** Share of household members engaged in non-farm enterprise 0.07 0.14 -0.08*** 0.06 0.14 -0.09*** Share of household members employed for wage 0.05 0.13 -0.09*** 0.03 0.12 -0.09*** Share of household members engaged in ganyu/casual jobs 0.17 0.10 0.07*** 0.27 0.14 0.13*** Household head employed in any activity 0.81 0.85 -0.05*** 0.74 0.79 -0.05*** Head engaged in own agriculture 0.67 0.52 0.15*** 0.48 0.41 0.07*** Head engaged in non-farm enterprise 0.09 0.19 -0.10*** 0.09 0.20 -0.11*** Head employed for wage 0.09 0.24 -0.15*** 0.06 0.22 -0.16*** Distance (in km) from services Road 11.3 7.2 4.09*** 11.3 7.4 3.83*** Population center (>20,000 people) 37.1 29.3 7.81*** 36.9 28.6 8.26*** ADMARC 8.1 7.3 0.85*** 8.6 7.5 1.18*** Daily market 8.7 5.9 2.79*** 7.7 6.2 1.54*** Weekly market 5.4 4.5 0.89*** 5.4 8.7 -3.27*** Food security Food insecure in the past 7 days 0.40 0.22 0.18*** 0.78 0.49 0.28*** Food insecure in the past 12 months 0.61 0.36 0.25*** 0.87 0.59 0.29*** Source: Authors’ calculation based on IHS3 and IHS4. Notes: a. Employment information is based on economic activities in the past week. *** indicates significant at 1%, ** significant at 5% and * significant at 10%. 3.3 Agriculture Practice of the Poor Farmers As poverty is mainly a rural phenomenon in Malawi and most of the poor engage in crop production, this section briefly explores agricultural practice of the poor farmers. It investigates gaps between poor and non-poor households in landholding, technology, and access to extension services. Table 5 summarizes the farming practice of the poor and non-poor, and 21 Table A.1 digs deeper by comparing farm households in the bottom 10 percent, bottom 40 percent, and top 60 percent of the national welfare distribution. The Farm Input Subsidy Program (FISP), one of the key agriculture policy interventions implemented by the Government of Malawi, has increased applications of fertilizer in the country relative to neighboring countries or the Sub-Saharan Africa average. Analysis of IHS data and international comparison shows that fertilizer application is relatively high in Malawi. For example, with the exception of Zambia, the amount of fertilizer used per unit of arable land is higher in Malawi than other countries in the region. It is also higher than the Sub-Saharan Africa average fertilizer application intensity (Figure A.6). According to IHS data, 60 percent of farmers have applied fertilizer, and the intensity of application was 55 kg per acre in 2016. However, adoption of yield-improving technologies such as fertilizer and pesticide is much lower among poor farmers. In 2016, only 52 percent of the poor farmers applied chemical fertilizer, compared to 71 percent of non-poor farmers. The intensity of fertilizer application, as measured by quantity of fertilizer per acre of cultivated land, is also lower among poor farmers (27 kg per acre) than non-poor farmers (46 kg per acre) for all crops. The same is true for maize. Although the use of pesticide is not very common in Malawi, an even smaller share of poor farmers (3 percent) use chemicals to control pests (Table 5). Even though fertilizer application is high in Malawi, there has been a noticeable decrease between 2010 and 2016, especially among poor farmers, and this might be a result of the 2015/16 FISP reform. Intensity of fertilizer application has declined by 67 percent among poor Malawian farmers (from 45 kg to 27 kg per acre on all crops). A 53 percent decrease in fertilizer application intensity has also occurred among non-poor farmers. The share of farmers who have applied chemical fertilizer has also declined during 2010–2016. This rather unfavorable trend in fertilizer application could be the result of the 2015/16 FISP reform, which increased farmers’ contribution from MWK 500 to MWK 3,500 per 50 kg fertilizer voucher and introduced private sector retailing/distribution of fertilizer in selected districts. Access to extension services, one of the mediums through which farmers learn about new technologies and farming practices, is quite limited in Malawi and more so for poor farmers. Relatively more farmers (36 percent to 44 percent) have received advice on the use of improved seed and fertilizer than other technologies such as irrigation and pest control. Extension advice on livestock management and crop marketing is also not commonly provided to farmers. However, the