68418 v7 I. Introduction Analysis of the Tajikistan Living Standards Survey (TLSS) for 1999 and 2003 shows that rural poverty exceeds urban poverty by a wide margin. In both 1999 and 2003, rural poverty was more pervasive, deeper and more severe than urban poverty. Regionally, rural poverty was highest and deepest in the cotton growing areas of Khatlon and Soghd, and these two regions also made the least progress in terms of poverty reduction. Rural poverty fell by 6.9% and 12.8% in Khatlon and Soghd respectively, from 1999 to 2003, versus an overall decline of 15.9% in rural poverty. Rural poverty declined much faster in RRS and GBAO. Indeed rural poverty levels were less than urban poverty levels in RRS in 2003, indicating the advantages of a more diversified production base and proximity to a large urban market. Analysis by location adds support to this conclusion, in that peri-urban households had both the lowest level of rural poverty and the highest rate of poverty reduction (from 72.9% in 1999 to 49.7% in 2003). The valley farmers who produce Tajikistan’s cotton had both the highest levels of rural poverty and the lowest rate of rural poverty reduction, with poverty rates of 83.4% in 1999 and 68.6% in 2003. Aggregate measures of household food consumption give a misleading impression that food security is not a major issue in rural areas. Average daily calorie consumption per capita exceeded minimum dietary requirements (2,170 kcal) in all regions and all locations in 2003, except GBAO (2,008 kcals). This perception of adequate food security is inconsistent with high observed levels of child malnutrition, however, and the more general indicators of rural poverty. Disaggregated analysis shows that food consumption is well below minimum requirements in the lowest income quintile (Q1), and marginally below minimum requirements in Q2. Protein consumption is also very low among Q1 households – indicating that food insecurity is due to inadequacies in both the level and quality of food consumption. Moreover, cereals provide over 80% of the protein requirements for lower income households. These trends are consistent across all regions’ and locations and show that food insecurity is indeed a major issue for the poorest rural households. Extremely low access to basic services is a further indicator of the magnitude of rural poverty. Twenty-two percent of rural households had access to piped water in 2003, 10% had access to piped heating, and only 5% had access to telephones. Poor access to clean water and an inadequate, low quality diet are the main causes of high child malnutrition. Urban households have significantly better access to these basic services. 1.1 The Determinants of Rural Poverty and Poverty Reduction Empirical analysis shows that a wide range of factors determine rural poverty status, many of which have direct policy relevance. Of the “non-policy” parameters, two household demographic factors have a strong impact on poverty at all levels of analysis. Poverty rises with household size and falls with education, for both urban and rural households, and in all regions. Pronounced regional differences were also observed. Rural households were most likely to be poor in GBAO and Soghd, and urban households were most likely to be poor in RRS. Both rural and urban households were equally likely to be poor in Khatlon, indicative of the depressed nature of Khatlon’s regional economy. Increased distance to major urban centres also increases poverty, consistent with the impact of high transaction costs on economic activity. Income levels and composition had a major impact on poverty. Descriptive analysis shows that rural households that rely solely on farm income are much more likely to be poor than households with income from both farm and non-farm business. At a national level, the marginal effect of non-farm income on rural poverty was higher than any other policy-related variables, although further analysis shows that the magnitude of this effect differs widely by region. Non-farm business activity was not observed among TLSS sample households in GBAO, and very low levels of activity were observed in Khatlon. In contrast non-farm income, was more common in RRS and Soghd, and had a strong positive impact on poverty reduction. Wage employment levels had little observable impact on rural poverty in all regions, a result attributed to very low prevailing wage rates. 1 Given that poverty occurs more widely among households that rely solely on farm income, it is pertinent to consider the impact of measures to raise farm income. Analysis of cotton production, which is strongly linked to high poverty levels, shows that poverty falls in response to higher cotton farm wages and increased cotton yields. The marked difference between the level of response to these two factors is also instructive. The response elasticity for cotton wages is strong, with a 10% increase in wages leading to a 3% reduction in rural poverty. This is income that farmers receive directly. However a 10% increase in cotton yields during the same period, only leads to a 1% reduction in rural poverty. This discrepancy suggests that farmers are unable to translate increased cotton yields and output into a substantial increase in income. Remittance income is observed in 31% of rural households versus 23% of urban households. Higher remittance incomes are also associated with lower poverty levels, although causality appears to flow from wealth to remittances rather than in the opposite direction. Most rural migrants originated from RRS (the least poor region), and both migration and remittance income are highest in the upper income quartile. This suggests that migrants from less poor households have a greater capacity to meet the costs of migration (immigration, travel, job search etc), and are perhaps able to find work more easily due to better education. Unlike most countries land and livestock ownership have a very weak impact on rural poverty in Tajikistan. Land reform has resulted in a more even distribution of land ownership, but this redistribution has not reduced rural poverty, an outcome attributed to inappropriate land privatization and land use policies. Initially, land privatization emphasized the creation of collective dehqon farms, and the issue of Land Share Certificates that named the joint owners of the collective but didn’t give them individual rights. As management decisions were still made by former brigade leaders, and the remaining “owners” continued to work as paid laborers, this approach did little to change incentives or income. More recently, land privatization has allowed individuals to take their share of the collective dehqon and create individual farms. The owners of these farms receive a Land Use Title which gives them full land use rights. They also have greater freedom to apply these rights. However the ability to use their land is still heavily circumscribed by local government. In cotton areas local authorities coerce farmers to use 70% of their arable land for cotton, and in all areas local government has wide discretionary powers to appropriate land in the event that it is used “irrationally”. Thus, while land privatization has resulted in a more equitable distribution of land, farmers still lack adequate security of tenure, the freedom to produce and market as they choose, and the incentive to raise production and invest. Its impact on poverty has been negligible as a consequence. Regression analysis shows that the issue of land share certificates has no effect on poverty levels, and that the impact of land use titles is extremely small. A one percent increase in the number of land use titles issued results in a 0.06% decrease in the proportion of households below the poverty line. Livestock are widely owned. Within regions, there is no major difference in ownership levels across income quintiles, and these differences actually narrowed slightly from 1999 to 2003. Greater differences occur between regions, with lower levels of livestock ownership in RRS and high levels of ownership in GBAO. Most of the growth in livestock output value is due to higher prices, which lowers the ability of poor households to increase animal protein consumption. Livestock productivity increased slightly but remained very low and livestock numbers increased mainly in mountain areas where grazing land is more abundant. Other than in GBAO, these changes generated limited benefits for lower income rural households, particularly in the valley areas where rural poverty is concentrated. 