Examining the file media/Synopses/download?doi=10.1.1.518.953&rep=rep1&type.tagsearch.results.html:
This file was generated: 2020-01-19 18:44:23
Overall Tag Match Counts:
Searching for tag age:
(return to top)
...
p.000001: 4 Empirical results
p.000001:
p.000001: Appendix Tables 1-3 present the estimation results for three alternative models exploring the
p.000001: predictors of moving out of and into poverty. Model 1 looks at poverty mobility predictors where community-level
p.000001: characteristics are not included among the regressors (Appendix Table 1). It includes household head characteristics
p.000001: such as age, gender, education, and self-reported health status; household demographics such as the share of adults;
p.000001: employment status of household members; and regional (oblast) dummies. Model 1 is estimated using two
p.000001: specifications, with and without district (rayon) characteristics. For ease of comparison of the coefficients, the
p.000001: estimation results of the moving-out of poverty model are presented next to the estimation results of the becoming-poor
p.000001: model.
p.000001:
...
...
p.000022: 4.3 The main predictors of moving into poverty
p.000022:
p.000022: The sample of all rural households
p.000022: The factors that explain a household’s likelihood to fall into poverty are different from those that explain moving out
p.000022: of poverty (Appendix Tables 1 and 2). The probability of moving into poverty declines significantly once the share of
p.000022: working-age individuals in a household increases. The estimates suggest that the probability of falling into poverty
p.000022: declines from 55 to 10 per cent when the share of adults increases from 25 to 100 per cent (Panel A, Figure 10).
p.000022: Employment in public administration reduces the risk of impoverishment: the probability of falling into poverty
p.000022: drops from 20 to 3 per cent as the share of adults in administrative employment increases from 25 to 100
p.000022: per cent (Panel B, Figure 10). Household size is also a factor: the estimates suggest a U-shaped relationship between
p.000022: the probability of becoming poor and household size (Appendix Table 2). In other words, small households
...
...
p.000023: important implications, which are briefly discussed below.
p.000023:
p.000023: First, several household-level factors emerge as key predictors of poverty transition, suggesting the
p.000023: importance of continued investments to improve human capital outcomes. The odds for exiting poverty
p.000023: increase with the higher level of education and improved health status of the household head, as well as with the
p.000023: higher ratio of adults in wage employment. The risk of falling into poverty declines with a higher share of working-age
p.000023: people in the household and a larger share of adults working in public administration.
p.000023:
p.000023: The district-level data suggest that areas where cotton farming has a more prominent role are likely to
p.000023: have higher levels of poverty. The analysis of poverty mobility at the household level also indicates that households
p.000023: located in cotton-producing areas do not enjoy better odds of climbing out of poverty. The analysis actually reveals
...
Searching for tag capacity:
(return to top)
...
p.000001:
p.000001: The World Institute for Development Economics Research (WIDER) was established by the United Nations University
p.000001: (UNU) as its first research and training centre and started work in Helsinki, Finland in 1985. The Institute
p.000001: undertakes applied research and policy analysis on structural changes affecting the developing and
p.000001: transitional economies, provides a forum for the advocacy of policies leading to robust, equitable and
p.000001: environmentally sustainable growth, and promotes capacity strengthening and training in the field of economic and
p.000001: social policy making. Work is carried out by staff researchers and visiting scholars in Helsinki
p.000001: and through networks of collaborating scholars and institutions around the world.
p.000001:
p.000001: www.wider.unu.edu publications@wider.unu.edu
p.000001:
...
Searching for tag children:
(return to top)
...
p.000022: declines from 55 to 10 per cent when the share of adults increases from 25 to 100 per cent (Panel A, Figure 10).
p.000022: Employment in public administration reduces the risk of impoverishment: the probability of falling into poverty
p.000022: drops from 20 to 3 per cent as the share of adults in administrative employment increases from 25 to 100
p.000022: per cent (Panel B, Figure 10). Household size is also a factor: the estimates suggest a U-shaped relationship between
p.000022: the probability of becoming poor and household size (Appendix Table 2). In other words, small households
p.000022: (elderly people living alone) and very large households (usually households with many children) face a higher risk of
p.000022: falling into poverty than the average-sized household.
p.000022:
p.000022: Examining the impact of district/community level characteristics, we find that variations in rain fall are
p.000022: associated with the risk of becoming poor. Households located in communities with less than average amount of
p.000022: rain over the previous year face a 55 per cent chance of becoming poor versus 28 per cent for households in
...
Searching for tag dependence:
(return to top)
...
p.000011: Aij = fi1 1j + fi2 2j + ... + fiq qj + ij
p.000011: (1)
p.000011:
p.000011: In (1), Aij is known since it is one of the values describing whether or not household i has asset j. The term fik
p.000011: represents the observation for household i of the value of factor k which needs to be estimated. The term kj is the
p.000011: coefficient indicating the dependence of the observed asset variable j upon the factor k, this coefficient being also
p.000011: estimated. The residual, ij, is the error term. In other words, factor analysis produces an index
p.000011: representing (through the vector of common factors F) the data generating process underlying the actual
p.000011: observations Aij. This is done by finding the one dimension of the space in which the original observations are
p.000011: represented with the largest variance, from j = 1, ..., p to k = 1, ..., n with n p.000011:
...
Searching for tag economic:
(return to top)
...
p.000001: (Swedish International Development Cooperation Agency— Sida).
p.000001: ISSN 1810-2611 ISBN 978-92-9230-072-2
p.000001:
p.000001: Third, there is strong evidence of geographic poverty mobility traps in Tajikistan. Higher levels of
p.000001: poverty in a district appear to reduce significantly the chance of a household shedding poverty. Living in a
p.000001: region with overall slow economic growth is also found to undermine the odds of exiting poverty and to increase the
p.000001: risk of falling into poverty. Finally, several key household-level factors, such as the share of adults, education
p.000001: level, health status and participation in wage employment, also emerge as significant predictors of poverty
p.000001: mobility.
p.000001:
p.000001:
...
...
p.000001:
p.000001: The World Institute for Development Economics Research (WIDER) was established by the United Nations University
p.000001: (UNU) as its first research and training centre and started work in Helsinki, Finland in 1985. The Institute
p.000001: undertakes applied research and policy analysis on structural changes affecting the developing and
p.000001: transitional economies, provides a forum for the advocacy of policies leading to robust, equitable and
p.000001: environmentally sustainable growth, and promotes capacity strengthening and training in the field of economic and
p.000001: social policy making. Work is carried out by staff researchers and visiting scholars in Helsinki
p.000001: and through networks of collaborating scholars and institutions around the world.
p.000001:
p.000001: www.wider.unu.edu publications@wider.unu.edu
p.000001:
...
...
p.000001: households. We use a panel of rural households that have been observed during two time periods: June-July
p.000001: 2003 and July-November 2004.
p.000001:
p.000001: Analysing the dynamics of rural poverty in Tajikistan during this time period is particularly interesting
p.000001: in view of the drastic changes that have occurred in the country over the last several years. Emerging in 1999 from
p.000001: civil war and a prolonged period of economic collapse,1 the country’s economic performance has been impressive from the
p.000001: year 2000, with sustained real GDP annual growth rates of 7 to 9 per cent.2
p.000001:
p.000001: Economic growth has been accompanied with substantial reduction in poverty, dropping from 81 per cent of the population
p.000001: living below the poverty line (US$2.15 per day) in 1999 to 64 per cent in 2003 (World Bank 2006). Although
p.000001: poverty headcount fell during this period by 19 percentage points in the rural areas compared to 14 percentage points
p.000001: in urban centres, it remains higher in the rural regions: 65 per cent versus 59 per cent. As 73 per cent of the
p.000001: population live in the countryside, poverty in Tajikistan continues to be an overwhelmingly rural
p.000001: phenomenon. Economic growth and the resultant poverty reduction are explained by three major factors: (i) conflict
p.000001: cessation, which allowed economic activity to resume and markets to develop; (ii) initial impact of the macroeconomic
p.000001: stability and agricultural reforms in the non-cotton sector that enabled farmers to diversify production
p.000001: and increase productivity; and (iii) large increase in migrant workers exiting Tajikistan for Russia and other
p.000001: countries. However, there have been concerns that once the initial benefits of these ‘special’ factors dry out,
p.000001: Tajikistan’s poverty reduction trends may not be sustainable (World Bank 2006).
p.000001:
p.000001: In view of the sound economic growth rates, markedly reduced but still very high rural poverty, and concerns over the
p.000001: sustainability of the country’s poverty reduction trend, it is important from a policy perspective to understand
p.000001: the key factors at the micro (household/community) level that explain the transition of rural households in and
p.000001: out of poverty. This is the main objective of this paper.
p.000001:
p.000001: The paper contributes to the literature on welfare dynamics in general, and to the studies of poverty in Tajikistan in
...
...
p.000001:
p.000001: accounting for 19 per cent of sector output, expanded by 61 per cent.6 The agricultural production developments of
p.000001: this period can be well illustrated with data on cotton production. Cotton traditionally has been a major
p.000001: agriculture commodity, and continues to account for about two-thirds of total crop output value (Ukaeva
p.000001: 2005). The cotton sector has experienced substantial output fluctuations during periods of civil conflict
p.000001: and economic transition. Between 1991-99, cotton output declined 62 per cent, from 820,000 tons to 313,000
p.000001: tons, but increased 73 per cent between 1999 and 2003, still accounting for only 65 per cent of the 1991 level. The
p.000001: increase in output is mostly a reflection of improved yields (1.1 tons per hectare in 1999 to 1.8 tons per
p.000001: hectare in 2003, Figure 3) as well as an increase in cultivation area.
p.000001: The cotton sector in Tajikistan has been severely hit by declining global prices. Despite output increasing by more
p.000001: than two-thirds between 1999 and 2003, the declining global prices reduced the real value of output by 7 per cent
...
...
p.000008: wages, and they are less likely to be paid on time. We find a very weak correlation between the level of the
p.000008: cotton-farm debt (per capita) and poverty at the district level.
p.000008:
p.000008: The extent of decline in output during the 1990s is strongly (positively) associated with poverty. Based on cotton
p.000008: output data, we find that districts with greater output gaps between 1991 (the peak output year before
p.000008: economic collapse) and 2003 are much more likely to be poorer. The deviations in cotton output and its value between
p.000008: 1999-2003 are not correlated with the levels of poverty at the district level.
p.000008:
p.000008: The above examination of the key correlates of poverty at the district level provides a solid basis for analysing the
p.000008: (asset-based) poverty mobility at the household level in the next section.
p.000008:
...
...
p.000016: characteristics. The level of poverty in a district is a significant predictor of poverty mobility at the
p.000016: household level. The estimates indicate that, after controlling for other characteristics, for a household located in a
p.000016: district with 30 per cent poverty headcount (based on the US$2.15 per day poverty line), there is a 70 per cent
p.000016: probability of escaping poverty, while the probability is a mere 5 per cent for a household located in a district
p.000016: with a poverty headcount of 90 per cent (Panel A, Figure 8).19 Thus living in a region with
p.000016: weak economic growth performance
p.000016:
p.000016:
p.000016:
p.000016: 18 For calculating the predicted probabilities of moving out of/into poverty for the overall rural sample, we use the
p.000016: regression results given in Appendix Table 2 (with interaction terms); for the sample of rural households living in
...
...
p.000024: increase in this type of farming has not yet improved the chances of mobility out of poverty. This is
p.000024: likely a reflection of the fact that land ownership transfers are often on paper only, and thus are not accompanied by
p.000024: improvements in farm productivity.
p.000024:
p.000024: There is strong evidence of geographic poverty mobility traps. A higher level of poverty in a district significantly
p.000024: reduces the chances of a household of moving out of poverty. Living in a region with an overall slow economic
p.000024: growth rate is also found to undermine the odds of escaping poverty and increase the odds of falling into poverty.
p.000024: The risk of impoverishment significantly increases for households in regions that experienced drought. In
p.000024: other words, everything else being equal, the geographical location of a household matters considerably in
p.000024: terms of its chances of escaping or falling into poverty. It is worth noting that this observation
p.000024: regarding geographical poverty traps on the part of rural households in Tajikistan confirms numerous similar findings
...
Searching for tag education:
(return to top)
...
p.000001: ISSN 1810-2611 ISBN 978-92-9230-072-2
p.000001:
p.000001: Third, there is strong evidence of geographic poverty mobility traps in Tajikistan. Higher levels of
p.000001: poverty in a district appear to reduce significantly the chance of a household shedding poverty. Living in a
p.000001: region with overall slow economic growth is also found to undermine the odds of exiting poverty and to increase the
p.000001: risk of falling into poverty. Finally, several key household-level factors, such as the share of adults, education
p.000001: level, health status and participation in wage employment, also emerge as significant predictors of poverty
p.000001: mobility.
p.000001:
p.000001:
p.000001: Authors notes
...
...
p.000010: households representing 8,368 individuals; 589 of the households are rural.12 The 2004 HES used the same sample frame
p.000010: (list of clusters) as the 2003 TLLS. A comparison of the distribution of the basic variables from the panel
p.000010: sample against the 2003 cross- section indicates that the panel sample is fairly representative of the overall
p.000010: population, both at rural and urban levels.
p.000010:
p.000010: Both surveys collected information on such household attributes as demographics, education and health, income
p.000010: and expenditures, assets, and consumption. The analysis of poverty dynamics here uses the panel component of the two
p.000010: surveys, and is based on an asset index. Construction of the asset index is described below.
p.000010:
p.000010: In addition to utilizing household-level data, the empirical analysis at the micro (household) level
p.000010: exploits a few key district-level variables to capture agricultural reform and various policy changes
...