overall gap in access to extension services between poor and non-poor farmers is quite low (Table 5). The average maize yield of poor farmers was 38 percent lower than that of non-poor farmers in 2016, and this is possibly a reflection of their limited landholding, technology adoption, and access to extension services. The yield gap has increased by 9 percentage points since 2010. The maize yield of poor farmers also declined from 612 kg per acre in 2010 to 430 kg per acre in 22 2016, which happens to be a drought year. Even though the maize yield of non-poor farmers also declined between 2010 and 2016, the decrease was only 22 percent compared to the 42 percent for poor farmers. The average landholding of the poor farmers is 0.34 acres (19 percent) lower than that of the non-poor farmers in 2016. (Table 5). Table 5. Profile of the poor farm households 2010/11 2016/17 Non- Differ Non- Differ Poor poor ence Poor poor ence. Cultivated land area (in acre) 1.80 2.10 -0.38*** 1.50 1.80 -0.34*** Technology adoption Application of fertilizer 0.66 0.81 -0.15*** 0.52 0.71 -0.19*** Share of land treated with fertilizer 0.60 0.70 -0.13*** 0.51 0.67 -0.16*** Intensity of fertilizer use (kg/acre), all crop 45.00 70.70 -25.7*** 27.0 46.2 -19.2*** Intensity of fertilizer use (kg/acre), maize 51.30 79.30 -28.1*** 31.5 54.1 -22.6*** Application of pesticide 0.03 0.04 -0.01 0.03 0.05 -0.02*** Access to extension services: Improved seed use 0.25 0.34 -0.08*** 0.45 0.44 0.02 Pest control 0.12 0.18 -0.06*** 0.08 0.10 -0.02*** Fertilizer application 0.24 0.31 -0.08*** 0.36 0.38 -0.02 Irrigation 0.16 0.27 -0.11*** 0.20 0.23 -0.03*** Commodity market 0.07 0.13 -0.06*** 0.08 0.11 -0.03*** Livestock management 0.07 0.14 -0.07*** 0.07 0.13 -0.07*** Maize yield 612.1 855.9 -243.8** 429.8 698.7 -268.9*** Source: Authors’ calculation based on IHS3 and IHS4. Note: *** indicates significant at 1%, ** significant at 5% and * significant at 10%. The difference in agricultural practice persists between the bottom 10 percent and bottom 40 percent as well as the bottom 40 percent and top 60 percent farm households. Table A.1 presents a comparison of landholding, technology adoption, and access to extension services for these three groups. The bottom 10 percent of farm households have smaller plots, lesser adoption of yield-improving technologies, and limited access to extension advices compared to the bottom 40 percent. Similarly, those in the bottom 40 percent have lower access to technologies and extension services than the top 60 percent. The maize yield is also lower for farmers in the bottom of the welfare distribution. 4. Conclusion Due to recurring shocks and resulting weak economic growth, Malawi has not experienced meaningful poverty reduction since 2004. In fact, poverty has been on the raise in rural areas 23 and has reached 59.5 percent by 2016. Poverty remains a rural phenomenon. While urban poverty has declined significantly between 2004 and 2010, this progress has been stalled after 2010. Since 2010, consumption of households at the bottom of the welfare distribution has increased significantly. This has reduced inequality and ultra-poverty. Between 2010 and 2016, consumption of those in the bottom 10 percent and bottom 40 percent grew by 3.1 percent and 1.7 percent per year, while that of the top 60 percent shrank by 1.5 percent per year. The consumption growth experienced by those in the bottom of the welfare distribution is mainly the result of increases in private and public transfer of food, likely as part of the food security response following the 2016 drought. Decomposition of poverty changes indicates that favorable redistribution has contributed to poverty reduction after 2010, but lack of growth hinders progress in poverty reduction. For example, lack of growth increased rural poverty by 5.3 percentage points between 2010 and 2016 but a more equitable distribution of consumption has decreased poverty by 2.5 percentage points. Before 2010, on the other hand, favorable growth contributed to poverty reduction while regressive distribution hampered poverty reduction. Despite seemingly unchanging poverty since 2004, there is enormous seasonality in well- being, and poverty is much higher during lean season. Analysis of poverty and consumption across quarters of a given agricultural year (from harvest to lean season) shows that poverty is relatively low