2 II. Profile of Income Poverty in Tajikistan in 2003 The Tajikistan Living Standard Survey (TLSS) data for 1999 and 2003 is used to examine poverty in the country. Three poverty measures are employed in the study to understand the evolution of poverty in the five regions (oblasts) of Tajikistan. These include the headcount index, the poverty gap, and the squared poverty gap. The headcount index is the percentage of the population living in households with income per capita below the poverty line. However, the headcount index ignores the amounts by which the expenditures of the poor fall short of the poverty line. Hence, the poverty gap index is a better measure as it gives the mean distance below the poverty line as a proportion of the poverty line is also computed. Here the mean is taken over the whole population, counting the non-poor as having zero poverty gaps. For example, a poverty gap of 20 percent means that the average expenditure of the poor is 20 percent below the poverty line. The squared poverty gap index which indicates the severity of poverty is computed by weighting the individual poverty gaps by the gaps themselves, so as to reflect inequality amongst the poor (Foster- Greer-Thorbecke, 1984). The squared poverty gap index tends to exhibit useful analytical properties, because it is sensitive to changes in distribution among the poor. Specifically, while a transfer of expenditures from a poor person to a poorer person will not change the headcount index or poverty index, it will decrease the squared gap index. The results from the analysis of the three indices are presented in Table 1. The findings indicate that there are regional differences in poverty. These differences are quite large and striking in the Republic. Overall, the proportion of the population living below the poverty line (Headcount index) declined by almost 20 percentage points from about 77 percent in 1999 to about 57 percent in 2003. However, differences were recorded across the individual oblasts. The lowest poverty rate in 1999 was recorded in the capital Dushanbe followed by RRS while poverty was highest in the GBAO and Khatlon oblasts. The poverty measures for 2003 all report the same ranking. The lowest poverty level in 2003 was recorded for the capital Dushanbe, with about 33 percent of the population falling below the poverty line, followed by the RRS, with 41 percent. The highest poverty levels were in Khatlon and Soghd, which had poverty levels of 76.13 percent and 62.3 percent, respectively. Both figures were above the national average of 56.8 percent. The two oblasts are the major cotton growing areas in the country. For the country as a whole, all the three poverty measures indicate that rural poverty exceeds urban poverty for both 1999 and 2003, and by a wide margin. Thus, the extent of poverty, the depth of poverty and the severity of poverty are all higher in rural than urban areas. Even at the regional level, rural poverty appears to be higher than urban poverty, with the notable exception of the RRS. In RRS rural poverty declined substantially from a headcount index of 70.6 to 40.0 percent compared to urban poverty, which declined from 56.3 to 45.0 percent. This resulted in a higher poverty rate in urban RRS. What is also obvious from the poverty indices is the enormous variation across the country. Another interesting classification of households according to location considers geographical differences. In this regard, households in rural areas can be classified into three categories: valley areas, mountainous areas and peri-urban. The poverty profiles for these different locations are reported in the last three rows of Table 1. It can be observed from Table 1 that households in valley areas have the highest poverty levels. Households in these areas recorded a poverty level of 68 percent, a level that is higher than the national average for rural areas. On the other hand, the poverty level of households in mountain areas was 60 percent. As expected poverty levels was lowest in the city centers, followed by peri-urban areas. The poverty gap was also highest in the valley areas, with an income shortfall of 26 percent. The income shortfall was lowest in the city centers, with a poverty gap of just 12 percent. 3 Table 1: Poverty Measures (%) in Different Regions of Tajikistan, 1999 and 2003 2003 1999 P0 P1 P2 P0 P1 P2 Headcount Poverty Poverty Headcount Poverty Poverty poverty gap severity poverty gap severity Dushanbe, total/urban 33.33 10.13 4.59 53.98 17.31 7.45 GBAO, total 55.00 17.82 7.71 88.75 43.65 25.17 GBAO, urban 37.50 9.86 3.95 93.75 46.50 25.99 GBAO, rural 60.83 20.48 8.96 87.50 42.94 24.96 RRS, total 40.86 12.87 5.67 68.98 26.92 13.73 RRS, urban 45.00 12.98 4.92 56.25 21.61 10.40 RRS, rural 40.00 12.85 5.82 70.57 27.58 14.14 Soghd, total 62.30 21.25 9.84 78.78 32.93 16.93 Soghd, urban 51.00 18.80 9.30 75.00 31.17 16.37 Soghd, rural 67.56 22.40 10.10 80.32 33.65 17.15 Khatlon, total 76.13 31.43 15.84 83.95 37.00 20.01 Khatlon, urban 70.91 30.79 16.43 82.03 34.10 17.92 Khatlon, rural 77.50 31.60 15.69 84.38 37.64 20.47 Tajikistan, total 56.78 20.28 9.59 76.70 32.12 16.81 Tajikistan, urban 44.67 15.61 7.52 68.75 26.98 13.60 Tajikistan, rural 63.75 22.97 10.78 79.67 34.04 18.01 Valleys, rural 68.56 25.79 12.48 83.38 36.21 19.28 Mountainous areas, rural 60.20 20.24 8.90 76.91 33.05 17.56 Peri-urban areas, rural 49.69 17.92 8.79 72.92 28.90 14.78 Source: TLSS (1999, 2003), own calculations. 2.1 How has poverty evolved over the five years? Poverty generally declined from 1999 to 2003, but the rate of poverty reduction differed across the country. There was a higher decline in urban poverty than in rural poverty. While the headcount index declined by 24 percent in urban areas, the corresponding figure was 16 percent in rural areas. The pattern was similar for the regional changes in poverty, with the notable exception of the RRS, where rural poverty rate declined at a greater percent that urban poverty. The changes in poverty levels between 1999 and 2003 are also presented for three geographical regions of rural Tajikistan. The figures reveal that while the decline in the headcount poverty was highest in peri-urban areas, the poverty gap and poverty severity declined mostly in mountainous areas. 4 2.2 Income Distribution Having looked at the development of poverty levels between the two periods under consideration, it will be interesting to investigate how the distribution of income has evolved over this period. The most widely used index in measuring income distribution or inequality is the Gini-Coefficient. The computed values for 1999 and 2003, as well as the change in inequality are presented in Table 2. The figures indicate a slight increase in inequality for the country. A closer look reveals that whereas inequality in urban areas actually declined marginally from 0.377 to 0.375, it increased in the rural areas from 0.309 to 0.328. At the regional level, inequality increased slightly for GBAO and Khatlon, but declined marginally for Dushanbe, RRS and Soghd. With the notable exception of GBAO, inequality appears to be higher in urban areas than rural areas. Table 2: Income Distribution as Measured by the Gini-Coefficients for Rural and Urban Areas, 1999-2003 Change in Gini 2003 1999 Coefficient Urban Rural Urban Rural Urban Rural Dushanbe 0.3697 0.3770 -0.0072 GBAO 0.2658 0.3105 0.2343 0.3055 0.0315 0.0049 RRS 0.3448 0.2996 0.3901 0.3265 -0.0454 -0.0269 Soghd 0.3572 0.2970 0.3803 0.3006 -0.0230 -0.0036 Khatlon 0.3723 0.3463 0.2835 0.2857 0.0888 0.0606 Tajikistan 0.3745 0.3280 0.3769 0.3085 -0.0024 0.0195 Total 0.3560 0.3346 0.0215 Source: TLSS (1999, 2003), own calculations. 