...
p.000001: 4 Empirical results
p.000001:
p.000001: Appendix Tables 1-3 present the estimation results for three alternative models exploring the
p.000001: predictors of moving out of and into poverty. Model 1 looks at poverty mobility predictors where community-level
p.000001: characteristics are not included among the regressors (Appendix Table 1). It includes household head characteristics
p.000001: such as age, gender, education, and self-reported health status; household demographics such as the share of adults;
p.000001: employment status of household members; and regional (oblast) dummies. Model 1 is estimated using two
p.000001: specifications, with and without district (rayon) characteristics. For ease of comparison of the coefficients, the
p.000001: estimation results of the moving-out of poverty model are presented next to the estimation results of the becoming-poor
p.000001: model.
p.000001:
...
...
p.000016: Table 2). These include the size of the population point, distance to the nearest market, whether cotton
p.000016: is produced in the area, and the reported amount of rain compared to the previous rain. Model 2 is also
p.000016: estimated based on two specifications: excluding the interaction between the ‘cotton’ variable (dummy variable
p.000016: indicating if cotton produced in the community) and various other factors (specification 1). Specification 2
p.000016: includes the interactions between the ‘cotton’ variable and such factors as distance to market, share of household
p.000016: adults working in agriculture, education status and gender of the household head. These interaction terms are designed
p.000016: to gain a better understanding of the importance of the ‘cotton’ variable in explaining the poverty transitions.
p.000016:
p.000016: Model 3 uses a richer set of district-level characteristics that apply only to cotton- producing districts
p.000016: in order to get a better understanding of poverty mobility in these areas (Appendix Table 3). It is worth noting
p.000016: that two-thirds of the households in the rural panel sample reside in the cotton-producing districts. Here, we
...
...
p.000021: household mobility out of poverty (Appendix Table 1). The same is true with respect to the impact of living in a
p.000021: cotton-producing community (Appendix Table 2).
p.000021:
p.000021: At the household level, household head’s schooling is related to a significantly higher probability of escaping
p.000021: poverty. The estimates suggest that the probability of shedding poverty increases from the 25 per cent that applies to
p.000021: the household head with less than secondary education to 50 per cent for those with university education
p.000021: (Panel B, Figure 8). Better health status also improves the odds of moving out of poverty: the
p.000021: probability of exiting poverty rises from about the 17 per cent observed for household heads with (self-reported)
p.000021: bad/very bad health to almost 40 per cent for those enjoying good/very good health (Panel C, Figure 8). Finally,
p.000021: a larger share of adults in wage employment has a positive impact on the poverty exit probability. This improves
p.000021: from 30 per cent to 50 per cent as the share of adults in hired employment goes up from 25 to 100 per cent (Panel D,
...
...
p.000023: household level, we also look at the correlates of poverty at the district (rayon) level. The findings have
p.000023: important implications, which are briefly discussed below.
p.000023:
p.000023: First, several household-level factors emerge as key predictors of poverty transition, suggesting the
p.000023: importance of continued investments to improve human capital outcomes. The odds for exiting poverty
p.000023: increase with the higher level of education and improved health status of the household head, as well as with the
p.000023: higher ratio of adults in wage employment. The risk of falling into poverty declines with a higher share of working-age
p.000023: people in the household and a larger share of adults working in public administration.
p.000023:
p.000023: The district-level data suggest that areas where cotton farming has a more prominent role are likely to
p.000023: have higher levels of poverty. The analysis of poverty mobility at the household level also indicates that households
...
Searching for tag elderly:
(return to top)
...
p.000022: declines from 55 to 10 per cent when the share of adults increases from 25 to 100 per cent (Panel A, Figure 10).
p.000022: Employment in public administration reduces the risk of impoverishment: the probability of falling into poverty
p.000022: drops from 20 to 3 per cent as the share of adults in administrative employment increases from 25 to 100
p.000022: per cent (Panel B, Figure 10). Household size is also a factor: the estimates suggest a U-shaped relationship between
p.000022: the probability of becoming poor and household size (Appendix Table 2). In other words, small households
p.000022: (elderly people living alone) and very large households (usually households with many children) face a higher risk of
p.000022: falling into poverty than the average-sized household.
p.000022:
p.000022: Examining the impact of district/community level characteristics, we find that variations in rain fall are
p.000022: associated with the risk of becoming poor. Households located in communities with less than average amount of
p.000022: rain over the previous year face a 55 per cent chance of becoming poor versus 28 per cent for households in
...
Searching for tag gender:
(return to top)
...
p.000001: 4 Empirical results
p.000001:
p.000001: Appendix Tables 1-3 present the estimation results for three alternative models exploring the
p.000001: predictors of moving out of and into poverty. Model 1 looks at poverty mobility predictors where community-level
p.000001: characteristics are not included among the regressors (Appendix Table 1). It includes household head characteristics
p.000001: such as age, gender, education, and self-reported health status; household demographics such as the share of adults;
p.000001: employment status of household members; and regional (oblast) dummies. Model 1 is estimated using two
p.000001: specifications, with and without district (rayon) characteristics. For ease of comparison of the coefficients, the
p.000001: estimation results of the moving-out of poverty model are presented next to the estimation results of the becoming-poor
p.000001: model.
p.000001:
...
...
p.000016: Table 2). These include the size of the population point, distance to the nearest market, whether cotton
p.000016: is produced in the area, and the reported amount of rain compared to the previous rain. Model 2 is also
p.000016: estimated based on two specifications: excluding the interaction between the ‘cotton’ variable (dummy variable
p.000016: indicating if cotton produced in the community) and various other factors (specification 1). Specification 2
p.000016: includes the interactions between the ‘cotton’ variable and such factors as distance to market, share of household
p.000016: adults working in agriculture, education status and gender of the household head. These interaction terms are designed
p.000016: to gain a better understanding of the importance of the ‘cotton’ variable in explaining the poverty transitions.
p.000016:
p.000016: Model 3 uses a richer set of district-level characteristics that apply only to cotton- producing districts
p.000016: in order to get a better understanding of poverty mobility in these areas (Appendix Table 3). It is worth noting
p.000016: that two-thirds of the households in the rural panel sample reside in the cotton-producing districts. Here, we
...
Searching for tag health:
(return to top)
...
p.000001:
p.000001: Third, there is strong evidence of geographic poverty mobility traps in Tajikistan. Higher levels of
p.000001: poverty in a district appear to reduce significantly the chance of a household shedding poverty. Living in a
p.000001: region with overall slow economic growth is also found to undermine the odds of exiting poverty and to increase the
p.000001: risk of falling into poverty. Finally, several key household-level factors, such as the share of adults, education
p.000001: level, health status and participation in wage employment, also emerge as significant predictors of poverty
p.000001: mobility.
p.000001:
p.000001:
p.000001: Authors notes
p.000001:
...
...
p.000008:
p.000008: 3.1 Using an asset-based poverty line to identify poverty transitions
p.000008:
p.000008: The literature on poverty dynamics has increasingly recognized the importance of adopting an asset-based
p.000008: approach to study changes in wellbeing, especially in response to a wide range of different (climatic,
p.000008: health, political and other) shocks.10 Differentiating between stochastic and structural poverty
p.000008: transitions implies the availability of information on assets and expected levels of wellbeing. To illustrate the
p.000008: importance of the assets-based approach in capturing welfare-status changes, we use the conceptual framework
p.000008: advocated by Carter and May (2001) and Carter and Barrett (2006). This framework is presented in Figure 6.
p.000008:
p.000008: In essence, in any timeperiod a household can be regarded as structurally poor if household consumption
...
...
p.000010: households representing 8,368 individuals; 589 of the households are rural.12 The 2004 HES used the same sample frame
p.000010: (list of clusters) as the 2003 TLLS. A comparison of the distribution of the basic variables from the panel
p.000010: sample against the 2003 cross- section indicates that the panel sample is fairly representative of the overall
p.000010: population, both at rural and urban levels.
p.000010:
p.000010: Both surveys collected information on such household attributes as demographics, education and health, income
p.000010: and expenditures, assets, and consumption. The analysis of poverty dynamics here uses the panel component of the two
p.000010: surveys, and is based on an asset index. Construction of the asset index is described below.
p.000010:
p.000010: In addition to utilizing household-level data, the empirical analysis at the micro (household) level
p.000010: exploits a few key district-level variables to capture agricultural reform and various policy changes
...
...
p.000001: 4 Empirical results
p.000001:
p.000001: Appendix Tables 1-3 present the estimation results for three alternative models exploring the
p.000001: predictors of moving out of and into poverty. Model 1 looks at poverty mobility predictors where community-level
p.000001: characteristics are not included among the regressors (Appendix Table 1). It includes household head characteristics
p.000001: such as age, gender, education, and self-reported health status; household demographics such as the share of adults;
p.000001: employment status of household members; and regional (oblast) dummies. Model 1 is estimated using two
p.000001: specifications, with and without district (rayon) characteristics. For ease of comparison of the coefficients, the
p.000001: estimation results of the moving-out of poverty model are presented next to the estimation results of the becoming-poor
p.000001: model.
p.000001:
...
...
p.000021: cotton-producing community (Appendix Table 2).
p.000021:
p.000021: At the household level, household head’s schooling is related to a significantly higher probability of escaping
p.000021: poverty. The estimates suggest that the probability of shedding poverty increases from the 25 per cent that applies to
p.000021: the household head with less than secondary education to 50 per cent for those with university education
p.000021: (Panel B, Figure 8). Better health status also improves the odds of moving out of poverty: the
p.000021: probability of exiting poverty rises from about the 17 per cent observed for household heads with (self-reported)
p.000021: bad/very bad health to almost 40 per cent for those enjoying good/very good health (Panel C, Figure 8). Finally,
p.000021: a larger share of adults in wage employment has a positive impact on the poverty exit probability. This improves
p.000021: from 30 per cent to 50 per cent as the share of adults in hired employment goes up from 25 to 100 per cent (Panel D,
p.000021: Figure 8).
p.000021:
p.000021: The sample of rural households located in cotton-producing communities
...
...
p.000023: household level, we also look at the correlates of poverty at the district (rayon) level. The findings have
p.000023: important implications, which are briefly discussed below.
p.000023:
p.000023: First, several household-level factors emerge as key predictors of poverty transition, suggesting the
p.000023: importance of continued investments to improve human capital outcomes. The odds for exiting poverty
p.000023: increase with the higher level of education and improved health status of the household head, as well as with the
p.000023: higher ratio of adults in wage employment. The risk of falling into poverty declines with a higher share of working-age
p.000023: people in the household and a larger share of adults working in public administration.
p.000023:
p.000023: The district-level data suggest that areas where cotton farming has a more prominent role are likely to
p.000023: have higher levels of poverty. The analysis of poverty mobility at the household level also indicates that households
...
Searching for tag ill:
(return to top)
...
p.000001: increase in output is mostly a reflection of improved yields (1.1 tons per hectare in 1999 to 1.8 tons per
p.000001: hectare in 2003, Figure 3) as well as an increase in cultivation area.
p.000001: The cotton sector in Tajikistan has been severely hit by declining global prices. Despite output increasing by more
p.000001: than two-thirds between 1999 and 2003, the declining global prices reduced the real value of output by 7 per cent
p.000001: during the same period (Figure 3).7 Adverse developments in international cotton prices, coupled with the
p.000001: farmers’ ill-
p.000001:
p.000001:
p.000001: 6 See Ukaeva (2005) for a detailed discussion based on the decomposition of agricultural growth between
p.000001: 1999 and 2003.
p.000001: 7 Due to civil conflict and low cotton production, Tajikistan missed the opportunity to benefit from the
...
Searching for tag income:
(return to top)
...
p.000010: households representing 8,368 individuals; 589 of the households are rural.12 The 2004 HES used the same sample frame
p.000010: (list of clusters) as the 2003 TLLS. A comparison of the distribution of the basic variables from the panel
p.000010: sample against the 2003 cross- section indicates that the panel sample is fairly representative of the overall
p.000010: population, both at rural and urban levels.
p.000010:
p.000010: Both surveys collected information on such household attributes as demographics, education and health, income
p.000010: and expenditures, assets, and consumption. The analysis of poverty dynamics here uses the panel component of the two
p.000010: surveys, and is based on an asset index. Construction of the asset index is described below.
p.000010:
p.000010: In addition to utilizing household-level data, the empirical analysis at the micro (household) level
p.000010: exploits a few key district-level variables to capture agricultural reform and various policy changes
...