in the harvest season but it increases continuously and reaches its peak in the lean season. This seasonal variation in well-being is pronounced in drought years. The salient profile of poor Malawians include engagement in agriculture, particularly maize production; low human capital that reduces their access to rewarding employment; limited opportunity for nonfarm jobs, except causal labor (ganyu); remote locations that hinder their access to basic services; and high level of food insecurity. These poor farm households have a relatively smaller land plot compared to non-poor households. Even if the FISP has increased fertilizer application rate relative to neighboring countries, agriculture technology adoption is lower among poor farmers and is on the decline since 2010. As a result, crop yield is lower for poor farmers. 24 Appendix Figure A.1. Growth incidence during 2004–2016 (a) 2004–2010 in rural areas by region (b) 2010–2016 in rural areas by region 10 10 Rural North Rural Centre Rural North Rural Centre Rural South Rural South Annual growth rate (%) Annual growth rate (%) 5 5 0 0 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 -5 -5 -10 -10 Expenditure percentiles Expenditure percentiles -15 -15 The bottom 10% and 40% experienced a strong consumption growth during 2010–2016 (c) Annualized growth by region 6 Rural North Rural Centre Rural South 4 Annualized growth (%) Pro-poor 2 0 -2 Not pro-poor -4 -6 B10% B40% T60% B10% B40% T60% 2004-2010 2010-2016 Rural North -2.6 -1.5 -1.0 3.8 2.3 -0.6 Rural Centre -3.9 -2.2 0.5 3.6 1.4 -2.5 Rural South -2.5 -1.3 0.6 2.3 1.7 -1.1 25 Figure A.2. Density of consumption per capita 2010–2016 (in 2016 prices) National Rural Urban Figure A.3. Lorenz curves in urban centers, 2004–2016 100 2004 80 2010 Commulative consumption (%) 2016 Equitable distribution 60 40 20 0 0 10 20 30 40 50 60 70 80 90 100 Proportion of the population (%) 26 Poverty rate (%) Proportion of the poor by location of residence 0 2 4 6 8 10 12 0 100 10 20 30 40 50 60 70 80 90 Phalombe Lilongwe non-city Nsanje Mangochi Chitipa Dedza Machinga Kasungu Mulanje Machinga Thyolo Chiradzulu Chiradzulu Thyolo Chikwawa Mulanje Dedza Dowa Balaka Chikwawa Mangochi Phalombe Salima Ntcheu Source: Authors’ calculation based on IHS2–IHS4. Source: Authors’ calculation based on IHS2–IHS4. Nkhatabay Mchinji 2004 2004 Karonga Salima Zomba Non-City Balaka 27 2010 2010 Ntcheu Zomba Non-City Rumphi Nsanje Mwanza Nkhotakota 2016 2016 Ntchisi Lilongwe City Nkhotakota Karonga Kasungu Chitipa Figure A.5. Poverty rate by district Mchinji Nkhatabay Dowa Ntchisi Lilongwe non-… Blantyre Neno Rumphi Mzimba Mzimba Blantyre Figure A.4. Distribution of the poor population across districts Neno Likoma Blantyre City Lilongwe City Mwanza Zomba City Mzuzu City Mzuzu City Zomba City Blantyre City Likoma Figure A.6. Fertilizer application is generally high in Malawi, by regional and Sub-Saharan Africa standard 60 Sub-Saharan Africa Malawi Tanzania Zambia Mozambique Uganda Zimbabwe 50 (kg per hectare of arable land) Fertilizer application intensity 40 30 20 10 0 2008 2009 2010 2011 2012 2013 2014 2015 Source: WDI. 28 Figure A.7. Poverty rate by districts Source: Prepared by authors. 29 Table A.1. Difference in agriculture practice of farmer households in the bottom and top consumption deciles 2010/11 2016/17 Difference Difference Difference between Difference between between Bottom 10% between Bottom 10% Bottom and Bottom Bottom 40% and Bottom 40% and ^ 40% and Top 60% 40% Top 60% Cultivated land area (in acre) --- --- --- --- Technology adoption Application of fertilizer --- --- --- --- Share of land treated with fertilizer --- --- --- --- Intensity of fertilizer use (kg/acre), all crop --- --- - --- Intensity of fertilizer use (kg/acre), maize --- --- - --- Application of pesticide --- --- Access to extension services: Improved seed use --- --- --- Pest control --- -- --- Fertilizer application --- -- --- Irrigation --- --- -- --- Commodity market --- --- --- Livestock management --- --- - --- Maize yield --- - - Source: Authors’ calculation based on IHS3 and IHS4. Note: The symbols ---, --, and – (and +++, ++, and +) indicate significant negative (positive) difference between the group at 1 percent, 5 percent, and 10 percent, respectively. a. The bottom 40 percent in the comparison between bottom 10 percent and bottom 40 percent excludes those in the lowest decile. 30