2.3 Determinants of National and Regional Poverty A better understanding of the spatial distribution of poverty and the characteristics of the poor are crucial in determining the likely impact of broad national economic development trends and specific policies on the disadvantaged. In this section, probit regression analysis is performed to determine the nature and magnitudes of selected factors on poverty at the national and regional level. Thus, a regression is first performed to examine the factors that influence the probability of households in Tajikistan being poor. In a second step, the determinants of poverty at the regional level are analyzed to ascertain the nature and magnitudes of the factors that affect poverty at the regional level. The analysis at the national level makes use of the data in the 2003 Tajikistan Living Standard Survey, consisting of observations from 4160 households. In addition to obtaining estimates from pooled data of total households in the sample, each oblast is analyzed separately to determine the factors influencing poverty at the regional level. The poverty assessment update reported estimates over all households, using quantile regressions. We are more interested in examining the nature and magnitude of the determinants of poverty at the individual regions or oblasts. The variables used reflect households’ access to information, regional labor markets as well as schooling conditions and distances to central markets. To control for regional differences not covered by these variables, dummy variables are used for the five administration oblasts. The capital Dushanbe is the omitted 5 category from the estimation and is therefore the base. Tables 3a and 3b present data for selected demographic and location characteristics of the households in the sample by poverty status and place of residence. The presentation is done separately for rural and urban areas, as well as for the five oblasts. While Table 3a shows the household characteristics for rural and urban areas differentiated according to poverty status, Table 3b shows the differences between the individual oblasts, also differentiated according to poverty status. Overall, the data in both Tables reveal some interesting features of the households, although the differences between poor and nonpoor households are not very large. Nevertheless, it is evident in Table 6a that poor households have on average larger household sizes, relative to their non-poor counterparts. An interesting observation is the difference in the share of unemployed household members between the rural and urban households. While more than 41 percent of the household members in urban areas were unemployed, the corresponding figure for rural areas was about 26 percent and 28 percent for poor and nonpoor households. Table 3a: Selected Demographic and Household Characteristics of Households in Tajikistan by Poverty Status and Type of Region rural Urban Variable nonpoor poor nonpoor Poor Number of household members 6.26 7.41 4.30 5.98 Share of female hh-member 0.48 0.50 0.54 0.52 Share of hh-member aged up to 5 0.12 0.16 0.10 0.14 Share of hh-member aged from 6 to 15 0.22 0.27 0.19 0.24 Share of hh-member aged over 64 0.07 0.05 0.08 0.07 Share of unemployed hh-members 0.28 0.26 0.41 0.41 Average years of schooling per hh- member 10.12 9.70 11.56 10.04 Household receives transfers from family members or other persons (number of sources) 0.24 0.22 0.31 0.30 Household receives transfers from organizations (number of sources) 0.24 0.26 0.18 0.23 Household receives pension payments (dummy) 0.41 0.40 0.34 0.40 Household receives allowances (dummy) 0.03 0.04 0.02 0.04 Household operates nonfarm business 0.04 0.02 0.04 0.04 Distance to oblast administration center (km) 66.67 88.76 21.07 39.65 Distance to raion administration center (km) 15.58 18.48 3.04 3.01 Distance to capital (km) 250.80 278.73 142.34 181.67 Share of workers in non family business 0.15 0.15 0.23 0.14 Number of part-time workers 0.46 0.68 0.45 0.41 Number of full-time workers 0.37 0.38 0.42 0.35 Land owned by hh (Sotka) 32.33 27.48 2.74 2.71 Livestock owned by hh (LSU) 17.96 14.39 7.48 4.25 Source: Own Calculations from TLSS (2003). The data show about 26 percent of household members above the age of 15 participated in off-farm work. However, the Poverty Assessment Update of 2004 argues that paid labor did not actually appear to keep households out of poverty, because of the generally low wages. The average wage earned from off-farm activities was about 45 Somoni per month, while farm activities earned about 30 Somoni per month. There appears to be no significant difference in the labor participation patterns of urban and rural households. While 25.9 percent of rural household members were involved in off-farm activities, the corresponding figure for their urban counterparts was 26.6 percent. However, the wages earned 6 differed significantly between urban and rural household members, with rural household members earning about 33 Somoni on average, while their counterparts in the urban areas earned 69 Somoni. Figure 1a: Income Sources and Income Combinations, National Average Household recieves income from... 1.1% 10.4% 0.3% 0.8% 28.0% 20.9% 1.1% 37.3% agricultural production only agricultural production and paid work agricultural prod. + non-farm enterprise paid work only non-farm enterprise only non-farm enterprise and paid work all three income sources recieves only transfer Figure 1b: Income Sources and Income Combinations, Rural Average Household recieves income from... 1.3% 4.5% 0.1% 0.3% 5.0% 1.2% 37.8% 49.8% agricultural production only agricultural production and paid work agricultural prod. + non-farm enterprise paid work only non-farm enterprise only non-farm enterprise and paid work all three income sources recieves only transfer Source: TLSS (2003), own calculations. 7 As is evident in Figures 1a and 1b, agricultural production appears to be the most important income source for households in the country. Over one third of Tajikistan’s rural households are solely engaged in agricultural production (37.8 percent), while nearly half of the surveyed rural households (49.8 percent) combine agricultural production with paid work. Not surprising, these figures are considerably lower when looking at national average. Households engaged in agricultural production are partly producing for subsistence and partly for commercial purposes. Around 21 percent of all households and 5 percent of rural households generate their income solely from paid labor. With regards to poverty status, the data generally reveal that poor households are more likely to have larger household sizes, lower average years of schooling, receive lower amounts of remittances from family members, receive more pension payments, and are far from administrative centers and oblast’s capital in terms of distance. The oblast differentials are more apparent in distance to oblast center, access to transfers from organizations and remittances from family members, as well as family structure variables. Poorer households in rural areas generally depend on agriculture for their livelihoods, while the better off combine crop with livestock holdings and widespread engagement in non-farm self-employment activities. Raising productivity in agriculture could therefore be a primary rural development goal. 8 Table 3b: Selected Demographic and Household Characteristics of Households in Tajikistan by Poverty Status and Oblast Tajikistan Khatlon Soghd GBAO RRS Dushanbe Variable Nonpoor Poor Nonpoor Poor Nonpoor Poor Nonpoor Poor Nonpoor Poor Nonpoor Poor Number of household members 5.34 7.00 5.55 7.50 4.82 6.38 5.36 7.03 7.07 8.43 4.14 5.50 Share of female hh-member 0.51 0.51 0.51 0.50 0.50 0.50 0.51 0.52 0.49 0.50 0.54 0.53 Share of hh-member aged up to 5 0.11 0.15 0.13 0.17 0.09 0.14 0.11 0.12 0.14 0.16 0.11 0.16 Share of hh-member aged from 6 to 15 0.21 0.26 0.24 0.29 0.20 0.24 0.19 0.23 0.24 0.27 0.18 0.25 Share of hh-member aged over 64 0.07 0.06 0.06 0.04 0.08 0.06 0.08 0.06 0.05 0.07 0.08 0.08 Share of unemployed hh-members 0.34 0.30 0.33 0.28 0.38 0.32 0.25 0.24 0.25 0.28 0.44 0.42 Average years of schooling per hh-member 10.78 9.80 10.31 9.80 10.79 9.94 11.20 10.60 9.78 8.57 11.85 9.92 Household receives transfers from family members or other persons (number of sources) 0.28 0.24 0.23 0.20 0.25 0.23 0.44 0.32 0.22 0.24 0.31 0.33 Household receives transfers from organizations 17 (number of sources) 0.21 0.25 0.09 0.14 0.12 0.17 0.95 1.02 0.14 0.19 0.08 0.12 Household receives pension payments (dummy) 0.38 0.40 0.39 0.36 0.35 0.39 0.54 0.53 0.40 0.52 0.30 0.32 Household receives allowances (dummy) 0.02 0.04 0.02 0.03 0.03 0.03 0.01 0.07 0.04 0.08 0.02 0.03 Distance to oblast administration center (km) 45.34 74.64 75.40 81.17 81.66 104.42 108.30 107.68 0.47 1.14 0.16 0.32 Distance to raion administration center (km) 9.71 14.04 14.22 14.63 13.39 15.21 9.56 20.43 10.37 12.05 2.62 2.58 Distance to capital (km) 200.07 250.83 147.53 159.65 368.14 380.88 532.80 545.28 73.81 69.79 4.30 3.23 Share of workers in non family business 0.18 0.15 0.22 0.19 0.18 0.14 0.18 0.10 0.12 0.09 0.23 0.13 Number of part-time workers 0.46 0.60 0.54 0.82 0.38 0.56 0.50 0.37 0.53 0.47 0.40 0.39 Number of full-time workers 0.39 0.37 0.65 0.56 0.37 0.26 0.44 0.33 0.20 0.22 0.44 0.31 Land owned by hh (Sotka) (rural regions only) 32.33 27.48 43.42 33.49 23.57 20.43 17.68 19.81 39.27 35.52 --- --- Livestock owned by hh (LSU) (rural regions 17.96 14.39 23.85 16.42 11.81 9.38 21.17 18.17 18.22 17.25 --- --- only) Source: Own Calculations from TLSS (2003). III. Household Assets and Rural Poverty 3.1 Impact of land and livestock assets on rural poverty Recent understandings of rural poverty place considerable emphasis on the ownership or access to assets that can be put to productive use as the building blocks by which the poor can help themselves out of poverty. For example, the World Food Program (2005) suggests that one of the main causes of food security in rural areas of Tajikistan is limited access to livelihood opportunities in both the agricultural sector and employment/labor market. Hence, for meaningful policy recommendations, it is important to examine the structure of ownership among the rural families. Land, livestock and labor (active adults in the household) are the key assets of rural families in Tajikistan. As a result of its crucial role in the household economy, Action Against Hunger, for example, focuses on building livestock assets of households as part of its food security program in the country. Since a large share of the rural labor force is engaged in agriculture, land ownership and food security remain linked due to the poor incomes earned by workers in the cotton sector. As a result of the extremely low incomes earned, a household's own food production (and the amount of land a family has access to) is critical to their food security. We therefore examine the relationship of asset holding to relative ability to generate a viable living by comparing assets across per capita income quintiles. Information on land holdings is particularly attractive because possession of land could be a major determinant of individuals’ productive capacity and their ability to invest. Figure 2: Land Assets by Expenditure Quintile and Geographical Area, 1999 and 2003 100.00 1999 90.00 2003 80.00 70.00 60.00 Sotka 50.00 40.00 30.00 20.00 10.00 0.00 Mean Bottom Top Quintile Valley areas Mountaineous Peri-urban Quintile areas areas Source: TLSS (1999, 2003), own calculations. Comparisons of average land holdings are presented in Table 4, as well as Figures 2 and 3 for rural households. Land distribution shows differences across oblasts and income quintiles, as well as geographical areas in both 1999 and 2003. The distribution in 1999 indicates that the highest quintile owned across all households more than three times the amount owned by the lowest income quintile. Considering the fact that these are averages, with wide variables around the mean, one could safely conclude that households in the wealthy class have acquired extensive amounts of land. Although the ownership structure changed immensely in 2003, the disparities between income groups still remained. The average land owned by the highest quintile averaged 38 Sotkas, while that owned by the lowest quintile averaged about 26 Sotkas. In the RRS oblast where land distribution was highly concentrated in 1999, the ownership structure for 2003 shows that on average, the highest income quintile still owned more than two and half times land than the lowest quintile. With regard to geographical locations, the distribution indicates a dramatic change in ownership in peri-urban areas in 2003, compared to 1999. Specifically, the average land owned declined from 70 Sotkas to just over 20 10 Sotkas. On the other hand, the observed changes in valley and mountainous areas appear to be similar, with both changes in land ownership being less than the peri-urban areas. Table 4: Mean Land Owned and Cultivated (in Sotkas) by Expenditure Quintiles and Oblast, 1999 and 2003 Expenditure quintiles Region Year 1 2 3 4 5 Mean Tajikistan 1999 28.12 73.53 41.60 56.28 89.04 56.28 2003 26.44 28.19 26.52 29.80 38.18 29.24 GBAO 1999 24.76 32.27 37.85 16.00 30.00 28.44 2003 20.72 20.74 17.43 18.28 17.18 18.98 RRS 1999 15.69 164.09 34.62 32.58 114.81 69.75 2003 17.11 44.07 36.96 37.86 44.25 37.77 Soghd 1999 15.95 26.95 26.47 20.16 22.14 22.58 2003 25.72 21.31 14.43 18.58 32.89 21.44 Khatlon 1999 40.80 73.33 59.84 108.86 127.63 75.66 2003 30.28 32.67 41.92 42.04 46.05 35.73 Source: TLSS (1999, 2003), own calculations. Given that the samples for 1999 and 2003 are not similar, a comparison was also made for household distribution by land owned in the two periods. This is shown in Figure 3. The distribution shows that in 2003, about 10 percent of households owned no land at all and a further 10 percent owned holdings up to 5 Sotkas. An interesting observation is the ownership pattern in 1999 and 2003. While over 30 percent of household owned more than 30 Soktas of land in 1999, a little over 20 percent are reported to have owned more than 30 Sotkas. This observation could be attributed to the land reform and redistribution exercise that is currently taking place. It implies that land might have been redistributed from households with larger ownership to landless households, or those with just small land ownership. The disparities even appear to be more significant, when comparison is made between income quintiles within the different oblasts. For example, in RRS the highest quintile owned more than seven times the amount of land owned by the lowest quintile, while the corresponding difference in Khatlon was about three times in favor of the highest income quintile. On the other hand, the observed variations in land ownership across income quintiles do not appear to be substantial in the GBAO and Soghd oblasts. The patterns of livestock holdings found in the five regions, which are shown in Table 5 and Figure 4, are also quite interesting. When livestock holdings are compared across quintiles (Table 5) a conventional picture emerges of a substantial difference in livestock ownership between the top and the other income groups. Specifically, a comparison across quintiles over 1999 and 2003 shows that the highest income quintile owned across all households more than seventy percent the amount owned by the lowest income quintile. This does not appear to be an especially large differential between the better-off and the poor for this society. However, it is important to note that these are average figures that are associated with wide variations around the mean, with several households possessing no livestock. Both Figure 4 and Table 5 indicate that in all oblasts livestock ownership is widespread, although there are sharp distinctions in ownership between oblasts. For example, in the Khatlon oblast, the highest income quintile owned almost twice as much livestock as the lowest income quintile in 2003, while this difference was not substantial in the GBAO oblast. It is also noteworthy that in some cases mean holding of the top quintile is lower than that of the third and fourth quintiles. In Khatlon, for 11 example, the mean holding of the top quintile was lower than that of the third and fourth quintiles in 2003. While livestock assets per household were almost the same in the different geographical areas in 1999, this pattern has changed in 2003. In contrast to peri-urban households which owned less livestock in 2003, compared to 1999, households in mountainous areas on average own nearly 5 livestock units more than in 1999. Livestock assets in the valley areas appear not to have changed considerably over the two periods. Figure 3: Distribution of Land Owned by Households, 1999 and 2003 35 1999 2003 30 25 % of Households 20 15 10 5 0 0 0-5 5-10 10-15 15-20 20-25 25-30 >=30 Sotka Source: TLSS (1999, 2003), own calculations. Table 5: Livestock Assets (Livestock Units) by Expenditure Quintile and Oblast, 1999 and 2003 Expenditure quintiles Region Year 1 2 3 4 5 Mean Tajikistan 1999 10.11 14.81 14.03 14.93 19.38 14.43 2003 11.80 14.13 18.05 16.58 19.37 15.68 GBAO 1999 19.29 15.28 18.25 6.56 41.47 18.43 2003 17.05 17.83 20.20 20.86 21.56 19.21 RRS 1999 9.85 15.03 13.42 13.95 15.93 13.82 2003 16.71 12.34 20.66 14.41 21.85 17.86 Soghd 1999 8.29 12.57 10.59 12.48 14.92 11.67 2003 7.16 10.46 10.47 11.49 11.74 10.16 Khatlon 1999 9.96 16.24 16.92 18.43 26.47 16.46 2003 12.18 16.49 25.77 23.87 22.99 18.10 Source: TLSS (1999, 2003), own calculations. 12 Figure 4: Livestock Assets by Expenditure Quintile and Geographical Area, 1999 and 2003 25.00 1999 2003 20.00 its (LSU) 15.00 tockun 10.00 Lives 5.00 0.00 Mean Bottom Top Quintile Valley areas Mountaineous Peri-urban Quintile areas areas Source: TLSS (1999, 2003), own calculations. 