...
p.000024: Baschieri, A., and J. Falkingham (2005). ‘Developing a Poverty Map of Tajikistan: A Technical Note’. S3RI
p.000024: Applications and Policy Working Papers, A05/11. Southampton: Southampton Statistical Sciences Research
p.000024: Institute.
p.000024: Becerra, C. A. V. (2004). ‘The World Cotton Market: A Long-Term Outlook’. Working Paper presented at the WTO African
p.000024: Regional Workshop on Cotton, 23-24 March, Cotonou.
p.000024: Canto, O. (2002). ‘Climbing Out of Poverty, Falling Back in: Low Income Stability in Spain’. Applied Econometrics, 34
p.000024: (15): 1903-16.
p.000024: Carter, M. R., and J. May (2001). ‘Poverty, Livelihood and Class in Rural South Africa’. World
p.000024: Development, 27 (1): 1-20.
p.000024: Carter, M. R., and B. C. Barrett (2006). ‘The Economics of Poverty Traps and Persistent Poverty: An Asset-Based
p.000024: Approach’. Journal of Development Studies, 42 (2): 178- 99.
p.000024: Filmer, D., and L. Pritchett (1998). ‘Estimating Wealth Effects without Income or Expenditure Data’. WB
p.000024: Policy Research Paper 1994. Washington, DC: World Bank.
p.000024: Jalan, J., and M. Ravallion (2002). ‘Geographic Poverty Traps? A Micro Model of Consumption Growth in Rural
p.000024: China’. Journal of Applied Econometrics, 17: 329-46.
p.000024: Lawley, D., and A. Maxwell (1971). Factor Analysis as a Statistical Method. London: Butterworth & Co.
p.000024: Sahn, D., and D. Stifel (2000). ‘Poverty Comparisons over Time and across Countries in Africa’. World Development, 28
...
...
p.000024:
p.000024:
p.000024: 24
p.000025:
p.000025: Sahn, D., and D. Stifel (2003). ‘Exploring Alternative Measures of Wealth in the Absence of Expenditure
p.000025: Data’. The Review of Income and Wealth, 49 (4): 463-89.
p.000025: World Bank (2005). ‘Priorities for Sustainable Growth: A Strategy for Agriculture Sector Development in
p.000025: Tajikistan’. Washington, DC: World Bank. Mimeo
p.000025: World Bank (2006). ‘Republic of Tajikistan Poverty Assessment Update’. Report No.
p.000025: 30853-TJ. Washington, DC: World Bank.
p.000025: Ukaeva, U. (2005). ‘Decomposition of Agricultural Growth in Tajikistan’. Washington, DC: World Bank. Mimeo.
...
Searching for tag political:
(return to top)
...
p.000008:
p.000008: 3.1 Using an asset-based poverty line to identify poverty transitions
p.000008:
p.000008: The literature on poverty dynamics has increasingly recognized the importance of adopting an asset-based
p.000008: approach to study changes in wellbeing, especially in response to a wide range of different (climatic,
p.000008: health, political and other) shocks.10 Differentiating between stochastic and structural poverty
p.000008: transitions implies the availability of information on assets and expected levels of wellbeing. To illustrate the
p.000008: importance of the assets-based approach in capturing welfare-status changes, we use the conceptual framework
p.000008: advocated by Carter and May (2001) and Carter and Barrett (2006). This framework is presented in Figure 6.
p.000008:
p.000008: In essence, in any timeperiod a household can be regarded as structurally poor if household consumption
...
Searching for tag poor:
(return to top)
...
p.000005:
p.000005: A U-curve relationship exists between the share of land under cultivation in mountainous terrain
p.000005: and poverty headcount, while the share of pastoral land is not correlated with poverty. Overall, about
p.000005: 60 per cent of Tajikistan is covered by mountainous terrain, with significant differences across the
p.000005: rayons. The data suggest that both types of territories, whether encompassing an insignificant or a
p.000005: significant percentage of mountainous terrain, are likely to be very poor, with a poverty headcount of about 80 per
p.000005: cent (Panel A, Figure 4). The share of pastoral land does not seem to be a factor (Panel C, Figure 4).
p.000005:
p.000005: A higher share of irrigated farming land is associated with somewhat lower levels of poverty. However, the
p.000005: level of irrigation is a very weak correlate of poverty. The data indicate that even well-irrigated areas are likely to
p.000005: have huge variations in the level of poverty, ranging from a high of 80 per cent to a low of 40 per cent (Panel B,
...
...
p.000008: health, political and other) shocks.10 Differentiating between stochastic and structural poverty
p.000008: transitions implies the availability of information on assets and expected levels of wellbeing. To illustrate the
p.000008: importance of the assets-based approach in capturing welfare-status changes, we use the conceptual framework
p.000008: advocated by Carter and May (2001) and Carter and Barrett (2006). This framework is presented in Figure 6.
p.000008:
p.000008: In essence, in any timeperiod a household can be regarded as structurally poor if household consumption
p.000008: falls below the consumption poverty line u* and its stock of
p.000008:
p.000008:
p.000008:
p.000008: 9 The graph is not presented here, but available from the authors on request.
...
...
p.000009:
p.000009: Source: Carter and Barrett (2006).
p.000009:
p.000009:
p.000009: assets falls below the asset poverty line A.11 Such a state is described by point B in Figure 6. A
p.000009: household can be regarded as stochastically poor if it holds assets above level A, yet its level of consumption is
p.000009: below the poverty line u* (described by point E above). A household that has over time moved from below to above the
p.000009: consumption poverty line u* could be regarded as having made a stochastic transition out of poverty if its assets are
p.000009: still mapped below the asset poverty line A. This case is represented in by the shift from point B to point C, which
p.000009: may occur because of increased crops in a given year due to favourable weather conditions. As a result, the
p.000009: livelihood function shifted upwards from u (A) to u’ (A), reflecting the increased returns on existing assets.
...
...
p.000013:
p.000013:
p.000013: 0 0.2 0.4 0.6 0 0.2
p.000013: 0.4 0.6
p.000013:
p.000013: Rural households that became non-poor
p.000013: Panel C: % of HHs acquiring household items Panel D: % of HHs disposing off household items
p.000013:
p.000013:
p.000013:
p.000013:
...
...
p.000013:
p.000013: 0 0.2 0.4 0.6 0 0.2
p.000013: 0.4 .0.6
p.000013:
p.000013:
p.000013: Rural HHs that became poor
p.000013: Panel E: % of HHs acquiring household items Panel F: % of HHs disposing off household items
p.000013:
p.000013:
p.000013:
p.000013:
...
...
p.000013:
p.000013:
p.000013: 13
p.000014:
p.000014: About 28 per cent of the households that had exited poverty bought a new refrigerator (Panel C, Figure 7), while about
p.000014: 22 per cent had disposed of one (Panel D, Figure 7).15 Households that had become (asset) poor displayed a
p.000014: substantial shedding of assets (Panel F, Figure 7). Only wood stoves and black and white TVs were among the goods
p.000014: acquired by this household group (Panel E, Figure 7). As mentioned above, scoring coefficient for a wood
p.000014: stove is close to zero, while it is negative for a black and white TV. The presented data clearly suggest that the
p.000014: ownership of assets by rural households in Tajikistan has been far from a static process, even in a period as short as
p.000014: one year.
p.000014:
p.000014: We set up an asset-based poverty line at a level equivalent to the 50th percentile of the asset index distribution
p.000014: (using 2003 distribution). The chosen cut-off level is rather arbitrary, but is consistent with the fact
p.000014: that over half of the population—based on a welfare indicator of per capita consumption and the poverty line of
p.000014: US$2.25 per day—is estimated to be poor. The 50th percentile cut-off level was also used in a previous study of poverty
p.000014: dynamics in urban and rural Tajikistan (Angel-Urdinola, Mete and Cnobloch 2008).16
p.000014:
p.000014: Given this asset-based poverty line, what was the extent of the poverty mobility among rural households in the panel?
p.000014: The poverty mobility matrix is presented in Table 2. Out of 322 households qualifying as asset-poor in 2003 (base year)
p.000014: 113 households, or 35 per cent, had moved out of poverty a year later. Out of 267 households classified
p.000014: as non-poor in the base year, 91 households, or 34 per cent, had become poor one year later. In terms of
p.000014: the share of the total panel sample, 19 per cent of households had shed poverty and 15 per cent had become
p.000014: impoverished. These findings confirm substantial mobility in asset holdings even over a relatively short
p.000014: period of time. The regression analysis in the following section attempts to explain this mobility with
p.000014: an array of variables at the household, community and district level.
p.000014:
...
...
p.000014: % of total
p.000014: % of row
p.000014: total N
p.000014: % of total
p.000014: % of row total
p.000014: Poor 209 35.5 64.9 113 19.2 35.1 322
p.000014: 54.7 100.0
p.000014: Non-poor 176 29.9 65.9 91 15.4 34.1 267
p.000014: 45.3 100.0
p.000014:
p.000014: Total 385 65.4 65.4 204 34.6 34.6 589
p.000014: 100.0 100.0
p.000014: Source: Authors’ estimates, based on 2003 LSMS and 2004 EHS.
...
...
p.000001: Pr( 0
p.000001: 0
p.000001: i,t -1
p.000001: = 0; Di, t-1, Xi,t-1, XR,,t-1, β) = Φ(Di, t-1, Xi,t-1, XR,,t-1, β) (2)
p.000001:
p.000001: Equation (1) models the probability of a household to be non-poor in period t (2004) conditional on being poor in
p.000001: period t-1 (2003). Equation (2) models the probability of a household to be poor in period t (2004) conditional
p.000001: on being non-poor in period t-1 (2003). As is already clear from the discussion, P0 is the indicator of being
p.000001: poor based on the asset poverty line. Both equations are modelled conditional on a household’s distance
p.000001: from the poverty line in period t-1, which is denoted by Di, t-1.17 Xi,t-1 denotes various household-level
p.000001: characteristics, and XR,t-1 denotes various regional-level characteristic at the region (oblast),
p.000001: district (rayon) and community (village) level. β denotes the vector of parameters. It is important to
p.000001: note that although for ease of presentation XR,t-1 indicates that the variable is expressed in levels at
p.000001: time t-1, some variables in the model actually capture changes occurring before t-1. For instance, a
...
...
p.000001: predictors of moving out of and into poverty. Model 1 looks at poverty mobility predictors where community-level
p.000001: characteristics are not included among the regressors (Appendix Table 1). It includes household head characteristics
p.000001: such as age, gender, education, and self-reported health status; household demographics such as the share of adults;
p.000001: employment status of household members; and regional (oblast) dummies. Model 1 is estimated using two
p.000001: specifications, with and without district (rayon) characteristics. For ease of comparison of the coefficients, the
p.000001: estimation results of the moving-out of poverty model are presented next to the estimation results of the becoming-poor
p.000001: model.
p.000001:
p.000001:
p.000001:
p.000001: 17 The poverty mobility literature often uses a specification in which the event is conditional on the
...
...
p.000022: working-age individuals in a household increases. The estimates suggest that the probability of falling into poverty
p.000022: declines from 55 to 10 per cent when the share of adults increases from 25 to 100 per cent (Panel A, Figure 10).
p.000022: Employment in public administration reduces the risk of impoverishment: the probability of falling into poverty
p.000022: drops from 20 to 3 per cent as the share of adults in administrative employment increases from 25 to 100
p.000022: per cent (Panel B, Figure 10). Household size is also a factor: the estimates suggest a U-shaped relationship between
p.000022: the probability of becoming poor and household size (Appendix Table 2). In other words, small households
p.000022: (elderly people living alone) and very large households (usually households with many children) face a higher risk of
p.000022: falling into poverty than the average-sized household.
p.000022:
p.000022: Examining the impact of district/community level characteristics, we find that variations in rain fall are
p.000022: associated with the risk of becoming poor. Households located in communities with less than average amount of
p.000022: rain over the previous year face a 55 per cent chance of becoming poor versus 28 per cent for households in
p.000022: non-drought communities (Panel C, Figure 10). Agricultural employment in cotton producing areas is associated with a
p.000022: higher risk of impoverishment compared to similar employment in non-cotton producing areas.20 According to estimates,
p.000022: if half of the adults in a cotton- producing district work in the agricultural sector, the risk of poverty
p.000022: in the cotton- producing areas is 55 per cent, but only 25 per cent in non-cotton areas. However,
p.000022: increasing the share of household members employed in agriculture in cotton-producing areas improves the odds of not
...
...
p.000022: its impact on the probability of moving out of poverty). Households in the RRS region face a 65 per cent probability of
p.000022: falling into poverty, while this is 20 per cent for those living in Khatlon (Panel A, Figure 11). A
p.000022: higher share of land under cotton cultivation in the district increases the risk of falling into poverty: once the
p.000022: share of cotton-cultivated land increases from 40 to 60 per cent, this compounds the odds from 20 to 40 per
p.000022: cent (Panel B, Figure 11). Distance to market is also a factor: the estimates suggest that as distance increases,
p.000022: the probability of becoming poor generally declines (Panel C, Figure 11). The explanation here is likely to be the same
p.000022: as discussed above in the case of moving out of poverty. A larger proportion of adults in agricultural employment
p.000022: reduces the probability of falling into poverty. But the effect is very marginal. As the share of
p.000022: agricultural employers increases from 25 to 75 per cent, the odds of becoming poor decline from 22 to 16 per cent
p.000022: (Panel D, Figure 11).
p.000022:
p.000022:
p.000022: 20 The household survey data do not specify how many adults are actually employed in the cotton farm (no cotton is
p.000022: produced on household plots). However, it is safe to assume that agricultural employment in cotton-producing
...
