Quantitative and Qualitative Approach to Understand Poverty and Inequality Dynamics in Uganda 1

Transcription

1 Quantitative and Qualitative Approach to Understand Poverty and Inequality Dynamics in Uganda 1 Collection of Life History Data based on 1992/99 Household Panel David Lawson April This research was undertaken with support from The Faculty of Humanities, General Research Support Fund, University of Manchester, and Department of International Development (DFID) Uganda. I gratefully acknowledge the assistance of David Hulme, James Muwange, Anthony Matovu and Vincent Ssennono, and comments on earlier drafts from Arthur van Diesen. 1

2 EXECUTIVE SUMMARY Despite an increased focus in combining qualitative and quantitative ( Q2 ) research to further our understanding of poverty, there are still relatively few LDC based attempts that have successfully achieved this. By using the sampling frame of Uganda s 1992/99 (2 wave) quantitative household panel data we follow up a small number of households to obtain life history information, with the objective being to further explain some of the Propagators, Maintainers, and Interrupters of Poverty, and Inequality, dynamics. In this case, and primarily due to budgetary constrains, we restrict our focus to three main issues of Ill health/hiv/aids, Assets and Gender, but also comment on broader findings. In particular, we find: Suggestions of direct causality between Ill health/hiv/aids and movements into poverty, and explanations regarding the processes that underpin this. Households preference order the types of assets sold in times of crisis (e.g. luxury goods such as radios are commonly sold first). However, the willingness, order and extent to which assets are sold varies according to age, geographical location and socio-economic characteristics of the household members. Households commonly asset smooth i.e. if chronic or transiently poor households do not have immediately available resources to pay for medical bills/transport etc. as a first resort food consumption was virtually always reduced, before selling assets or obtaining loans. Many individuals, and particularly women, commented that the lack of effective social networks and limited contacts in the job market, was a probable cause of gender inequality, especially with regards to formal labour market participation. Increased (and larger) household sizes were commonly identified as a major cause of persistent and transient poverty. In particular, a very common finding was for households to experience increased wealth, for household sizes then to increase, but only 5-6 years later do the delayed child costs of school fees etc. start causing the major monetary problems that impact on a households ability to avoid poverty. The Q2 approach was particularly useful in identifying household vulnerability. Limited access to land, and cases of disputed land, were particularly prevalent amongst poorer households. This has deteriorated the welfare status and long term security and planning of such households. From a policy perspective, preliminary micro level findings indicate that there could be a limiting of agricultural productivity, due to lack of access/ability to plan long term. Also, given the impact of illness, death and increasing pressures from having no land access, such findings perhaps point to the importance of social protection as having key role to avoid persistent poverty and inequality. From a methodological perspective, the use of Q2 approaches clearly assisted in providing clarity of understanding regarding the processes that underpin poverty and inequality dynamics. Even with a small life history sample size, and based on relatively old panel data we were able to complement the quantitative data with life history, and further data collection, to reveal some of the real processes that underpin how health, assets and gender interconnect with poverty and inequality dynamics. 2