3.2 Impact of land and livestock assets on rural poverty In view of the significance of land and livestock holdings in combating rural poverty, an analysis was undertaken to examine how land and livestock assets, access to nonfarm activities, as well as other household and location variations influence rural poverty in the country. Estimations were carried out for rural areas in the entire country, and then for rural areas in the individual oblasts, using probit models. The poverty line used in the analysis is $2.15 per day. The estimated results and marginal effects are presented for the entire country and individual oblasts in Tables 6 and 7, respectively. The results in Table 6 clearly show that the amount of land owned by a household influences its poverty status, although the magnitude of the impact is very small. Specifically, households possessing greater amounts of land are less likely to be poor, while households without any land have a higher probability of being poor. Similarly livestock assets possession tends to be a statistically significant determinant of poverty, whereby household that possess livestock holdings are less likely to be poor. Although the marginal effects of both land and livestock holdings are quite small, what is interesting is the finding that the impact of livestock holdings is twice that of land holdings, a result that is different from the experience of other countries. For example, the study by Haddad and Ahmed (2002) on poverty in Egypt revealed that the amount of land owned had a much higher impact on poverty reduction than the value of livestock assets of a household. Normally land appears to be a more important determinant of rural poverty than livestock holdings, but probably because the full impact of land possession is yet to be realized. In addition, livestock is easily used to generate income in the current economic condition. However, these two findings still confirm the notion that helping households to accumulate land and livestock assets is an important policy instrument in any poverty alleviation exercise. Another interesting finding relates to participation in nonfarm activities and poverty. Besides being statistically significant, the marginal effect of the variable for nonfarm business is quite high compared to the other policy variables, clearly showing that households with access to nonfarm activities are less likely to be poor. This finding supports the notion that nonfarm income could be used as a route to improve incomes among the poorest. As pointed out by Reardon (1997), inequality in access to scarce land translates to inequality to nonfarm business opportunities because agricultural cash incomes and the confounding land wealth and political pull are determinants of nonfarm business starts. Moreover, in areas where credit constraints are binding, households generally have to rely on earnings from cash crops or savings to invest in non-cropping activities. 13 Table 6: Results for the Determinants of Poverty of Rural Households, 2003 Coefficient z-value Marginal Effects Number of household members 0.1162 (10.17) 0.0423 Share of female hh-member 0.3028 (1.61) 0.1102 Share of hh-member aged up to 5 1.2763 (5.65) 0.4647 Share of hh-member aged from 6 to 15 0.9832 (5.41) 0.3580 Share of hh-member aged over 64 0.7331 (1.97) 0.2669 Share of unemployed hh-members -0.0312 (-0.20) -0.0114 Average years of schooling per hh-member -0.0944 (-5.94) -0.0344 hh receives transfers from family members or -0.0581 (-0.98) -0.0212 other persons (Dummy) hh receives transfers from organizations 0.1380 (1.77) 0.0503 (Dummy) hh receives pension payments (dummy) -0.0596 (-0.78) -0.0218 hh receives allowances (dummy) 0.4158 (2.50) 0.1363 hh operates nonfarm business -0.5425 (-3.13) -0.2107 Land owned by hh (Sotka) -0.0016 (-3.10) -0.0006 Livestock owned by hh (LSU) -0.0042 (-4.16) -0.0015 distance to oblast administration center (km) 0.0009 (1.90) 0.0003 distance to raion administration center (km) 0.0014 (0.71) 0.0005 distance to capital (km) 0.0005 (2.57) 0.0002 Khatlon (Dummy variable) 0.5170 (3.73) 0.1796 Soghd (Dummy variable) 0.2520 (2.09) 0.0896 RRS (Dummy variable) -0.6884 (-4.31) -0.2622 Constant -0.2953 (-1.12) Likelihood Ratio 521.27 Pseudo R2 0.1742 Number of Observations 2296 Source: TLSS (2003), own calculations. Many studies show that rural households in Tajikistan generally find it difficult to gain access to credit. A recent survey of rural households by World Food program (2005) showed that less than 40 percent of the households interviewed had access to credit. However, there appeared to be regional differences in that 80 percent of the sample communities in the survey had access to credit as compared to only about one-third in each of the other regions. An interesting observation from the study was the fact that about half of the communities in GBAO with access to credit could obtain credit from the bank or credit union or NGOs, whereas in Soghd, 82 percent of the respondents obtained credit from relatives and friends. Given the crucial role that access to institutional credit facilities could play in promoting income generating activities in the rural areas, it is important to strengthen rural financial intermediaries to increase the access of rural households to credit. Such a measure could go a long way to help them increase agricultural production and also diversify into nonfarm activities to increase their incomes and get out of poverty. The variables for distance to administrative center and capital are used to capture transaction costs involved in searching for information about employment and other income generation activities, since areas closer to the regional capital or administrative centers offer the best trade and employment 14 opportunities. The positive and statistically significant coefficients for the two variables suggests that households located in remote areas face higher transaction costs in income generating activities and are therefore more likely to be poor. The dummies for oblasts also indicate that households in Khatlon and Soghd are more likely to be poor, compared to their counterparts in RRS, supporting the earlier findings of higher poverty levels in Khatlon and Soghd. The marginal effects of the location dummies clearly show that households in Khatlon are twice more likely to be poor than their counterparts in Soghd. Table 7: Results for the Determinants of Poverty of Households in Rural Areas, 2003 Khatlon Soghd GBAO RRS Coefficient Coefficient Coefficient Coefficient (z-value) (z-value) (z-value) (z-value) 0.1738 0.1153 0.2127 0.0750 Number of household members (7.04) (4.97) (4.50) (4.13) 0.0119 0.6140 0.4124 -0.0042 Share of female hh-member (0.03) (1.82) (0.81) (-0.01) 1.4125 1.3148 0.8066 1.3313 Share of hh-member aged up to 5 (3.43) (3.05) (1.33) (2.86) 1.1582 0.9755 0.2186 0.9684 Share of hh-member aged from 6 to 15 (3.36) (3.15) (0.42) (2.45) 0.5631 0.5631 1.1636 0.7541 Share of hh-member aged over 64 (0.79) (0.83) (1.31) (0.93) -0.0017 -0.3626 0.1612 0.3848 Share of unemployed hh-members (-0.01) (-1.35) (0.35) (1.23) Average years of schooling per hh- -0.1207 -0.0737 -0.0622 -0.1312 member (-3.43) (-2.71) (-1.16) (-4.35) Household receives transfers from family -0.0828 -0.0028 -0.1722 -0.0698 members or other persons (Dummy) (-0.74) (-0.03) (-1.08) (-0.49) Household receives transfers from 0.2672 0.0781 0.0378 0.0626 organizations (Dummy) (1.50) (0.55) (0.21) (0.38) Household receives pension payments -0.2245 0.0062 -0.2332 0.1959 (dummy) (-1.52) (0.04) (-1.15) (1.24) 0.5153 -0.1634 0.7984 0.6821 Household receives allowances (dummy) (1.16) (-0.58) (1.39) (2.51) Household operates nonfarm business 0.3181 -0.6490 ---a -1.0344 (dummy) (0.50) (-0.58) (-1.68) -0.0027 -0.0016 -0.0056 -0.0004 Land owned by hh (Sotka) (-3.06) (-1.86) (-1.05) (-0.39) -0.0044 -0.0090 -0.0191 -0.0018 Livestock owned by hh (LSU) (-3.42) (-2.75) (-2.99) (-0.87) 0.0027 0.0004 0.0005 0.0004 Distance to capital (km) (2.91) (1.21) (1.33) (0.56) 0.0149 -0.0541 -0.3972 -0.4269 Constant (0.03) (-0.14) (-0.52) (-1.11) Likelihood Ratio 128.51 96.50 43.16 77.24 Pseudo R2 0.1552 0.1081 0.1163 0.1150 Number of Observations a only one household in GBAO operates a non-farm business Source: TLSS (2003), own calculations. The results for the individual oblasts presented in Table 7 also reveals that land ownership has a statistically significant impact on poverty in Khatlon and Soghd, but not in GBAO and RRS. This is probably because the Khatlon and Soghd oblasts are cotton growing areas, where access to land is quite important for livelihood strategies, in the absence of limited employment opportunities in the nonfarm sector. However, livestock ownership appears to be important in reducing poverty in the Khatlon, Soghd and GBAO oblasts, but not in RRS. Similar to the findings for the national level, education appears to be statistically significant in reducing poverty in all oblasts, with the notable 15 exception of GBAO. Household size also appears to be positive and significantly associated with poverty in all the oblasts, supporting the earlier finding for the national level that the larger the household size, the higher the probability of the household falling below the poverty level. 