Searching for tag poverty:
(return to top)
...
p.000001:
p.000001:
p.000001:
p.000001:
p.000001: Research Paper No. 2008/26
p.000001: Asset-Based Poverty in Rural Tajikistan
p.000001: Who Climbs out and Who Falls in?
p.000001: Oleksiy Ivaschenko and Cem Mete*
p.000001: March 2008
p.000001:
p.000001: Abstract
p.000001:
p.000001: Tajikistan’s rural sector has witnessed substantial development since the country began to emerge from civil conflict
p.000001: in 1999. Gross agricultural output increased 64 per cent from 1999 to 2003, and there were significant developments in
p.000001: the agricultural reform agenda. This paper uses the panel component of two surveys conducted in Tajikistan at one-year
p.000001: interval (2003 and 2004) to explore the major determinants of the transition out of/into poverty of rural households.
p.000001: Poverty status is measured in the asset space, thus indicating structural rather than transitory poverty
p.000001: movements. The empirical analysis reveals several interesting findings that are also important from a
p.000001: policy perspective: first, cotton farming seems to have no positive impact on poverty levels, nor on mobility out
p.000001: of poverty. Second, the rate of increase in the share of private farming at the district level had
p.000001: little impact on poverty levels and poverty mobility.
p.000001: …/.
p.000001: Keywords: welfare, poverty, Tajikistan JEL classification: I3
p.000001:
p.000001: Copyright © UNU-WIDER 2008
p.000001: * ECSHD, World Bank, Washington, DC: emails: oivaschenko@worldbank.org (O. Ivaschenko);
p.000001: cmete@worldbank.org (C. Mete)
p.000001: This is a revised version of a paper originally prepared for the UNU-WIDER conference on Fragile
...
...
p.000001: Department for International Development—DFID. Programme contributions are also received from the governments
p.000001: of Denmark (Royal Ministry of Foreign Affairs), Norway (Royal Ministry of Foreign Affairs) and Sweden
p.000001: (Swedish International Development Cooperation Agency— Sida).
p.000001: ISSN 1810-2611 ISBN 978-92-9230-072-2
p.000001:
p.000001: Third, there is strong evidence of geographic poverty mobility traps in Tajikistan. Higher levels of
p.000001: poverty in a district appear to reduce significantly the chance of a household shedding poverty. Living in a
p.000001: region with overall slow economic growth is also found to undermine the odds of exiting poverty and to increase the
p.000001: risk of falling into poverty. Finally, several key household-level factors, such as the share of adults, education
p.000001: level, health status and participation in wage employment, also emerge as significant predictors of poverty
p.000001: mobility.
p.000001:
p.000001:
p.000001: Authors notes
p.000001:
...
...
p.000001: endorsement by the Institute or the United Nations University, nor by the programme/project sponsors, of any of the
p.000001: views expressed.
p.000001:
p.000001: 1 Introduction
p.000001:
p.000001: This paper exploits an asset-based approach to study the (asset-based) poverty dynamics of Tajikistan rural
p.000001: households. We use a panel of rural households that have been observed during two time periods: June-July
p.000001: 2003 and July-November 2004.
p.000001:
p.000001: Analysing the dynamics of rural poverty in Tajikistan during this time period is particularly interesting
p.000001: in view of the drastic changes that have occurred in the country over the last several years. Emerging in 1999 from
p.000001: civil war and a prolonged period of economic collapse,1 the country’s economic performance has been impressive from the
p.000001: year 2000, with sustained real GDP annual growth rates of 7 to 9 per cent.2
p.000001:
p.000001: Economic growth has been accompanied with substantial reduction in poverty, dropping from 81 per cent of the population
p.000001: living below the poverty line (US$2.15 per day) in 1999 to 64 per cent in 2003 (World Bank 2006). Although
p.000001: poverty headcount fell during this period by 19 percentage points in the rural areas compared to 14 percentage points
p.000001: in urban centres, it remains higher in the rural regions: 65 per cent versus 59 per cent. As 73 per cent of the
p.000001: population live in the countryside, poverty in Tajikistan continues to be an overwhelmingly rural
p.000001: phenomenon. Economic growth and the resultant poverty reduction are explained by three major factors: (i) conflict
p.000001: cessation, which allowed economic activity to resume and markets to develop; (ii) initial impact of the macroeconomic
p.000001: stability and agricultural reforms in the non-cotton sector that enabled farmers to diversify production
p.000001: and increase productivity; and (iii) large increase in migrant workers exiting Tajikistan for Russia and other
p.000001: countries. However, there have been concerns that once the initial benefits of these ‘special’ factors dry out,
p.000001: Tajikistan’s poverty reduction trends may not be sustainable (World Bank 2006).
p.000001:
p.000001: In view of the sound economic growth rates, markedly reduced but still very high rural poverty, and concerns over the
p.000001: sustainability of the country’s poverty reduction trend, it is important from a policy perspective to understand
p.000001: the key factors at the micro (household/community) level that explain the transition of rural households in and
p.000001: out of poverty. This is the main objective of this paper.
p.000001:
p.000001: The paper contributes to the literature on welfare dynamics in general, and to the studies of poverty in Tajikistan in
p.000001: particular on several fronts:
p.000001:
p.000001: i) utilizing an assets-based approach to better capture the permanent (as opposed to transitory) component of
p.000001: welfare changes for rural households;
p.000001: ii) investigating explicitly the importance of community/local factors versus household/individual level
p.000001: characteristics to explain movements in and out of (asset-based) poverty.
p.000001: It is worth noting that a study of the general factors affecting poverty transition in Tajikistan has
p.000001: been undertaken by Angel-Urdinola, Mete and Cnobloch (2008). Our
p.000001:
p.000001:
p.000001: 1 Looking at cotton output across rayons (the smallest administrative unit), we observe that between 1991
p.000001: and 1999, there was an average output decline of 62 per cent. Cotton output has increased since 1999 by an average of
...
...
p.000001:
p.000001: 1
p.000002:
p.000002: study expands on this work by focusing specifically at the determinants of welfare dynamics in rural
p.000002: areas. More specifically, we combine household survey data with district (rayon) level data to take into
p.000002: account the poverty impact of community-level factors such as the share of private (dekhan) farms, per
p.000002: hectare of land under cotton cultivation, level of debt and distance to market or district centre.3
p.000002:
p.000002: The paper is structured as follows. Section 2 discusses the main developments in Tajikistan’s rural
p.000002: sector since 1999, as well as the correlates of rural poverty at the district (rayon) level. Section 3
p.000002: provides the theoretical and empirical framework for the analysis of the (asset-based) poverty mobility at the
p.000002: household level. It also describes the data and the constructed asset index. Section 4 presents the empirical results,
p.000002: and section 5 concludes with a summary of the main findings and their policy implications.
p.000002:
p.000002:
p.000002: 2 Tajikistan’s rural sector developments since 1999, and correlates of rural poverty at the district (rayon) level
p.000002:
p.000002: 2.1 Rural sector developments since 1999
p.000002:
p.000002: Tajikistan’s rural sector has witnessed substantial changes since the country emerged from civil conflict
p.000002: in 1999. These include agricultural reform, and specifically the rapidly changing structure of land ownership;
...
...
p.000004: 4
p.000005:
p.000005: advised pricing arrangements with investors, have resulted in dubious ‘debts’ that the producers are struggling to
p.000005: repay.8 These developments are expected to have an impact on the standards of living of Tajikistan’s rural population.
p.000005:
p.000005: 2.2 The correlates of rural poverty at the rayon level
p.000005:
p.000005: How are the developments of the rural sector, as outlined above, correlated with poverty? To gain some
p.000005: insights into this issue, we look at the correlation of selected key variables of the rural section at the rayon level
p.000005: with the poverty headcount at a similar level, as obtained from the poverty mapping conducted in Tajikistan
p.000005: (Baschieri and Falkingham 2005). Some of the variables will be used later to explain poverty mobility at the
p.000005: household level. A number of interesting findings emerge from simple scatter plots of the district-level data
p.000005: (Figures 4 and 5).
p.000005:
p.000005: A U-curve relationship exists between the share of land under cultivation in mountainous terrain
p.000005: and poverty headcount, while the share of pastoral land is not correlated with poverty. Overall, about
p.000005: 60 per cent of Tajikistan is covered by mountainous terrain, with significant differences across the
p.000005: rayons. The data suggest that both types of territories, whether encompassing an insignificant or a
p.000005: significant percentage of mountainous terrain, are likely to be very poor, with a poverty headcount of about 80 per
p.000005: cent (Panel A, Figure 4). The share of pastoral land does not seem to be a factor (Panel C, Figure 4).
p.000005:
p.000005: A higher share of irrigated farming land is associated with somewhat lower levels of poverty. However, the
p.000005: level of irrigation is a very weak correlate of poverty. The data indicate that even well-irrigated areas are likely to
p.000005: have huge variations in the level of poverty, ranging from a high of 80 per cent to a low of 40 per cent (Panel B,
p.000005: Figure 4).
p.000005:
p.000005: A larger portion of land under dekhan cultivation is correlated with lower poverty levels. However, the
p.000005: increase in dekhan farming land between 2000-04 shows no correlation with poverty. The level of dekhan farming
p.000005: in 2000 (prior to its substantial increase in the structure of land ownership) seems to be negatively
p.000005: correlated with poverty headcount (Panel D, Figure 4). However, additional dekhan cultivation has not been reflected
p.000005: in poverty levels (Panel E, Figure 4). This finding is consistent with other evidence which suggests that the
p.000005: increasing ratio of dekhan cultivation has not been accompanied by improved productivity on these farms (World
p.000005: Bank 2005). Moreover, many of these dekhan farms are in cotton production, and have thus been affected by the sector’s
p.000005: adverse development.
p.000005:
p.000005:
...
...
p.000005:
p.000005:
p.000005: 5
p.000006:
p.000006: Figure 4
p.000006: Correlations between rayon (district) poverty headcount* and land variables
p.000006:
p.000006:
p.000006:
p.000006:
p.000100: 100
...
...
p.000051:
p.000040: 40
p.000041:
p.000030: 30
p.000031:
p.000031: Panel (A): Poverty vs. mountainous terrain
p.000031:
p.000031:
p.000031:
p.000100: 100
p.000101:
...
...
p.000051:
p.000040: 40
p.000041:
p.000030: 30
p.000031:
p.000031: Panel (B): Poverty vs. % of farming land which is irrigated
p.000031:
p.000020: 20
p.000021:
p.000021:
p.000010: 10
...
...
p.000011:
p.000011:
p.000011:
p.000011:
p.000100: 100
p.000101: Panel (C): Poverty vs. % of pastoral land
p.000101:
p.000101:
p.000101: 100
p.000101: Panel (D): Poverty vs. % of farming under dekhan farms
p.000101:
p.000090: 90
p.000091: 90
p.000091:
p.000080: 80
...
...
p.000031:
p.000031:
p.000031:
p.000031:
p.000100: 100
p.000101: Panel (E): Poverty vs. change in area under dekhan farms
p.000101:
p.000101:
p.000101: 100
p.000101: Panel (F): Poverty vs. % of arable land under household plots
p.000101:
p.000090: 90
p.000091: 90
p.000091:
p.000080: 80
...
...
p.000031:
p.000031:
p.000006: 6
p.000007:
p.000007: Figure 5
p.000007: Correlations between rayon (district) poverty headcount* and cotton production
p.000007:
p.000007:
p.000007:
p.000007:
p.000100: 100
p.000101:
p.000101: Panel (A): Poverty vs. % of arable land under cotton
p.000101:
p.000101:
p.000101: 100
p.000101:
p.000101: Panel (B): Poverty vs. mean wage on cotton farms
p.000101:
p.000101: 90 90
p.000101:
p.000101: 80 80
p.000101:
...
...
p.000031: mean monthly wage on cotton farms, Somoni (2003)
p.000031:
p.000031:
p.000031:
p.000100: 100
p.000101: Panel (C): Poverty vs. cotton debt (per capita)
p.000101:
p.000101: 100
p.000101: Panel (D): Poverty vs. cotton output (2003 as % of 1991)
p.000101:
p.000101: 90 90
p.000101:
p.000101: 80 80
p.000101:
...
...
p.000031: cotton output, tones (2003 as % of 1991)
p.000031:
p.000031:
p.000031:
p.000100: 100
p.000101: Panel (E): Poverty vs. % change in cotton output (1999 to 2003)
p.000101:
p.000101: 100
p.000101: Panel (F): Poverty vs. % change in value of cotton (1999 t
p.000101:
p.000101: 90 90
p.000101:
p.000101: 80 80
p.000101:
...