3 1. Introduction/Background to The Study Over the last few years there has been an increased focus on genuinely combining qualitative and quantitative ( Q2 ) research methodologies to further our understanding of poverty. Despite this however, there are still relatively few LDC based attempts that have successfully achieved this. Most applied papers focus on the use of participatory techniques and simply cross-reference to separate quantitative findings (e.g. Barahona et al for Malawi, Lawson et al for Uganda). 2 While in some respects such work is breakthrough, in the main it has made only a limited contribution to deepening analysis of poverty because they rarely closely integrate the sampling frame, they use group based methods in public (participatory appraisal) and the analytical methods applied to data collected by participatory appraisal remain opaque. 3 Considering the above, for this research we integrate qualitative and quantitative research, using Ugandan panel data, visiting the same households to obtain life history information. A number of direct findings are provided in relation to some accepted areas highlighted in the PMAU Research agenda. For example, issues covered under several of the cross cutting themes mentioned in Research Programme to Support the Eradication of Poverty in Uganda are directly investigated. In particular we limit our investigation to the main issues highlighted by households as being particularly important in relation to poverty dynamics. These include issues of gender inequalities and empowerment, HIV/AIDS/Chronic Illness, and Assets. The report is structured as follows. The next section reviews the main attempts that have been made at combining Q2 research within Uganda. We then provide a detailed outline of the approach used for this paper. Section 4 then provides the main empirical analysis. Disaggregating the section by the 3 main themes, we undertake a focused analysis of the 1992/99 quantitative data, outlining the main findings, and then further these with the information obtained from the life history interviews. The final section of the paper summarise the main findings. 2 See further Q2 literature such as Adoto, M, F. Lund and P Mhlongo (2004), Parker B.,and V. Kozel (2004), Barahona and Levy (2004), all Q2 conference papers: 3 The use of group based methods in public (participatory appraisal) means that the quality of data is often compromised and there are profoundly ethical issues (e.g. by classifying named individuals as poor in public, reinforces social stigma). A better, and more complementary methods to quantitative methods, is a life history research approach. The analytical methods applied to data collected by participatory appraisal remain opaque. While some analysts claim to present the Voices of the Poor this is clearly untrue as the materials presented are selective and structured, some form of interpretive lens has been utilized but is not explained 3

4 2. Previous Q2 Work In Uganda Poverty research in Uganda is very well developed with both of Uganda s Participatory Poverty Assessment Process (UPPAP s) providing widely quoted qualitative findings and a series of quantitative papers having been produced by the likes of Deininger and Okidi (2003), and Appleton (1999). However, there is relatively limited genuinely integrated poverty research for the country. McGee (2000) and Appleton and Booth (2001) were perhaps the first to note the advantages of combining research methods to further our understanding of poverty/poverty dynamics. From a methodological perspective, McGee (2000) focused on how the nationally representative quantitative household surveys (the Uganda National Household Survey - UNHS) and the Participatory Poverty Assessment (PPA) could be combined. She concluded that the PPA could be used to: Further examine the downturn in welfare indicators between 1995/6 and 1996/7 as derived by the UNHS; Provide insights to refine the UNHS questionnaires e.g. including questions on risk and vulnerability; Visit some of the same households for both the PPA and UNHS; Try and standardise poverty trends assessment in PPA and repeat the PPA for subsequent waves. (Kanbur 2001a: 19). Further to the above, Appleton and Booth (2001) highlighted the need to take advantage of the panel data in furthering our understanding of poverty and other issues. Of the research that has applied the combined methods, Lawson et al. (2003) combined qualitative and quantitative research to try and understand the factors and processes underlying poverty transitions and persistence. They used the nationally based UPPAP findings, and two wave panel data, of 1992/99, to undertake econometric analysis. The research showed that the qualitative sources added substantially to the information available from the panel survey data alone, by helping to identify key issues to investigate using the survey data and by providing important additional insights not available from the survey data, including about processes and contextual factors. Despite not inter-linking the sampling frame/households interviewed, of particular note they found: Factors such as the lack of key physical assets, high dependency ratios and increased household size were identified by both the qualitative and quantitative approaches, as major factors influencing poverty transitions and persistence. In other instances though the qualitative approach identifies additional factors not so easily identified quantitatively - for example the impacts of excessive alcohol consumption in many cases. A more notable attempt at genuine Q2 research for rural Uganda, is that undertaken by Bird and Shinyekwa (2003). They build on the qualitative (participatory) and quantitative livelihoods research undertaken in 3 Districts and undertake in-depth life history interviews with the heads of nearly 25 households in 3 villages. The choice to talk to someone about their life, rather than adopting a thematic approach, meant they 4