3.3 Impacts of Institutional Changes on Poverty Several important institutional changes have taken place between 1999 and 2004 that would be expected to have effects on household incomes and poverty. Some of these changes represented a continuation of the move towards market liberalization, while others involved land reforms and reforms in the cotton sector. Land reform was initialized in 1992 with the objective of privatizing the existing sovkhoz and kolkhoz farms by dividing them into smaller and collective units and then conferring inheritable rights to the owners. However, the civil war retarded progress until 1999 when efforts were intensified to complete the process by December 2005. Under legislation passed in 1996, individuals or groups within privatized, collective dehqon farms have the right to claim use rights to their share of the collective farm, and to create a separate farming entity on this basis. It has been argued that the privatization process has been limited to wealthier farmers who are able to take advantage of the non-transparency of the process and to influence the decisions of the District Land Committee. Although a number of NGOs have been working to enhance the access of all farmers to the privatization process, a large majority of farmers have not yet been able to benefit from it. In this section we use the information collected on land privatization to explore the association between the land reform exercise and poverty. Specifically, we employ the number of land share certificates and land use titles issued between 1999 and 2003, as well as the reduction of the land holdings in state or collective farms as additional factors influencing the probability of a household being poor. While land share certificates are the paper shares that every farmer receives in the restructuring process, land use titles capture the impact of effective farm restructuring, where farmers are issued with clear land use rights. Other variables included reflect the access to information, regional labor markets as well as distances to central markets. The analysis makes use of the data in the 2003 TLSS. This data set contains 208 observations of point of sampling units (PSU) that covers the entire nation. Since we focus on the impact of land reforms and changes in the cotton production, only 126 PSU located in rural areas are used for the following analysis. The location and household characteristics of the households in this PSU are provided in the data set. The descriptive statistics of the variables considered in the analysis are given in Table 8. The dependent variable employed in the analysis is the share of households of a PSU living below the poverty line of PPP US$ 2.15 per day. The analysis was done with ordinary least squares method. Table 8: Descriptive Statistics – District Level Poverty Mean Std. Dev. Min Max Share of households living below 47.06 Somoni/month 0.64 0.21 0.10 1.00 Number of land use titles issued 447.71 746.70 5.00 3319.00 Number of land share certificates issued 10024.88 10226.73 21.00 48249.00 Change in land in state or collective farms -0.39 0.41 -0.99 2.98 Distance to oblast administration center (km) 80.87 87.07 0.00 350.00 Distance to raion administration center (km) 17.68 16.53 0.00 85.00 Distance to capital (km) 269.39 237.96 0.00 1019.00 Average wage of cotton-farm worker (Somoni/Month) (cotton-producing regions only, N=91) 25.75 8.86 6.47 44.92 Change in cotton yield 1999-2003 (0-1) (cotton-producing regions only, N=91) 0.88 0.96 -0.23 3.45 Source: TLSS (2003), own calculations. 16 Given that cotton is the most important agricultural commodity in Tajikistan and changes in the cotton sector would be expected to affect incomes of rural households and consequently rural poverty. On average, cotton yields increased by almost 46 percent in the country, although there were variations between districts. To capture the impact of productivity increases in the cotton sector on rural poverty, the analysis was conducted exceptionally for cotton producing areas. Changes in cotton productivity between 1999 and 2003 and average wages of workers on cotton farms were added to the variables that were used in Table 9 below. Table 9: Determinants of District Level Poverty, Impact of Land Reforms, 2003 Variables Coefficient t-value Elasticities Number of land use titles issued -0.0001 (-1.99) -0.0367 Number of land share certificates issued 0.0000 (1.85) 0.0613 Change in land in state or collective farms -0.0189 (-0.43) 0.0116 Distance to oblast administration center (km) 0.0005 (1.88) 0.0634 Distance to raion administration center (km) 0.0021 (1.82) 0.0576 Distance to capital (km) 0.0000 (0.45) 0.0172 Constant 0.7561 (4.51) Observations =126; R2 = 0.27 Source: TLSS (2003), own calculations. The estimated coefficients and elasticities computed at the means are presented in Table 9. These results clearly show that while the number of land use titles issued in a district has a negative and statistically significant impact on the probability of a household being poor, the higher the number of land share certificates issued in a district, the higher the likelihood of households being poor in the region. The magnitudes of the elasticities which are quite small, indicate that an increase of the number of land use titles by one percent results in a decrease in the share of poor households by 0.04 percent, while an increase in the number of land share certificates leads to an increase in the share of poor households in the district by 0.06 percent These findings suggest that restructuring exercise that clearly confers property rights on farmers or households, encourages such households to invest in productivity enhancing inputs and production methods to increase output and farm incomes. This measure is more effective in combating poverty than reforms that merely issue land share certificates to farm workers without any indication of property rights. Changes in land in state or collective farms appear not to have any significant influence on the share of poor households in the district being poor, suggesting that the current practice of restructuring the state and collective farms into larger number of associations of dehqon farms are less effective in alleviating poverty. The results presented in Table 10 again indicate that land use titles reduce poverty, while land share certificates exert no significant influence on poverty in a region, again indicating that land restructuring can only be effective if clearly defined property rights are conferred on farmers. Noteworthy is the negative and significant impact of wages of cotton workers and cotton yields on poverty in an area, indicating that productivity increases in cotton and higher wages paid to workers on cotton farms significantly lower poverty rates in these districts. The elasticity for this variables show that a 10 percent increase in cotton wages in a district leads to a more than 3 percent decline in the level of poverty in such a cotton producing district. The impact of cotton productivity increase is much lower, averaging less than 1 percent decline in the poverty rate. The results also revealed that the higher the growth in cotton yield between the period 1999 and 2003, the lower the level of poverty, indicating that poor households also benefited from 17 productivity growth in the cotton sector. As Table 11 indicates, regions with a high increase in cotton productivity could not take profit from this increase as poverty is higher in these regions as in regions with lower productivity change. Lowest poverty can be found in regions with a moderate increase in cotton yields. Table 10: Estimation Results – Regional Poverty, Impact of Cotton Production and Land Reforms, Cotton Producing Regions Variables Coefficient (t-value) Elasticity Average wage of worker on cotton farm (Somoni/month) -0.0091 (-2.74) -0.3596 Change in cotton yield -0.0605 (-1.83) -0.0809 Number of land use titles issued -0.0001 (-2.19) -0.0521 Number of land share certificates issued 0.0000 (-0.45) -0.0241 Change in land in state or collective farms -0.0895 (-1.73) 0.0491 Distance to oblast administration center (km) 0.0013 (2.66) 0.1189 Distance to raion administration center (km) 0.0002 (0.12) 0.0042 Distance to capital (km) 0.0001 (0.78) 0.0310 Constant 0.8119 (4.69) Observations = 91; R2 = 0.46 Source: TLSS (2003), own calculations. As in the other estimations, the distance to oblast administration center is positive and statistically significant, suggesting that household located further from the administrative center are more likely to be poor, compared to those located closer to the center. This finding supports the notion that households located further away from the center face higher transaction costs in searching for information on prices and employment opportunities in the nonfarm sector. Normally, declines in the cost of information and transport flows due to good infrastructure reduce transaction costs and improve the efficiency with which rural labor and financial markets channel inputs into activities yielding highest returns. Table 11: Headcount Poverty Rates by Cotton Yield Change Headcount Poverty in Change in Change in cotton yield 2003 Headcount Poverty 1999-2003 Bottom 25% 64.32 -9.98 Average 61.63 -12.10 Top 25% 74.35 -12.15 Source: TLSS (1999, 2003), own calculations. 