...
p.000031:
p.000031:
p.000031:
p.000007: 7
p.000008:
p.000008: A larger share of arable land under household plots is weakly associated with lower poverty levels. The
p.000008: share of this type of land has remained quite stable over time at about 20 per cent. It would appear that
p.000008: districts with a somewhat higher than average share of land under household plots exhibit lower levels of poverty
p.000008: (Panel F, Figure 4). The importance of this type of farming to overall agricultural production is, however, unlikely to
p.000008: change.
p.000008:
p.000008: The districts with more extensive cotton cultivation are likely to have higher poverty levels (Panel A,
p.000008: Figure 5). As can be expected, there is a negative correlation between a district’s average cotton-farm wages and
p.000008: poverty levels (Panel B, Figure 5). Although, this relationship is at least partly driven by the fact that in poorer
p.000008: districts, cotton-farm workers are paid lower wages. We also find that the average wage arrears per person are higher
p.000008: in the poorer areas than elsewhere.9 These findings suggest that in poorer areas cotton-farm labourers receive smaller
p.000008: wages, and they are less likely to be paid on time. We find a very weak correlation between the level of the
p.000008: cotton-farm debt (per capita) and poverty at the district level.
p.000008:
p.000008: The extent of decline in output during the 1990s is strongly (positively) associated with poverty. Based on cotton
p.000008: output data, we find that districts with greater output gaps between 1991 (the peak output year before
p.000008: economic collapse) and 2003 are much more likely to be poorer. The deviations in cotton output and its value between
p.000008: 1999-2003 are not correlated with the levels of poverty at the district level.
p.000008:
p.000008: The above examination of the key correlates of poverty at the district level provides a solid basis for analysing the
p.000008: (asset-based) poverty mobility at the household level in the next section.
p.000008:
p.000008:
p.000008: 3 Theoretical and empirical framework for an asset-based analysis of poverty mobility
p.000008:
p.000008: 3.1 Using an asset-based poverty line to identify poverty transitions
p.000008:
p.000008: The literature on poverty dynamics has increasingly recognized the importance of adopting an asset-based
p.000008: approach to study changes in wellbeing, especially in response to a wide range of different (climatic,
p.000008: health, political and other) shocks.10 Differentiating between stochastic and structural poverty
p.000008: transitions implies the availability of information on assets and expected levels of wellbeing. To illustrate the
p.000008: importance of the assets-based approach in capturing welfare-status changes, we use the conceptual framework
p.000008: advocated by Carter and May (2001) and Carter and Barrett (2006). This framework is presented in Figure 6.
p.000008:
p.000008: In essence, in any timeperiod a household can be regarded as structurally poor if household consumption
p.000008: falls below the consumption poverty line u* and its stock of
p.000008:
p.000008:
p.000008:
p.000008: 9 The graph is not presented here, but available from the authors on request.
p.000008: 10 See February 2006 special issue of the Journal of Development Studies, 42 (2) for the set of papers presenting the
...
...
p.000008:
p.000008:
p.000008: 8
p.000009:
p.000009: Figure 6
p.000009: Poverty transitions: consumption poverty line vs. asset poverty line
p.000009:
p.000009:
p.000009:
p.000009: Utility
p.000009: Asset poverty line
p.000009:
p.000009:
p.000009:
p.000009:
p.000009:
p.000009:
p.000009:
p.000009:
p.000009: Consumption poverty line
p.000009:
p.000009:
p.000009:
p.000009:
p.000009:
...
...
p.000009:
p.000009:
p.000009: Source: Carter and Barrett (2006).
p.000009:
p.000009:
p.000009: assets falls below the asset poverty line A.11 Such a state is described by point B in Figure 6. A
p.000009: household can be regarded as stochastically poor if it holds assets above level A, yet its level of consumption is
p.000009: below the poverty line u* (described by point E above). A household that has over time moved from below to above the
p.000009: consumption poverty line u* could be regarded as having made a stochastic transition out of poverty if its assets are
p.000009: still mapped below the asset poverty line A. This case is represented in by the shift from point B to point C, which
p.000009: may occur because of increased crops in a given year due to favourable weather conditions. As a result, the
p.000009: livelihood function shifted upwards from u (A) to u’ (A), reflecting the increased returns on existing assets.
p.000009:
p.000009: The stochastic transition into poverty is represented here by the shift from point D to point E, whereby household
p.000009: consumption drops below the poverty line, as returns to existing assets temporarily diminish, but the level of
p.000009: asset holdings stays above the asset poverty line. This is exemplified by the household that has experienced
p.000009: a temporary consumption decline because of a negative shock (e.g., drought), but is expected to bounce
p.000009: back to a level of consumption above the poverty line. A household that has shifted over time across the
p.000009: consumption poverty line u* could be regarded as having made a structural transition out of poverty, if it has
p.000009: also accumulated sufficient additional assets to move above the asset poverty line A (represented here by the shift
p.000009: from point B to point D). Conversely, the shift from point D to point B would represent the structural transition
p.000009: into poverty. As Figure 6 indicates, there could be multiple options for poverty transition paths, but
p.000009: the main point here is that changes in
p.000009:
p.000009:
p.000009: 11 The asset poverty line in Figure 1 is simply the level of assets corresponding to the level of wellbeing equal to
p.000009: the consumption poverty line.
p.000009:
p.000009:
p.000009: 9
p.000010:
p.000010: consumption that are not accompanied by changes in the assets base can be regarded as stochastic rather than structural
...
...
p.000010: (list of clusters) as the 2003 TLLS. A comparison of the distribution of the basic variables from the panel
p.000010: sample against the 2003 cross- section indicates that the panel sample is fairly representative of the overall
p.000010: population, both at rural and urban levels.
p.000010:
p.000010: Both surveys collected information on such household attributes as demographics, education and health, income
p.000010: and expenditures, assets, and consumption. The analysis of poverty dynamics here uses the panel component of the two
p.000010: surveys, and is based on an asset index. Construction of the asset index is described below.
p.000010:
p.000010: In addition to utilizing household-level data, the empirical analysis at the micro (household) level
p.000010: exploits a few key district-level variables to capture agricultural reform and various policy changes
p.000010: (discussed earlier). The analysis also uses community survey data from the 2003 TLLS, to allow us to identify
p.000010: whether cotton was grown in a particular community, whether rainfall during the survey year was
p.000010: better/worse relative to the previous year, as well as certain other important community-level
p.000010: characteristics that are likely to be associated with a household’s mobility out of or into poverty.
p.000010:
p.000010: 3.3 Constructing the asset index and asset-based poverty line
p.000010:
p.000010: In order to construct an asset index, we rely on principal component analysis (Lawley and Maxwell 1971).13 The
p.000010: principal component constitutes a linear index capturing most of the information (variance) common to all
p.000010: the variables. Denoted by Aij the observation for household i and asset j (for example, whether or not a household
p.000010: has a
...
...
p.000012: However, as discussed in greater detail later, according to the data, significant changes in rural households concern
p.000012: the possession of durable goods (which make up the asset index).14 In fact, rural households accumulated major durable
p.000012: goods faster than urban households (Table 1).
p.000012:
p.000012: Moreover, as our attempt is to understand the determinants of rural households in moving in/out of
p.000012: assets-based poverty, the exclusion of agricultural assets and livestock in the asset index may even be advisable. This
p.000012: is because household agricultural assets and community agricultural characteristics (explanatory variables in
p.000012: our regression model) are likely to be determined simultaneously by such factors as agricultural reform,
p.000012: thus representing an endogeneity problem.
p.000012:
p.000012: Analysing changes in the possession of the durables making up our asset index, we find that the average (per
...
...
p.000012: It is important to note that the scoring coefficients are estimated with the pooled 2003 and 2004 samples
p.000012: (panel), which make the estimated asset indices fully comparable between the two years.
p.000012:
p.000012: It is worth noting that there was a substantial mobility in the ownership of various assets, with households acquiring
p.000012: and disposing of assets. For instance, about 20 per cent of rural households overall acquired a colour TV (Panel A,
p.000012: Figure 7), while this figure was almost 60 per cent among the households who had moved out of poverty (Panel C, Figure
p.000012: 7). About 15 per cent of rural households bought a video player (Panel A, Figure 7), while at the same time about 5 per
p.000012: cent of these households got rid of one (Panel B, Figure 7).
p.000012:
p.000012:
p.000012: 14 Another important consideration is that an assets index that includes only durables may be better at capturing
p.000012: transitions out of structural poverty, but not transitions into structural poverty (unless rural household, when faced
p.000012: with hardships, prefer to sell durable goods before selling agricultural assets or livestock). However, application of
p.000012: the asset index to the panel data indicates that in the rural regions the share of households escaping poverty (35.1
p.000012: per cent) is almost equivalent to those who fall into poverty (34.1 per cent).
p.000012:
p.000012:
p.000012: 12
p.000013:
p.000013: Figure 7
...
...
p.000013: Source: Authors’ estimates, based on 2003 LSMS and 2004 EHS.
p.000013:
p.000013:
p.000013: 13
p.000014:
p.000014: About 28 per cent of the households that had exited poverty bought a new refrigerator (Panel C, Figure 7), while about
p.000014: 22 per cent had disposed of one (Panel D, Figure 7).15 Households that had become (asset) poor displayed a
p.000014: substantial shedding of assets (Panel F, Figure 7). Only wood stoves and black and white TVs were among the goods
p.000014: acquired by this household group (Panel E, Figure 7). As mentioned above, scoring coefficient for a wood
p.000014: stove is close to zero, while it is negative for a black and white TV. The presented data clearly suggest that the
p.000014: ownership of assets by rural households in Tajikistan has been far from a static process, even in a period as short as
p.000014: one year.
p.000014:
p.000014: We set up an asset-based poverty line at a level equivalent to the 50th percentile of the asset index distribution
p.000014: (using 2003 distribution). The chosen cut-off level is rather arbitrary, but is consistent with the fact
p.000014: that over half of the population—based on a welfare indicator of per capita consumption and the poverty line of
p.000014: US$2.25 per day—is estimated to be poor. The 50th percentile cut-off level was also used in a previous study of poverty
p.000014: dynamics in urban and rural Tajikistan (Angel-Urdinola, Mete and Cnobloch 2008).16
p.000014:
p.000014: Given this asset-based poverty line, what was the extent of the poverty mobility among rural households in the panel?
p.000014: The poverty mobility matrix is presented in Table 2. Out of 322 households qualifying as asset-poor in 2003 (base year)
p.000014: 113 households, or 35 per cent, had moved out of poverty a year later. Out of 267 households classified
p.000014: as non-poor in the base year, 91 households, or 34 per cent, had become poor one year later. In terms of
p.000014: the share of the total panel sample, 19 per cent of households had shed poverty and 15 per cent had become
p.000014: impoverished. These findings confirm substantial mobility in asset holdings even over a relatively short
p.000014: period of time. The regression analysis in the following section attempts to explain this mobility with
p.000014: an array of variables at the household, community and district level.
p.000014:
p.000014: Table 2
p.000014: Poverty mobility (based on assets) for rural households
p.000014:
p.000014: Retained status, 2004 Changed status, 2004 Total
p.000014:
p.000014:
p.000014: Poverty status, 2003 N
p.000014: % of total
p.000014: % of row
p.000014: total N
p.000014: % of total
p.000014: % of row
...
...
p.000014:
p.000014: 14
p.000015:
p.000015: 3.4 The empirical model
p.000015:
p.000015: The event of a household transiting out of (or into) (asset-based) poverty is modelled within the probability
p.000015: framework. The (ex-post) realization of the event (experience of the transit into (out of) poverty between 2003-04) is
p.000015: used to define the samples at risk of leaving (falling into) poverty. The probability of experiencing a transit out of
p.000015: (or into) poverty is modelled as follows:
p.000015:
p.000015:
p.000015: Pr( 0
p.000000: 0
p.000001: i,t -1
...
...
p.000001: = 0; Di, t-1, Xi,t-1, XR,,t-1, β) = Φ(Di, t-1, Xi,t-1, XR,,t-1, β) (2)
p.000001:
p.000001: Equation (1) models the probability of a household to be non-poor in period t (2004) conditional on being poor in
p.000001: period t-1 (2003). Equation (2) models the probability of a household to be poor in period t (2004) conditional
p.000001: on being non-poor in period t-1 (2003). As is already clear from the discussion, P0 is the indicator of being
p.000001: poor based on the asset poverty line. Both equations are modelled conditional on a household’s distance
p.000001: from the poverty line in period t-1, which is denoted by Di, t-1.17 Xi,t-1 denotes various household-level
p.000001: characteristics, and XR,t-1 denotes various regional-level characteristic at the region (oblast),
p.000001: district (rayon) and community (village) level. β denotes the vector of parameters. It is important to
p.000001: note that although for ease of presentation XR,t-1 indicates that the variable is expressed in levels at
p.000001: time t-1, some variables in the model actually capture changes occurring before t-1. For instance, a
p.000001: community reports rainfall shocks between t-2 and t-1. We also investigate the impact on the poverty mobility of the
p.000001: share of cultivation under private farming at time t-4, as well as the impact of the change in this variable between
p.000001: t-4 and t-1. In other words, some of the explanatory variables are lagged by more than one year; and that
p.000001: some explanatory variables are actually changes rather than levels. These equations are estimated using the
p.000001: maximum-likelihood estimator.
p.000001:
p.000001:
p.000001: 4 Empirical results
p.000001:
p.000001: Appendix Tables 1-3 present the estimation results for three alternative models exploring the
p.000001: predictors of moving out of and into poverty. Model 1 looks at poverty mobility predictors where community-level
p.000001: characteristics are not included among the regressors (Appendix Table 1). It includes household head characteristics
p.000001: such as age, gender, education, and self-reported health status; household demographics such as the share of adults;
p.000001: employment status of household members; and regional (oblast) dummies. Model 1 is estimated using two
p.000001: specifications, with and without district (rayon) characteristics. For ease of comparison of the coefficients, the
p.000001: estimation results of the moving-out of poverty model are presented next to the estimation results of the becoming-poor
p.000001: model.
p.000001:
p.000001:
p.000001:
p.000001: 17 The poverty mobility literature often uses a specification in which the event is conditional on the
p.000001: distance from the poverty line (e.g., Canto 2002). This improves the overall fit of the model and
p.000001: allows one to obtain more accurate parameter estimates on other variables of interest.
p.000001:
p.000001:
p.000015: 15
p.000016:
...