5 were able to identify the most common covariant and idiosyncratic shocks which triggered a decline into poverty (drivers) and the constraints which prevented accumulation, investment and movement out of poverty (maintainers) (Bird and Shinyekwa, 2003:10). As with UPPAP, Bird and Shinyekwa found: Poor gender relations and excessive alcohol consumption to be key factors, with the latter often financed by the sale of household assets and a major cause of domestic violence. However, a common finding was that the poorest had suffered recurrent and composite shocks and personal tragedies compared with the persistently nonpoor who had simply managed to avoid personal disaster, allowing them to retain their assets and even continue to accumulate (Bird and Shinyekwa, 2003:31). More recently Ellis et al. (2006) completed a ladder study for the three districts of Mbale, Kamuli and Mubende Panel households were interviewed and the information complemented with community qualitative interviews. They found reasons for downward trajectories in welfare status to be: Farm sub-division, chronic illness, death of household head, livestock disease, theft, and profligacy (spending money on gambling and drink and thence getting into debt) (Ellis et al. 2006:14). Even more recent attempts at Q2 attempts have been made by The Government of Uganda, with support from The World Bank, to understand the dynamics of upward mobility from poverty. The idea was to follow up households from the 1992/99 panel that had moved out of poverty. Results are yet to be finalised, but despite few panel households having been located, the work is likely to produce value added for the quantitative/qualitative survey design of future research. 4 Livelihoods and Diversification Directions Explored by Research (LADDER). 5

6 3. This Approach Quantitative Data Specifically, life history interviews were carried out on the households that were in a nationally representative sample of 1,103 households that had been interviewed for quantitative data in both 1992 and 1999 (a two wave panel). 5 Based on the country s accepted monetary based poverty measures, a sample of households that are Chronically Poor (Poor in both periods), Never Poor, Moving Into Poverty, and Moved Out Of Poverty were selected (see Appendix Table 1 for a summary of the national figures of households defined as chronically poor, moved into poverty, moved out of poverty, never poor). The sampling of households to be selected for further interview was based on the proportions in the panel, with more than 40 households interviewed (comprising roughly equally proportions of households, across several districts and 3 regions, that were persistently poor, moved into poverty, moved out of poverty and have never been poor). For logistical, purposes and due to a restrictive budget, the majority (34) of the interviews took place in the 3 Central Region districts of Masaka, Mukono and Luwero. A further 9 households were collected in the Eastern Region and Northern Regions some households of which formed a three wave quantitative panel form the 1992/99 national household panel and the 2005 Northern Ugandan Survey. Qualitative Data - Life History and Timeline The main qualitative method of detailed life histories followed the approach of a semi-structured format providing comparative information about households as well as record responses to open-ended questions that arise during the course of interviews. The latter focused on critical incidents, events and factors identified by households and information that households identify as important but was not part of the questionnaire design. The life history and timeline traced an individual s life from childhood to the present day, focusing on key events. As with Bird and Shinyekwa (2003), by talking to a person about their life we also hoped to identify path-determination in individuals lives and to pinpoint key moments of choice or the absence of choice, but with the advantage of also having robust quantitative panel data to underpin this. 5 To ensure that the panel households were the same in both periods, a two part matching process was undertaken. The first stage matched the sex and age of the household head, allowing for an acceptable error range given uncertainty about precise ages etc. A second stage focused on those households whose head had changed over the period, for example where a household head had died and another member of the family had become the new head. See Lawson et al. (2006) for further details. 6