18 I.V Household Poverty Levels and Nutrition Although poverty levels and the depth of poverty declined in all regions of Tajikistan between 1999 and 2003, recent evidence appears to suggest that malnutrition has not improved over the last couple of years. Malnutrition has many dimensions to be addressed, such as food security, socio-cultural behavior, dietary behavior, consumer preferences, and access to safe drinking water and health facilities. On the other hand, chronic malnutrition rates are associated with micro-nutrient deficiency, in particular, iron deficiency, iodine deficiency and vitamin A deficiency. According to the FAO, food security exists when all people, at all times, have access to sufficient, safe and nutritious food to meet their dietary needs and food preferences for an active and healthy life. The broader definition of food security also includes non-food inputs such as clean water, sanitation and health care. Recent studies on food security in the country indicate that there are pockets of the population that are food insecure. However, the definition of food security employed in most of the studies involves the broad definition of diet quality. For example, the WFP (2005) study classifies households as chronically food insecure based on their diet quality. According to the findings, 10 percent of sampled households have very poor dietary consumption, relying on daily consumption of bread/wheat as their staple food. The dietary diversity of these households is quite low and they do not have a balanced diet in terms of nutritional quantity and quality. In addition, 17 percent of households were classified as very vulnerable to food insecurity because they do not consume sufficient quantities of food rich in protein, and also access their food mainly through purchase. In the sections that follow, the dietary behavior and consumption patterns of the households are examined, followed by their access to safe drinking water and other amenities. In addition to the food expenditure pattern of households, the calorie and protein availability for the individual households is also presented. The estimates are done for different income groups and for rural and urban areas. In addition, the calorie and protein consumption of households differentiated according to geographic location is also presented. Table 12 provides the total household expenditure and total food expenditure profiles, as well as the budget allocation pattern of the households, derived from the 2003 TLSS. They are calculated from the budget shares of each of the 4160 households in the survey, averaged over the whole sample in column 1 and then over the bottom and top quintiles of per capita expenditure in columns 2 and 3, respectively. The average of the per capita household expenditure is 53.97 Somoni per person per month; the corresponding means for the bottom and top quintiles are 19.63 and 116.84 Somoni, respectively. The average household food expenditure reveals a similar pattern, with a sample average of 35.56 Somoni per person per month, and 14.06 and 73.15 Somoni for the bottom and top quintiles, respectively. The food expenditure profile reveals that poorer households allocate a substantial proportion of their food expenditure to cereals and cereal products. Almost 45 percent of their food outlay is allocated to this food group. Although animal source foods, such as meat, fish and eggs have been identified as a compact and efficient source of micronutrients, the poorer households appear to be spending a smaller proportion of their food outlay on this group of food. While the top quintile allocates almost 20 percent of food expenditure to meat, fish and eggs, the bottom quintile allocates only about 5 percent to this group of food. Calorie and protein availability are computed from the basic data using the USDA Nutrient Database for Standard Reference, as well as the conversion Table obtained from the WFP office in Dushanbe. The full detail of reported food consumption is employed, with weights converted to calories and protein using the calorie and protein content factors. In computing the calorie and protein consumption, it is assumed that total of food consumed g is made up of individual groups of food j. These individual groups of food are taken to be homogenous, so that a kilogram of food j has a constant calorie and protein available kj no matter who buys it. Total calories or protein Y is then given by Y = ∑∑ g ( q gj k gj ) , where qgj is the total quantity of food. The computed results for calories and j∈g proteins are presented in Table 13.below. 19 Table 12: Household Expenditure Patterns by Quintiles, 2003 Mean Bottom Quintile Top Quintile Per capita total 53.97 19.63 116.84 expenditures (Somoni/month) Total household food 35.56 14.06 73.15 expenditures (Somoni/month) Total household food 65.89% 71.63% 62.61% expenditures (share of total expenditure) Share of food expenditures Cereals and Cereal 37.64% 44.42% 30.09% Products Fruits and Vegetables 15.02% 13.82% 15.39% Meat, Fish and Eggs 13.80% 4.60% 19.53% Milk and Milk Products 3.69% 2.82% 3.39% Edible Oils 10.70% 10.79% 10.34% Source: TLSS (2003), own calculations. Columns 2-4 show the distribution of calories and protein over the various food groups. The second row shows that per capita daily calories are 2,676 on average and 1,549 and 4,340 in the two extreme quintiles, respectively. The average of 1,549 for the bottom quintile is significantly below the average for developing countries as a whole, indicating that much needs to be done to improve the calorie intake of poor households. The third row indicates that per capita daily protein consumption is 80.99 on average and 45.72 and 134.35 in the two extreme quintiles, respectively. The differences between the bottom and top income quintiles are quite significant, and show that poorer households have difficulty in meeting both their calorie and protein requirements. It, however, needs to be noted that these wide variations in calorie availability across income quintiles are quite common in developing countries. For example, a recent study by Abdulai and Aubert (2004) in Tanzania revealed the per capita daily calories for the lowest and highest income deciles were 1,414 and 2,270, respectively, with mean consumption of 2,270 calories. An earlier study by Subramanian and Deaton (1996) for India showed per capita daily calories for the lowest and highest income deciles of 1,429 and 3,167, respectively, and mean consumption of 2,098 calories. The next panel of Table 13 (rows 5-9) shows the contribution of the individual food groups to calorie consumption of households. It is obvious from the figures that cereals and cereal products group is the largest source of calories for Tajik households, with about 67% on average. This food group is particularly important in covering energy needs of poor households, since its calorie share amounts to 75.36% of total calorie availability of the bottom quintile of the households in the present analysis. However, as is evident in the Table, the significance of this group as a source of calories declines with increasing income. The group contributes about 60 percent of the energy needs of the top income quintile. The meat, fish and eggs group is less important in providing calories to households. Compared to cereals and cereal products group, which is a cheap source of calories to all households, the meat, fish and eggs group is an expensive source of calories. It provides 6.57% on average, 9.93% for high-income households and only 2.57% for the poorest income group. Rows 11-15 present the contribution of the individual food groups to protein consumption of the sampled households. On average, households derive about 73% of their protein supply from cereals and cereal products. Quite interesting is the observation that this group contributes about 85.17% of the protein needs of the poorest households, and about 64.08% of the wealthiest households. Households in the top income quintile derive on average 25.45% of their protein needs from meat, fish 20 and eggs, while this group supplies only 7.16% of the protein requirements of the bottom income quintile. Table 13: Mean Household Consumption Patterns by Quintiles, 2003 Top Mean Bottom Quintile Quintile Calories, total (kcal) 2676 1549 4340 Protein, total (g) 80.99 45.72 134.35 Share [percent of total