...
p.000016: is produced in the area, and the reported amount of rain compared to the previous rain. Model 2 is also
p.000016: estimated based on two specifications: excluding the interaction between the ‘cotton’ variable (dummy variable
p.000016: indicating if cotton produced in the community) and various other factors (specification 1). Specification 2
p.000016: includes the interactions between the ‘cotton’ variable and such factors as distance to market, share of household
p.000016: adults working in agriculture, education status and gender of the household head. These interaction terms are designed
p.000016: to gain a better understanding of the importance of the ‘cotton’ variable in explaining the poverty transitions.
p.000016:
p.000016: Model 3 uses a richer set of district-level characteristics that apply only to cotton- producing districts
p.000016: in order to get a better understanding of poverty mobility in these areas (Appendix Table 3). It is worth noting
p.000016: that two-thirds of the households in the rural panel sample reside in the cotton-producing districts. Here, we
p.000016: investigate a few, very important agricultural policy variables that impact on poverty mobility: the share of total
p.000016: arable land under cotton, cotton farm debt (per hectare of cotton-cultivated land), share of arable land
p.000016: under dekhan (private farms) in 2000 (prior to its significant increase), and the change in the share of dekhan land
p.000016: between 2000-04. Again, Model 3 is estimated with two specifications: specification 1 includes the interaction terms
p.000016: for the ‘cotton’ variable, while specification 2 ignores the interaction terms.
p.000016:
p.000016: While the main purpose of this exercise is to explore the effects of various additional variables, it also enables us
p.000016: to investigate the robustness of the regression results. Next, we discuss the effects of the variables that had
p.000016: statistically significant coefficient estimates across different specifications.18
p.000016:
p.000016: 4.1 The main predictors of moving out of poverty
p.000016:
p.000016: The sample of all rural households
p.000016: The probability of climbing out of poverty is significantly affected by both geographical factors and household-level
p.000016: characteristics. The level of poverty in a district is a significant predictor of poverty mobility at the
p.000016: household level. The estimates indicate that, after controlling for other characteristics, for a household located in a
p.000016: district with 30 per cent poverty headcount (based on the US$2.15 per day poverty line), there is a 70 per cent
p.000016: probability of escaping poverty, while the probability is a mere 5 per cent for a household located in a district
p.000016: with a poverty headcount of 90 per cent (Panel A, Figure 8).19 Thus living in a region with
p.000016: weak economic growth performance
p.000016:
p.000016:
p.000016:
p.000016: 18 For calculating the predicted probabilities of moving out of/into poverty for the overall rural sample, we use the
p.000016: regression results given in Appendix Table 2 (with interaction terms); for the sample of rural households living in
p.000016: cotton-producing areas, we use the regression results given in Appendix Table 1 (with interaction terms).
p.000016: However, as already mentioned earlier, in calculating these predictions we focus only on the variables
p.000016: which produced robust effects across the different specifications.
p.000016: 19 The horizontal line in the graphs indicates the predicted probability of climbing out of, or falling into, poverty
p.000016: at the means of variables in the estimation sample. In other words, this line indicates the average odds
p.000016: of poverty transition.
p.000016:
p.000016:
p.000016: 16
p.000017:
p.000017: Figure 8
p.000017: Determinants of the probability of MOVING OUT of poverty
p.000017:
p.000017:
p.000017: Panel (A): Prob. of moving into poverty vs. % of adults in the household Panel (B): Prob. of moving
p.000017: into poverty vs. % of adults in admin. employment
p.000017:
p.000017:
p.000017: share of adults=25% share of adults=50% share of adults=75% share of adults=100%
p.000017:
p.000017: share of adults=25% share of adults=50% share of adults=75% share of adults=100%
...
...
p.000017:
p.000017:
p.000017:
p.000017:
p.000017:
p.000017: Panel (C): Prob. of moving into poverty vs. rain shock Panel (D): Prob. of
p.000017: moving into poverty vs. cotton & % of adults in agriculture
p.000017:
p.000017:
p.000017: less rain normal rain
p.000017:
p.000017: Cotton & Share = 50 %
...
...
p.000017:
p.000017:
p.000017: 17
p.000018:
p.000018: Figure 9
p.000018: Determinants of the probability of MOVING OUT of poverty (cotton-producing districts only)
p.000018:
p.000018:
p.000018: Panel (A): Prob. of moving out of poverty vs. region Panel (B):
p.000018: Prob. of moving out of poverty vs. % of land under dekhan farms
p.000018:
p.000018:
p.000018: region=Khatlon region=GBAO region=Sugd region=RRS
p.000018:
p.000018: dekhan farms = 10% dekhan farms = 15% dekhan farms = 20%
...
...
p.000018:
p.000018:
p.000018:
p.000018:
p.000018:
p.000018: Panel (C): Prob. of moving out of poverty vs. cotton farm debt per ha ($) Panel (D): Prob. of
p.000018: moving out of poverty vs. distance to the market
p.000018:
p.000018:
p.000018: cotton farm debt per ha ($) = 0.5 cotton farm debt per ha ($) = 1 cotton farm debt per ha ($) = 1.5
p.000018:
p.000018: distance = 5 km. distance = 10 km. distance = 15 km.
...
...
p.000018:
p.000018:
p.000018: 18
p.000019:
p.000019: Figure 10
p.000019: Determinants of the probability of MOVING INTO poverty
p.000019:
p.000019:
p.000019: Panel (A): Prob. of moving into poverty vs. % of adults in the household Panel (B): Prob. of moving
p.000019: into poverty vs. % of adults in admin. employment
p.000019:
p.000019:
p.000019: share of adults=25% share of adults=50% share of adults=75% share of adults=100%
p.000019:
p.000019: share of adults=25% share of adults=50% share of adults=75% share of adults=100%
...
...
p.000019:
p.000019:
p.000019:
p.000019:
p.000019:
p.000019: Panel (C): Prob. of moving into poverty vs. rain shock Panel (D): Prob. of
p.000019: moving into poverty vs. cotton & % of adults in agriculture
p.000019:
p.000019:
p.000019: less rain normal rain
p.000019:
p.000019: Cotton & Share = 50 %
...
...
p.000019:
p.000019:
p.000019: 19
p.000020:
p.000020: Figure 11
p.000020: The determinants of the probability of MOVING INTO poverty (cotton-producing districts only)
p.000020:
p.000020:
p.000020: Panel (A): Prob. of moving into poverty vs. region Panel (B):
p.000020: Prob. of moving into poverty vs. % of land under cotton
p.000020:
p.000020:
p.000020: region=Khatlon region=RRS
p.000020:
p.000020: % of land under cotton = 40
...
...
p.000020:
p.000020:
p.000020:
p.000020:
p.000020:
p.000020: Panel (C): Prob. of moving into poverty vs. distance to the market Panel (D): Prob. of
p.000020: moving into poverty vs. % of adults in agr. employme
p.000020:
p.000020:
p.000020: distance = 5 km. distance = 10 km. distance = 15 km.
p.000020:
p.000020: share of adults=25% share of adults=50% share of adults=75%
...
...
p.000020:
p.000020:
p.000020:
p.000020: 20
p.000021:
p.000021: significantly reduces the chances of moving out of poverty. Even when the district poverty headcount is
p.000021: controlled for, living in the RRS region is associated with more than 30 per cent lower chance of shedding poverty than
p.000021: in Khatlon (the reference region in the regression) (Appendix Table 2). This reflects the fact that between
p.000021: 1999 and 2003, RRS had the lowest rate of per capita GDP growth, averaging annually only 2 per cent while it was
p.000021: 14 per cent in Khatlon (GBAO and Sugd had comparable rates of growth). Location in neither a
p.000021: cotton-producing district or a cotton-producing community has no bearing on the odds of moving out of poverty.
p.000021: Controlling for other factors, cotton production in a district is found to have no statistically significant impact on
p.000021: household mobility out of poverty (Appendix Table 1). The same is true with respect to the impact of living in a
p.000021: cotton-producing community (Appendix Table 2).
p.000021:
p.000021: At the household level, household head’s schooling is related to a significantly higher probability of escaping
p.000021: poverty. The estimates suggest that the probability of shedding poverty increases from the 25 per cent that applies to
p.000021: the household head with less than secondary education to 50 per cent for those with university education
p.000021: (Panel B, Figure 8). Better health status also improves the odds of moving out of poverty: the
p.000021: probability of exiting poverty rises from about the 17 per cent observed for household heads with (self-reported)
p.000021: bad/very bad health to almost 40 per cent for those enjoying good/very good health (Panel C, Figure 8). Finally,
p.000021: a larger share of adults in wage employment has a positive impact on the poverty exit probability. This improves
p.000021: from 30 per cent to 50 per cent as the share of adults in hired employment goes up from 25 to 100 per cent (Panel D,
p.000021: Figure 8).
p.000021:
p.000021: The sample of rural households located in cotton-producing communities
p.000021: Using the sample of households located in cotton-producing districts only, we can explore the impact on
p.000021: poverty mobility of several variables related to the structural (and exogenous) changes that have taken place in
p.000021: the agricultural sector of Tajikistan. Regional factor has a substantial impact on poverty mobility in
p.000021: cotton-producing districts (similar to its impact for all rural households). In the cotton-producing districts of the
p.000021: country, the probability of moving out of poverty ranges from 10 per cent in RRS to 43 per cent Khatlon (Panel A,
p.000021: Figure 9). Larger initial fraction of land under private farming (dekhan) improves the odds: the chances of
p.000021: exiting poverty increase from 30 per cent when a tenth of the land is dekhans (the average level in 2000) to 70 per
p.000021: cent when the share of these farms increases to 30 per cent (Panel B, Figure 9). However, the rate of
p.000021: increase in the share of dekhan farming between 2000 and 2004 shows no association with the chances of shedding
p.000021: poverty. The estimated effect of this variable is not statistically significant (Appendix Table 3). The extent of the
p.000021: cotton farm debt (per hectare of land under cotton cultivation) has a strong impact on poverty mobility.
p.000021: It is estimated that if the debt were to double from US$0.5 to 1.00 per hectare, the probability of moving out of
p.000021: poverty would drop from 40 per cent to 10 (Panel C, Figure 9). Also, distance to market in cotton-producing areas
p.000021: affects poverty mobility. A somewhat counterintuitive finding is the observation that greater distances to market or
p.000021: the district centre improve the odds of moving out of poverty (Panel D, Figure 9). However, one needs to bear in
p.000021: mind that several earlier studies on Tajikistan (World Bank 2006) indicate a high degree of government control
p.000021: and regulation of the cotton market. It may well be that our finding indicates that being in the proximity
p.000021: of the ‘watchful eye of the state’ does not promote the sharing of benefits from cotton production. It
p.000021: would be useful to explore this finding further in future research.
p.000021:
p.000021:
p.000021:
p.000021: 21
p.000022:
p.000022: 4.3 The main predictors of moving into poverty
p.000022:
p.000022: The sample of all rural households
p.000022: The factors that explain a household’s likelihood to fall into poverty are different from those that explain moving out
p.000022: of poverty (Appendix Tables 1 and 2). The probability of moving into poverty declines significantly once the share of
p.000022: working-age individuals in a household increases. The estimates suggest that the probability of falling into poverty
p.000022: declines from 55 to 10 per cent when the share of adults increases from 25 to 100 per cent (Panel A, Figure 10).
p.000022: Employment in public administration reduces the risk of impoverishment: the probability of falling into poverty
p.000022: drops from 20 to 3 per cent as the share of adults in administrative employment increases from 25 to 100
p.000022: per cent (Panel B, Figure 10). Household size is also a factor: the estimates suggest a U-shaped relationship between
p.000022: the probability of becoming poor and household size (Appendix Table 2). In other words, small households
p.000022: (elderly people living alone) and very large households (usually households with many children) face a higher risk of
p.000022: falling into poverty than the average-sized household.
p.000022:
p.000022: Examining the impact of district/community level characteristics, we find that variations in rain fall are
p.000022: associated with the risk of becoming poor. Households located in communities with less than average amount of
p.000022: rain over the previous year face a 55 per cent chance of becoming poor versus 28 per cent for households in
p.000022: non-drought communities (Panel C, Figure 10). Agricultural employment in cotton producing areas is associated with a
p.000022: higher risk of impoverishment compared to similar employment in non-cotton producing areas.20 According to estimates,
p.000022: if half of the adults in a cotton- producing district work in the agricultural sector, the risk of poverty
p.000022: in the cotton- producing areas is 55 per cent, but only 25 per cent in non-cotton areas. However,
p.000022: increasing the share of household members employed in agriculture in cotton-producing areas improves the odds of not
p.000022: falling into poverty (Panel D, Figure 10).
p.000022:
p.000022: The sample of rural households located in cotton-producing communities
p.000022: Using the sample of only those households that are located in cotton-producing districts, we note the following major
p.000022: findings. Region of residence has a substantial impact on poverty mobility in cotton-producing districts (similar to
p.000022: its impact on the probability of moving out of poverty). Households in the RRS region face a 65 per cent probability of
p.000022: falling into poverty, while this is 20 per cent for those living in Khatlon (Panel A, Figure 11). A
p.000022: higher share of land under cotton cultivation in the district increases the risk of falling into poverty: once the
p.000022: share of cotton-cultivated land increases from 40 to 60 per cent, this compounds the odds from 20 to 40 per
p.000022: cent (Panel B, Figure 11). Distance to market is also a factor: the estimates suggest that as distance increases,
p.000022: the probability of becoming poor generally declines (Panel C, Figure 11). The explanation here is likely to be the same
p.000022: as discussed above in the case of moving out of poverty. A larger proportion of adults in agricultural employment
p.000022: reduces the probability of falling into poverty. But the effect is very marginal. As the share of
p.000022: agricultural employers increases from 25 to 75 per cent, the odds of becoming poor decline from 22 to 16 per cent
p.000022: (Panel D, Figure 11).
p.000022:
p.000022:
p.000022: 20 The household survey data do not specify how many adults are actually employed in the cotton farm (no cotton is
...