7 4. Empirical Analysis This section is disaggregated into the three main research themes of HIV/AIDS/Ill Health, Assets, and Gender Inequality and their interconnections with poverty/inequality dynamics. However, we also extend the section by providing a relatively succinct summary of the other key findings. We restrict our primary analysis to the aforementioned areas as these were themes that highlighted by the interviewed households as being particularly important, but were also priority research areas in the Research Programme of PMAU, MFPED. For each of the aforementioned areas we firstly highlight what we can establish by undertaking a pure quantitative analysis. Even considering the large volume of quantitative poverty research, we differentiate from a large section of this prior research by utilising the 1992/99 panel data and focusing exclusively on dynamics. Although previously published, and unpublished, work has utilised such data, it has rarely, if at all, provided exclusive focus on such issues. Hence, these quantitative findings in themselves add value to our understanding of such issues. We then follow up the quantitative findings with examples of how life history (and timeline) analysis were then used to further our understanding of the propagators, maintainers and interrupters of poverty. Ill Health/HIV/AIDS Of the previous econometric work that has touched on ill health and poverty dynamics (Deininger 2003, Lawson 2004) an association between the two has been established. However, by using the 1992/99 panel data we can see from both descriptive and econometric analysis that ill health not only appears to be associated in some way with both chronic and transient poverty, but particularly with households that have moved into poverty, or are chronically poor. For example, Table 1 indicates that initial health status of the household head appears to be associated with households that move into poverty (chronically poor) above average proportions, 28.3% (21.8%) of households that moved into poverty (chronically poor) were headed by somebody who was sick. Above average proportions (8.08%) of households, that moved into poverty, were headed by somebody who was suffering from a long term sickness. 6 Table 1: Chronic and Transient Poverty By Health Status Chronic Poor Moving out of Poverty Moving in to Poverty Never In Poverty (1) (2) (3) (4) (5) Household Head Health Status Proportion of households with head who has a long term sickness (>10 days), in % 6.44% 8.08% 6.92% 6.87% Proportion of households with head sick, in % 15.6% 28.3% 18.9% 17.9% All 6 NOTE: Ideally, when looking at long term sickness, in the context of microdata, one would like to analyse HIV/AIDS. However as this was not accurately recorded in the UNHS 1999 survey, a proxy of long term sickness is used for HIV/AIDS and for those individuals with more severe sickness. There are obvious drawbacks to this assumption, therefore limiting the conclusions that can be drawn about HIV/AIDS. 7

8 We can also see from econometric analysis in Appendix Table 2 that: If a household head is sick, in 1992, then the household is significantly associated with an increased probability of moving into poverty (by 3.5 percentage points) and significantly associated with a reduced probability of never being poor (6.7 percentage points). 7 In addition, households headed by females who are ill face a higher probability of moving into poverty, compared with households headed by ill men. 8 In fact, households headed by women and suffering from (long term) sickness are up to (15) 7 percentage points more likely to fall into poverty than similarly defined male headed households. Such findings add value to our understanding of such issues, however, they still do not tell us the full story regarding the processes through which ill health might then result in a household moving into poverty or becoming chronically poor? The direction of causality is unclear, for example, are households more likely to be chronically poor/move into poverty and as a result of being unable to eat then fall sick, or vica versa. In addition, the quantitative data analysis was also unable to clarify how HIV/AIDS was related to poverty dynamics primarily because large scale household data on such issues tends to very poor. Q2 research enlightens us further. In the life history interviews, a number of households highlighted ill health to be a major influence in their deteriorating welfare status. We use a timeline example, and life history extracts, from a household that moved into poverty to highlight some common findings regarding the impact of Ill health/hiv/aids on welfare status. The first point to note is that, upon corroboration of the quantitative micro panel data, we find that it correctly suggested the household to have moved into poverty. However the time line and detailed qualitative information only support this after the mid 1990 s coffee boom ended. In fact, when originally interviewed the household head identified the 1990 s as very good times. She explains the primary reason for this as being: She.could would work for labour and brew beer In addition she used to dig, produce, collect and sell maize. Somebody next door used to buy the maize. Several of her children used to crop the land and trade coffee. We could choose what to eat whenever we wanted.. and even owned a radio. However, in the mid to late 1990 s all of the adult working children died, with this appearing to be the major cause of the downturn (see figure 1) and subsequent movement into poverty. The household head remarks that four of the main income earning children died.. one of my sons and daughters died of AIDS (1999, 1992 respectively), and two sons died of Malaria (1992, and 1996). 7 All significant results are at or below the 5% level. 8 Significant at the 10% level for sick regressions and at the 5% level for long term sickness results. 8