calorie consumption] Cereals and cereal products 66.89% 75.36% 60.08% Vegetables and Fruits 6.91% 6.38% 7.31% Meat, Fish and Eggs 6.57% 2.57% 9.93% Milk 1.98% 1.37% 2.06% Edible Oils 12.09% 11.06% 13.22% Share [percent of total protein consumption] Cereals and cereal products 73.14% 85.17% 64.08% Vegetables and Fruits 5.56% 5.12% 5.80% Meat, Fish and Eggs 17.28% 7.16% 25.45% Milk 3.37% 2.40% 3.43% Edible Oils 0.00% 0.00% 0.00% Source: TLSS (2003), own calculations. Table 14 presents the budget allocation profile and sources of calories and protein of the households across locations. The Table shows that on average urban households have more calories and protein available to them compared to their rural counterparts. While rural households consume 2,560 calories per day, the corresponding figure for their urban counterparts is 2,814 calories. Similarly, urban households have a higher supply of protein than rural households. Also in the Table are calorie and protein supply for valley, mountain, peri-urban and urban centers. While the difference between the valley and mountain areas is not significant, the calorie and protein availability for peri-urban areas appear to be much higher than those of valley and mountain areas. On average, households located in peri-urban areas consume about 2,938 calories per day, while those in the valley and mountain areas consume 2,462 and 2,512 calories per day, respectively. Also evident from the Table is the observation that households located in rural areas derive a much greater proportion of their protein needs from cereals and cereal products, as compared to their counterparts in the urban areas. On the other hand, relative to rural households, the urban households derive a higher proportion of their protein needs from meat fish and eggs. These differences are also reflected in the differences between valley and mountain as a group and the city centers. Although the figures presented suggest that households in all locations consume sufficient quantities of calories and proteins, it needs to be noted that these are averages, with wide variations around the mean. As shown in the Table with income quintiles, poorer households generally have lower than the minimum daily calorie requirement, suggesting the possible need for intervention, as is practiced by the World Food Program in certain areas. In particular, location based targeting in rural areas could be 21 employed to increase the calorie availability of rural households in general, and specifically those households located in mountain and valley areas. Table 14: Mean Household Consumption Patterns by Areas, 2003 Urban Rural Valley Mountain Peri-Urban Calories, total (kcal) 2,814 2,560 2,462 2,512 2,938 Protein, total (g) 86.86 77.06 73.72 76.63 87.74 Share (percent of total calorie consumption) Cereals and cereal products 63.29% 69.74% 68.77% 70.90% 70.37% Vegetables and Fruits 7.72% 6.49% 7.46% 4.70% 6.81% Meat, Fish and Eggs 8.99% 5.12% 5.00% 5.37% 5.04% Milk 1.34% 2.41% 2.29% 3.09% 1.67% Edible Oils 12.15% 11.70% 11.89% 11.65% 11.30% Share (percent of total protein consumption) Cereals and cereal products 67.79% 76.71% 76.24% 76.65% 77.98% Vegetables and Fruits 6.17% 5.20% 5.96% 3.85% 5.40% Meat, Fish and Eggs 23.19% 13.56% 13.44% 13.96% 13.20% Milk 2.24% 4.12% 3.94% 5.21% 2.88% Edible Oils 0.00% 0.00% 0.00% 0.00% 0.00% Source: TLSS (2003), own calculations. 22 V. Migration of Household Members As indicated earlier, in many parts of the developing world, migration has often been used as a copying strategy in the absence of access to sufficient income from agriculture and the rural nonfarm sector. In particular, migration of younger men has been an essential step in acquiring capital to establish a livelihood back home, either in the farm or nonfarm sector. In rural communities, those without access to migration or local nonfarm income are likely to be at a severe disadvantage in terms of investing in income generating activities. Recent evidence suggests that Tajikistan’s robust economic growth during the past several years has been driven to a significant extent by strong inflows of migrant workers’ remittances, estimated to be close to 20 percent of GDP in 20041. According to these estimates, remittances have become the most important source of external finance during the past few years. These inflows have contributed to increased income and fuelled private consumption and thereby helped reduce poverty. As such, emigration, particularly seasonal migration to Russia has become an alternative poverty coping measure chosen by the poor. The PRSP indicate that several intergovernmental agreements, most importantly the agreement with Russia, have helped increase the safety and improve the general condition of the migrants. As demonstrated by Figure 5, migration appears to provide an important safety net for Tajik households. The figure indicates that migration increased dramatically over the period between 1998 and 2003. Figure 5: Number of Persons Living Abroad for Three Month or More at One Time 900 urban 800 rural 700 600 500 400 300 200 100 0 1998 1999 2000 2001 2002 2003 Source: TLSS (2003), own calculations. The most important destination of migrants is Russia, where 95 percent of the migrants move to. The other destinations are mostly other states of the former Soviet Union. Figure 6 clearly reveals that overwhelming majority of the persons going abroad stated that they were migrating for economic reasons, with more than 90 percent of migrants indicating that they were leaving to search for a better paid job. The rest of the migrants indicated their motives for moving from their original places to be; starting a new business somewhere, studying in a different place or joining family members. With regard to differences in the rates of migration, the data revealed that most migrants originated from RRS, followed by Soghd. GBAO produced the least number of migrants (see Figure 7). 1 IMF estimate based on official data. 23 Figure 6: Reasons for Going Abroad TO START A NEW JOB/BUSINESS TO LOOK FOR A BETTER PAID JOB STUDY OTHER Source: TLSS (2003), own calculations. Figure 7: Origin of persons living abroad for three month or more in 2003 Kahtlon Sugd GBAO RRS Dushanbe Source: TLSS (2003), own calculations. 5.1 Are rural households benefiting from remittances? It is also important to ask if rural households are benefiting from the remittances that are being sent from migrants. A detailed analysis of the data which is presented in Table 15 reveals that 31 percent of households in rural areas received remittances from relatives in 2003, compared to 23 percent of their counterparts in urban areas who reported receiving remittances. This suggests that migration and remittances provide an important safety net for rural households, given that almost a third of household receive such private transfers. Private transfers are defined as assistance (money or goods) from members no longer living in the household. Institutional transfers stem from any kind of organization or other sources. Quite interesting from the results in Table 15 is the observation that private transfers, which are mainly remittances, are highest for the top income quintile. More than 30 percent of this category of households received transfers, compared to the other categories, where 23-25 percent of households reported receiving private transfers. Wealthier households have better financial resources to migrate to other destinations, particularly foreign countries. They are also generally better educated and as such are able to secure employment more easily than lower income groups. It is therefore not surprising that most migrants were from RRS, which is the wealthiest oblast. This is also reflected in Figure 8, which shows the share of people migrating from households in the various income quintiles as well as for urban and rural areas. It is evident from the figure that households in the top income quintile had about 24 4.5 percent of household members migrating, while the corresponding figure for households in the bottom income quintile was 3.8 percent. Table 15: Share of households receiving transfers or other payments by income quintiles and location, 2003 Income Private transfers Institutional Pensions Allowances Quintile transfers 1 0.25 0.22 0.41 0.04 2 0.23 0.27 0.40 0.05 3 0.25 0.27 0.41 0.04 4 0.25 0.24 0.39 0.03 5 0.31 0.17 0.36 0.02 Mean 0.26 0.23 0.39 0.03 Urban areas 0.23 0.25 0.41 0.04 Rural areas 0.31 0.20 0.37 0.03 Source: TLSS (2003), own calculations. Figure 8: Share of Persons Migrating in 2003 by Income Quintiles and Region 4.60 4.47 4.49 4.40 4.18 4.20 4.07 4.05 4.02 4.00 3.83 3.80 3.60 3.40 1 2 3 4 5 urban rural Source: TLSS (2003), own calculations. 25