...
p.000023: in the agricultural reform agenda, including a rapidly changing structure of land ownership as the old
p.000023: soviet-era farms were dismantled and private-owned dekhan farms created. During this period there was a noticeable
p.000023: increase in crop yields, including cotton, the major agricultural commodity in Tajikistan. However, despite improved
p.000023: cotton yields, output value dropped because of declining international prices for cotton. Moreover, cotton
p.000023: farms accumulated substantial debts to creditors. This period of rapid changes makes the analysis of the
p.000023: process of poverty mobility among rural households very interesting.
p.000023:
p.000023: This paper uses the panel component of two surveys conducted in Tajikistan at an interval of one year to
p.000023: explore the major determinants of the transition of households out of (or into) poverty. Household poverty status is
p.000023: measured in the asset space which, compared to a welfare measure based on consumption, provides a better indication of
p.000023: structural poverty transition. In addition to analysing the determinants of poverty transitions at the
p.000023: household level, we also look at the correlates of poverty at the district (rayon) level. The findings have
p.000023: important implications, which are briefly discussed below.
p.000023:
p.000023: First, several household-level factors emerge as key predictors of poverty transition, suggesting the
p.000023: importance of continued investments to improve human capital outcomes. The odds for exiting poverty
p.000023: increase with the higher level of education and improved health status of the household head, as well as with the
p.000023: higher ratio of adults in wage employment. The risk of falling into poverty declines with a higher share of working-age
p.000023: people in the household and a larger share of adults working in public administration.
p.000023:
p.000023: The district-level data suggest that areas where cotton farming has a more prominent role are likely to
p.000023: have higher levels of poverty. The analysis of poverty mobility at the household level also indicates that households
p.000023: located in cotton-producing areas do not enjoy better odds of climbing out of poverty. The analysis actually reveals
p.000023: that having a higher share of land under cotton cultivation in the district increases the probability of falling into
p.000023: poverty. Moreover, there are indications that that living in a cotton- producing area located near to
p.000023: markets (or district centre) worsens the chances of escaping poverty. Furthermore, the accumulated debt of the
p.000023: cotton farms is estimated to present a substantial drawback in transiting out of poverty. These
p.000023: findings are disheartening, given the importance of cotton in Tajikistan’s agricultural production, and the
p.000023: number of people employed in the sector. A new critical look at the cotton sector is needed by
p.000023: policymakers in order to understand why cotton production does not broadly benefit the population of the
p.000023: cotton-producing areas.
p.000023:
p.000023: The rate of increase in the share of dekhan (private) farming in a district had little impact on poverty
p.000023: levels and poverty mobility. Examination of the poverty correlates at the district level indicates that lower poverty
p.000023: levels are associated with larger portions of arable land being transferred to dekhan (private) farms. However, the
p.000023: increase in this type of farming between 2000-04 showed no positive impact on poverty levels or poverty
p.000023: mobility. The analysis at the household level indicates that larger initial shares of private farming improve the
p.000023: odds of escaping poverty. Nevertheless, the rate of
p.000023:
p.000023:
p.000023: 23
p.000024:
p.000024: increase in this type of farming has not yet improved the chances of mobility out of poverty. This is
p.000024: likely a reflection of the fact that land ownership transfers are often on paper only, and thus are not accompanied by
p.000024: improvements in farm productivity.
p.000024:
p.000024: There is strong evidence of geographic poverty mobility traps. A higher level of poverty in a district significantly
p.000024: reduces the chances of a household of moving out of poverty. Living in a region with an overall slow economic
p.000024: growth rate is also found to undermine the odds of escaping poverty and increase the odds of falling into poverty.
p.000024: The risk of impoverishment significantly increases for households in regions that experienced drought. In
p.000024: other words, everything else being equal, the geographical location of a household matters considerably in
p.000024: terms of its chances of escaping or falling into poverty. It is worth noting that this observation
p.000024: regarding geographical poverty traps on the part of rural households in Tajikistan confirms numerous similar findings
p.000024: in other countries and settings, as in post-reform China (Jalan and Ravallion 2002).
p.000024:
p.000024:
p.000024: References
p.000024:
p.000024: Angel-Urdinola, D. F., C. Mete, and S. R. Cnobloch (2008). ‘Poverty Dynamics in Tajikistan’. Washington,
p.000024: DC: World Bank. Manuscript.
p.000024: Baschieri, A., and J. Falkingham (2005). ‘Developing a Poverty Map of Tajikistan: A Technical Note’. S3RI
p.000024: Applications and Policy Working Papers, A05/11. Southampton: Southampton Statistical Sciences Research
p.000024: Institute.
p.000024: Becerra, C. A. V. (2004). ‘The World Cotton Market: A Long-Term Outlook’. Working Paper presented at the WTO African
p.000024: Regional Workshop on Cotton, 23-24 March, Cotonou.
p.000024: Canto, O. (2002). ‘Climbing Out of Poverty, Falling Back in: Low Income Stability in Spain’. Applied Econometrics, 34
p.000024: (15): 1903-16.
p.000024: Carter, M. R., and J. May (2001). ‘Poverty, Livelihood and Class in Rural South Africa’. World
p.000024: Development, 27 (1): 1-20.
p.000024: Carter, M. R., and B. C. Barrett (2006). ‘The Economics of Poverty Traps and Persistent Poverty: An Asset-Based
p.000024: Approach’. Journal of Development Studies, 42 (2): 178- 99.
p.000024: Filmer, D., and L. Pritchett (1998). ‘Estimating Wealth Effects without Income or Expenditure Data’. WB
p.000024: Policy Research Paper 1994. Washington, DC: World Bank.
p.000024: Jalan, J., and M. Ravallion (2002). ‘Geographic Poverty Traps? A Micro Model of Consumption Growth in Rural
p.000024: China’. Journal of Applied Econometrics, 17: 329-46.
p.000024: Lawley, D., and A. Maxwell (1971). Factor Analysis as a Statistical Method. London: Butterworth & Co.
p.000024: Sahn, D., and D. Stifel (2000). ‘Poverty Comparisons over Time and across Countries in Africa’. World Development, 28
p.000024: (1): 2123-55.
p.000024:
p.000024:
p.000024:
p.000024:
...
...
p.000025:
p.000025: Sahn, D., and D. Stifel (2003). ‘Exploring Alternative Measures of Wealth in the Absence of Expenditure
p.000025: Data’. The Review of Income and Wealth, 49 (4): 463-89.
p.000025: World Bank (2005). ‘Priorities for Sustainable Growth: A Strategy for Agriculture Sector Development in
p.000025: Tajikistan’. Washington, DC: World Bank. Mimeo
p.000025: World Bank (2006). ‘Republic of Tajikistan Poverty Assessment Update’. Report No.
p.000025: 30853-TJ. Washington, DC: World Bank.
p.000025: Ukaeva, U. (2005). ‘Decomposition of Agricultural Growth in Tajikistan’. Washington, DC: World Bank. Mimeo.
p.000025:
p.000025:
p.000025:
...
...
p.000025:
p.000025: 25
p.000026:
p.000026:
p.000026: Appendix Table 1
p.000026: Determinants of the transition out of, and into, poverty for rural households (probit model) (with no community
p.000026: characteristics)
p.000026:
p.000026:
p.000026:
p.000026:
...
...
p.000026:
p.000026: Source: Authors’ estimates, based on 2003 LSMS and 2004 EHS.
p.000026:
p.000026:
p.000026: Appendix Table 2
p.000026: Determinants of the transition out of, and into, poverty for rural households (probit model) (with community
p.000026: characteristics and ‘cotton’ variable interactions)
p.000026:
p.000026:
p.000026:
p.000026:
...
...
p.000026:
p.000026: Source: Authors’ estimates, based on 2003 LSMS and 2004 EHS.
p.000026:
p.000026:
p.000026: Appendix Table 3
p.000026: Determinants of the transition out of, and into, poverty for rural households in cotton-producing districts (probit
p.000026: model) (with community characteristics and ‘cotton variable interactions)
p.000026:
p.000026:
p.000026:
p.000026:
...
Searching for tag risk:
(return to top)
...
p.000001: ISSN 1810-2611 ISBN 978-92-9230-072-2
p.000001:
p.000001: Third, there is strong evidence of geographic poverty mobility traps in Tajikistan. Higher levels of
p.000001: poverty in a district appear to reduce significantly the chance of a household shedding poverty. Living in a
p.000001: region with overall slow economic growth is also found to undermine the odds of exiting poverty and to increase the
p.000001: risk of falling into poverty. Finally, several key household-level factors, such as the share of adults, education
p.000001: level, health status and participation in wage employment, also emerge as significant predictors of poverty
p.000001: mobility.
p.000001:
p.000001:
p.000001: Authors notes
...
...
p.000015:
p.000015: 3.4 The empirical model
p.000015:
p.000015: The event of a household transiting out of (or into) (asset-based) poverty is modelled within the probability
p.000015: framework. The (ex-post) realization of the event (experience of the transit into (out of) poverty between 2003-04) is
p.000015: used to define the samples at risk of leaving (falling into) poverty. The probability of experiencing a transit out of
p.000015: (or into) poverty is modelled as follows:
p.000015:
p.000015:
p.000015: Pr( 0
p.000000: 0
...
...
p.000022: The sample of all rural households
p.000022: The factors that explain a household’s likelihood to fall into poverty are different from those that explain moving out
p.000022: of poverty (Appendix Tables 1 and 2). The probability of moving into poverty declines significantly once the share of
p.000022: working-age individuals in a household increases. The estimates suggest that the probability of falling into poverty
p.000022: declines from 55 to 10 per cent when the share of adults increases from 25 to 100 per cent (Panel A, Figure 10).
p.000022: Employment in public administration reduces the risk of impoverishment: the probability of falling into poverty
p.000022: drops from 20 to 3 per cent as the share of adults in administrative employment increases from 25 to 100
p.000022: per cent (Panel B, Figure 10). Household size is also a factor: the estimates suggest a U-shaped relationship between
p.000022: the probability of becoming poor and household size (Appendix Table 2). In other words, small households
p.000022: (elderly people living alone) and very large households (usually households with many children) face a higher risk of
p.000022: falling into poverty than the average-sized household.
p.000022:
p.000022: Examining the impact of district/community level characteristics, we find that variations in rain fall are
p.000022: associated with the risk of becoming poor. Households located in communities with less than average amount of
p.000022: rain over the previous year face a 55 per cent chance of becoming poor versus 28 per cent for households in
p.000022: non-drought communities (Panel C, Figure 10). Agricultural employment in cotton producing areas is associated with a
p.000022: higher risk of impoverishment compared to similar employment in non-cotton producing areas.20 According to estimates,
p.000022: if half of the adults in a cotton- producing district work in the agricultural sector, the risk of poverty
p.000022: in the cotton- producing areas is 55 per cent, but only 25 per cent in non-cotton areas. However,
p.000022: increasing the share of household members employed in agriculture in cotton-producing areas improves the odds of not
p.000022: falling into poverty (Panel D, Figure 10).
p.000022:
p.000022: The sample of rural households located in cotton-producing communities
p.000022: Using the sample of only those households that are located in cotton-producing districts, we note the following major
p.000022: findings. Region of residence has a substantial impact on poverty mobility in cotton-producing districts (similar to
p.000022: its impact on the probability of moving out of poverty). Households in the RRS region face a 65 per cent probability of
p.000022: falling into poverty, while this is 20 per cent for those living in Khatlon (Panel A, Figure 11). A
p.000022: higher share of land under cotton cultivation in the district increases the risk of falling into poverty: once the
p.000022: share of cotton-cultivated land increases from 40 to 60 per cent, this compounds the odds from 20 to 40 per
p.000022: cent (Panel B, Figure 11). Distance to market is also a factor: the estimates suggest that as distance increases,
p.000022: the probability of becoming poor generally declines (Panel C, Figure 11). The explanation here is likely to be the same
p.000022: as discussed above in the case of moving out of poverty. A larger proportion of adults in agricultural employment
p.000022: reduces the probability of falling into poverty. But the effect is very marginal. As the share of
...