9 In addition, since 2000 though the household head thinks her old age has caused things to get worse as she can t dig and she can t see properly and has weakness in her hands. As much as there were good times we now no longer have a radio (we sold it for 80,000 Ug Sh when he (the son with AIDS) was sick, to pay for the health centre drugs). After this the house started falling down etc. At the time of death, however a local organisation did help with the funeral costs. The household head currently eats 1 to 2 meals per day, compared with 3 meals in the mid 1990 s. Figure 1: TIMELINE (Drawn by Interviewee) 1 of 5 Siblings - 3 girls and 2 boys. Looked after livestock for father until the age of 18. Produced 12 children in 1 marriage. They grew a lot of coffee and a lot of maize, 3 bedrooms in the family home they still own this but it is very run down and they don t have the manpower to crop the land. 3 mature children working, household did well, traded, cropped etc., Times were good but children then died of AIDS (1995/6/9) sold assets. Lady is too old to crop land, mature children have died, daughter now in jail. Birth Although the above focuses only on one qualitative follow up example of how HIV/AIDS and general sickness can impact on a household that moved into poverty, the processes underpinning such a movement were commonly observed throughout the life history data collection. A very common sequence of ill health/poverty dynamics associated events identified by both chronically and transiently poor households was identified as follows: 1) Many households increasing welfare levels (often after the 1992 data had been collected), sometimes acquiring both productive (livestock etc.) and nonproductive assets (radios etc.) during the coffee boom of the mid 1990 s. 2) Such periods of increased wealth often resulted in an increased family size (and it appears an increase in the number of sexual partners particularly for men who increased incomes during this time). 3) Serious sickness of the main breadwinners followed, with the main income earners being unable to work but also draining finances and selling assets (commonly after reducing food consumption see next section) to pay for health bills or food consumption. 4) The additional household size that resulted form the boom then, at a later, further accentuates the households reduction in welfare particularly when the offspring 9

10 were of a schooling age and extra resourcing demands resulted (e.g. school fees, increased health costs, increased food demands of children). We have confirmed that the pure quantitative analysis indicates a clear association between sickness and poverty dynamics, but this appears to be more closely associated with households that were chronically poor or moving into poverty, than others. However, the life history information has provided clearer understanding regarding the underpinning process. If the households are to be believed the causality is one of sickness that then causes reduced welfare/movement below the (official monetary) poverty line. This appears acceptable for several reasons, but no more so than because many households reduced food consumption as a result of reduced incomes (i.e. the breadwinner being unable to work) - by implication such households are more likely to experience a decline in, officially defined, welfare as the accepted poverty line is derived from food consumption. Such a process appears to have been particularly accentuated during the high (but declining) period of HIV/AIDS prevalence, in Uganda. Assets There is relatively little prior quantitative analysis that has focused on the issue of assets. Perhaps of particular note is the work by Okidi (2004) that used the retrospective 1999 data to produce a descriptive analysis of the 1992/99 changes in assets. Inter-linking similar data with poverty dynamics, and extending the aforementioned results we find, amongst other things that (see Table 2): Chronically poor households are less likely to own cattle, and to own smaller quantities when they do. Both the chronically poor households and those that moved into poverty cultivated less land, and experienced smaller increases in land area cultivated, between 1992 and 1999, than compared with h average. Table 2: Land and Cattle Assets and Poverty Dynamics Poverty status 1992/99 Chronic Moving Moving Never In All Poor out of Poverty into Poverty Poverty (1) (2) (3) (4) (5) Asset ownership Average land area cultivated Average land area cultivated Average number of cattle owned Average number of cattle owned As with the ill health and poverty dynamics descriptive data, the data for assets and poverty dynamics are enlightening but tell us little regarding the interesting stories that underpin these movements for example, they don t inform us of: Why Chronically Poor Households Have Experienced Smaller Increases in Land Area Cultivated? If Ill Health Leads To Asset Depletion and Therefore Move Households into Poverty? 10