...
p.000023: important implications, which are briefly discussed below.
p.000023:
p.000023: First, several household-level factors emerge as key predictors of poverty transition, suggesting the
p.000023: importance of continued investments to improve human capital outcomes. The odds for exiting poverty
p.000023: increase with the higher level of education and improved health status of the household head, as well as with the
p.000023: higher ratio of adults in wage employment. The risk of falling into poverty declines with a higher share of working-age
p.000023: people in the household and a larger share of adults working in public administration.
p.000023:
p.000023: The district-level data suggest that areas where cotton farming has a more prominent role are likely to
p.000023: have higher levels of poverty. The analysis of poverty mobility at the household level also indicates that households
p.000023: located in cotton-producing areas do not enjoy better odds of climbing out of poverty. The analysis actually reveals
...
...
p.000024: improvements in farm productivity.
p.000024:
p.000024: There is strong evidence of geographic poverty mobility traps. A higher level of poverty in a district significantly
p.000024: reduces the chances of a household of moving out of poverty. Living in a region with an overall slow economic
p.000024: growth rate is also found to undermine the odds of escaping poverty and increase the odds of falling into poverty.
p.000024: The risk of impoverishment significantly increases for households in regions that experienced drought. In
p.000024: other words, everything else being equal, the geographical location of a household matters considerably in
p.000024: terms of its chances of escaping or falling into poverty. It is worth noting that this observation
p.000024: regarding geographical poverty traps on the part of rural households in Tajikistan confirms numerous similar findings
p.000024: in other countries and settings, as in post-reform China (Jalan and Ravallion 2002).
p.000024:
...
Searching for tag social:
(return to top)
...
p.000001: The World Institute for Development Economics Research (WIDER) was established by the United Nations University
p.000001: (UNU) as its first research and training centre and started work in Helsinki, Finland in 1985. The Institute
p.000001: undertakes applied research and policy analysis on structural changes affecting the developing and
p.000001: transitional economies, provides a forum for the advocacy of policies leading to robust, equitable and
p.000001: environmentally sustainable growth, and promotes capacity strengthening and training in the field of economic and
p.000001: social policy making. Work is carried out by staff researchers and visiting scholars in Helsinki
p.000001: and through networks of collaborating scholars and institutions around the world.
p.000001:
p.000001: www.wider.unu.edu publications@wider.unu.edu
p.000001:
p.000001:
...
Searching for tag special:
(return to top)
...
p.000001: population live in the countryside, poverty in Tajikistan continues to be an overwhelmingly rural
p.000001: phenomenon. Economic growth and the resultant poverty reduction are explained by three major factors: (i) conflict
p.000001: cessation, which allowed economic activity to resume and markets to develop; (ii) initial impact of the macroeconomic
p.000001: stability and agricultural reforms in the non-cotton sector that enabled farmers to diversify production
p.000001: and increase productivity; and (iii) large increase in migrant workers exiting Tajikistan for Russia and other
p.000001: countries. However, there have been concerns that once the initial benefits of these ‘special’ factors dry out,
p.000001: Tajikistan’s poverty reduction trends may not be sustainable (World Bank 2006).
p.000001:
p.000001: In view of the sound economic growth rates, markedly reduced but still very high rural poverty, and concerns over the
p.000001: sustainability of the country’s poverty reduction trend, it is important from a policy perspective to understand
p.000001: the key factors at the micro (household/community) level that explain the transition of rural households in and
...
...
p.000008: falls below the consumption poverty line u* and its stock of
p.000008:
p.000008:
p.000008:
p.000008: 9 The graph is not presented here, but available from the authors on request.
p.000008: 10 See February 2006 special issue of the Journal of Development Studies, 42 (2) for the set of papers presenting the
p.000008: conceptual framework and empirical evidence for an asset-based approach.
p.000008:
p.000008:
p.000008: 8
p.000009:
...
Searching for tag welfare:
(return to top)
...
p.000001: movements. The empirical analysis reveals several interesting findings that are also important from a
p.000001: policy perspective: first, cotton farming seems to have no positive impact on poverty levels, nor on mobility out
p.000001: of poverty. Second, the rate of increase in the share of private farming at the district level had
p.000001: little impact on poverty levels and poverty mobility.
p.000001: …/.
p.000001: Keywords: welfare, poverty, Tajikistan JEL classification: I3
p.000001:
p.000001: Copyright © UNU-WIDER 2008
p.000001: * ECSHD, World Bank, Washington, DC: emails: oivaschenko@worldbank.org (O. Ivaschenko);
p.000001: cmete@worldbank.org (C. Mete)
p.000001: This is a revised version of a paper originally prepared for the UNU-WIDER conference on Fragile
...
...
p.000001: In view of the sound economic growth rates, markedly reduced but still very high rural poverty, and concerns over the
p.000001: sustainability of the country’s poverty reduction trend, it is important from a policy perspective to understand
p.000001: the key factors at the micro (household/community) level that explain the transition of rural households in and
p.000001: out of poverty. This is the main objective of this paper.
p.000001:
p.000001: The paper contributes to the literature on welfare dynamics in general, and to the studies of poverty in Tajikistan in
p.000001: particular on several fronts:
p.000001:
p.000001: i) utilizing an assets-based approach to better capture the permanent (as opposed to transitory) component of
p.000001: welfare changes for rural households;
p.000001: ii) investigating explicitly the importance of community/local factors versus household/individual level
p.000001: characteristics to explain movements in and out of (asset-based) poverty.
p.000001: It is worth noting that a study of the general factors affecting poverty transition in Tajikistan has
p.000001: been undertaken by Angel-Urdinola, Mete and Cnobloch (2008). Our
p.000001:
...
...
p.000001: country in the Europe and Central Asia region.
p.000001:
p.000001:
p.000001: 1
p.000002:
p.000002: study expands on this work by focusing specifically at the determinants of welfare dynamics in rural
p.000002: areas. More specifically, we combine household survey data with district (rayon) level data to take into
p.000002: account the poverty impact of community-level factors such as the share of private (dekhan) farms, per
p.000002: hectare of land under cotton cultivation, level of debt and distance to market or district centre.3
p.000002:
p.000002: The paper is structured as follows. Section 2 discusses the main developments in Tajikistan’s rural
...
...
p.000002: contribution to sector output from 10 to 24 per cent. The share of
p.000002:
p.000002:
p.000002: 3 Angel-Urdinola, Mete and Cnobloch (2008) differentiate between urban and rural areas by including an urban/rural
p.000002: dummy in their regression that uses combined (urban and rural) panel sample, and thus they make no attempt to analyse
p.000002: welfare dynamics determinants that would be specific to the rural areas.
p.000002: 4 State-farms encompass ownership forms of both the sovhoz (soviet farms) and kolhoz (collective farms),
p.000002: which are effectively the same.
p.000002: 5 A thorough overview of agricultural reforms in Tajikistan is provided in World Bank (2005). The
p.000002: major findings of this study are: (i) the process of land restructuring has been rather inequitable;
p.000002: (ii) the reform of state-farms in especially cotton-producing areas has resulted in numerous
...
...
p.000008:
p.000008: The literature on poverty dynamics has increasingly recognized the importance of adopting an asset-based
p.000008: approach to study changes in wellbeing, especially in response to a wide range of different (climatic,
p.000008: health, political and other) shocks.10 Differentiating between stochastic and structural poverty
p.000008: transitions implies the availability of information on assets and expected levels of wellbeing. To illustrate the
p.000008: importance of the assets-based approach in capturing welfare-status changes, we use the conceptual framework
p.000008: advocated by Carter and May (2001) and Carter and Barrett (2006). This framework is presented in Figure 6.
p.000008:
p.000008: In essence, in any timeperiod a household can be regarded as structurally poor if household consumption
p.000008: falls below the consumption poverty line u* and its stock of
p.000008:
...
...
p.000010: the variables. Denoted by Aij the observation for household i and asset j (for example, whether or not a household
p.000010: has a
p.000010:
p.000010:
p.000010: 12 The majority of households in the panel are urban because the 2004 EHS over-sampled these areas.
p.000010: 13 For a comprehensive discussion of the pros and cons of an assets-based welfare analysis, see Filmer and Pritchett
p.000010: (1998), Sahn and Stifel (2000), and Sahn and Stifel (2003). An asset index retains certain properties necessary for
p.000010: proper welfare analysis, such as transparency in construction and ranking individuals credibly in terms of
p.000010: welfare. As argued by Filmer and Pritchett (1998), Carter and May (2001) and Carter and Barrett (2006), an asset index
p.000010: is likely to be a better indicator of the long-run household wealth than per capita household consumption.
p.000010: However, a significant limitation of an assets index is that it treats ownership of assets as giving
p.000010: similar utility without allowing for differences in unobserved quality.
p.000010:
p.000010:
...
...
p.000012:
p.000012: Asset indexes typically also include housing characteristics, such as the type of floor and walls of the
p.000012: dwelling. However, given that the analysis is based on a one-year panel, housing variables remained largely
p.000012: unchanged and were thus excluded. We also excluded variables related to the ownership of agricultural assets and
p.000012: livestock. If the accumulation of assets in the rural areas takes place largely through the acquisition of agricultural
p.000012: assets, the omission of these variables from the asset index is likely to underestimate welfare changes.
p.000012: However, as discussed in greater detail later, according to the data, significant changes in rural households concern
p.000012: the possession of durable goods (which make up the asset index).14 In fact, rural households accumulated major durable
p.000012: goods faster than urban households (Table 1).
p.000012:
p.000012: Moreover, as our attempt is to understand the determinants of rural households in moving in/out of
...
...
p.000012: between 2003-04, particularly in the rural regions. The average number of wood-burning stoves and black and white
p.000012: TVs declined during the same period. There was also a noticeable decline in the number of electric radiators,
p.000012: presumably because of rising electricity costs. The last column in Table 1 presents the asset scoring
p.000012: coefficient (from the factor analysis) for the major assets owned by households. The scoring coefficient
p.000012: effectively indicates the weight of a specific variable in estimating the total asset score. A positive
p.000012: coefficient suggests a positive association between having the particular asset and the overall welfare index. A
p.000012: higher value of the scoring coefficient suggests a stronger association. Note that most of the assets for which
p.000012: possession increased between 2003-04 display a positive scoring coefficient with magnitudes between 0.2 and 0.3.
p.000012: Finally, ownership of assets such as the wood stoves, usually associated with lower welfare, decreased between
p.000012: 2003-04. The scoring coefficient for wood stoves is close to zero, suggesting a rather flat association
p.000012: between having the asset and household welfare. The scoring coefficient for black and white TVs is even negative.
p.000012: It is important to note that the scoring coefficients are estimated with the pooled 2003 and 2004 samples
p.000012: (panel), which make the estimated asset indices fully comparable between the two years.
p.000012:
p.000012: It is worth noting that there was a substantial mobility in the ownership of various assets, with households acquiring
p.000012: and disposing of assets. For instance, about 20 per cent of rural households overall acquired a colour TV (Panel A,
...
...
p.000014: ownership of assets by rural households in Tajikistan has been far from a static process, even in a period as short as
p.000014: one year.
p.000014:
p.000014: We set up an asset-based poverty line at a level equivalent to the 50th percentile of the asset index distribution
p.000014: (using 2003 distribution). The chosen cut-off level is rather arbitrary, but is consistent with the fact
p.000014: that over half of the population—based on a welfare indicator of per capita consumption and the poverty line of
p.000014: US$2.25 per day—is estimated to be poor. The 50th percentile cut-off level was also used in a previous study of poverty
p.000014: dynamics in urban and rural Tajikistan (Angel-Urdinola, Mete and Cnobloch 2008).16
p.000014:
p.000014: Given this asset-based poverty line, what was the extent of the poverty mobility among rural households in the panel?
p.000014: The poverty mobility matrix is presented in Table 2. Out of 322 households qualifying as asset-poor in 2003 (base year)
...
...
p.000014: 15 It is worth noting that in many cases while a new asset is purchased, the old one will be just discarded unless it
p.000014: has tradable value on the market. When a household replaces an old TV, its asset position does not change, but if
p.000014: merely gets rid of the old set without buying a new one, its wealth position deteriorates.
p.000014: 16 In the study by Angel-Urdinola, Mete and Cnobloch (2008), the authors have an urban/rural dummy in their panel
p.000014: sample that includes households living in both urban and rural areas. They do not attempt to investigate
p.000014: the determinants of welfare dynamics that would be specific to rural areas.
p.000014:
p.000014:
p.000014: 14
p.000015:
p.000015: 3.4 The empirical model
...
...
p.000023: farms accumulated substantial debts to creditors. This period of rapid changes makes the analysis of the
p.000023: process of poverty mobility among rural households very interesting.
p.000023:
p.000023: This paper uses the panel component of two surveys conducted in Tajikistan at an interval of one year to
p.000023: explore the major determinants of the transition of households out of (or into) poverty. Household poverty status is
p.000023: measured in the asset space which, compared to a welfare measure based on consumption, provides a better indication of
p.000023: structural poverty transition. In addition to analysing the determinants of poverty transitions at the
p.000023: household level, we also look at the correlates of poverty at the district (rayon) level. The findings have
p.000023: important implications, which are briefly discussed below.
p.000023:
p.000023: First, several household-level factors emerge as key predictors of poverty transition, suggesting the
...