11 When we cross tabulate ill health, asset levels and poverty dynamics status of households (Table 3) we find: Comparing healthy and non-healthy headed households (column 11) that not only are land areas smaller for the sick than non-sick (3.54 acres and 4.59 acres respectively) but land increases for the sick are almost half that of the non-sick (65.7% compared with 36.7% for the sick). Similar (un-presented) trends were found to also exist for other enterprise assets such as chickens and cows. Once again however, although the data shows that sick households have lower land increases it does not inform us of the very processes that underpin this, in addition to the direction of causality, i.e. Does ill health propagate the asset depletion which then forces households to move into poverty? Does the movement into poverty cause ill health, by having less money to spend on food, reducing consumption etc.? and then asset sales occur to cover for the loss in income? Unfortunately, we can not provide clear answers to such questions from the 2 wave panel data, and even adding community based participatory information does not completely clarify the situation. However, by collecting life history information from the same households for the descriptive where generated allows us to at least start providing insightful answers to such questions. In this case, we explore this further, through the example of a Never Poor household, with a brief household summary provided as a background starting point (Box 3) BOX 3: Summary Box A Never Poor Household An agricultural/rural household that has met its needs over the years through mixes of farming, fishing and charcoal burning. They managed at one stage to accumulate enough to build a good quality house. Income has levelled out since they were stopped from charcoal burning in mid/late 1990s. Food production appears to have dropped in last 4 years as B has been unwell and unable to dig. Main problem they have faced is paying health costs had to sell a bicycle to cover A s costs and B has not had an operation she needs (80,000 Ug sh). Having a large family has reduced the possibility for accumulation of non-human assets and increased health/education expenses. There are also effects of B s 11 pregnancies on her health. 11

12 Table 3: Household Head Health Status and Assets Chronic Poor Moving out of Poverty Moving into Poverty Never In Poverty All Asset Levels and Change Sick (1) Not Sick (2) Sick (3) Not Sick (4) Sick (5) Not Sick (6) Sick (7) Not Sick (8) Sick (10) Not Sick(11) Number of Cows at Number of Cows at % Increase in Cows Numbers -63.6% -14.3% -61.5% 14.0% -2.4% -20.6% -4.7% 65.6% -22.1% 26.3% Table 4: Poverty Status - By Gender of Household Head (as defined in 1992) Poverty Status By column Unmarried (1) Male Headed Household (MHH) Married/ Cohabiting Divorced Widowed (2) (3) (4) All (5) Unmarried (6) Women Headed Household (WHH) Married/ Cohabiting Divorced Widowed (7) (8) (9) Chronic Poor 8.3% 20.0% 16.7% 16.7% 19.3% 10.5% 27.9% 10.6% 14.9% 17.5% 18.9% Moving out of Poverty 27.8% 29.8% 33.3% 16.7% 29.6% 21.1% 27.9% 19.1% 39.4% 30.3% 29.7% Moving into Poverty 13.9% 10.0% 6.7% 5.6% 9.9% 0.0% 17.6% 14.9% 8.5% 11.8% 10.4% Never In Poverty 50.0% 40.2% 43.3% 44.4% 41.1% 68.4% 26.5% 46.8% 37.2% 40.4% 40.9% All (10) All (11) 12

13 Even for this never poor household assets play a major role during times of crisis. However, the life history information reveals the story to be more complex than a household that just sells assets when faced with unmanageable health care, or other costs. For example we found that: There appears to be clear preferencing in relation to the types of assets sold (e.g. luxury goods such as radios were sold first), but across household we found that the willingness to sell any assets were dependant upon the age and geographical location (e.g. older household heads were substantially less willing, than younger headed households, to sell any livestock firmly believing that they were looking after the assets on behalf of future generations ). It was clear that such households realised that selling such assets would substantially reduce coping strategies in the future. However, such a process also involves an element of Asset Smoothing e.g. in order to pay for medical bills/transport etc., food consumption was commonly found to be reduced for a period of time. Only after experiencing reduced food consumption would assets then be sold. Therefore, by combining the aggregated national level quantitative finding with an admittedly small, but consistent, number of life histories it appears that virtually all households that resorted to desperation sales (particularly as a result of terminal sickness which accentuated health costs), had firstly asset smoothed. In other words, all such households reduced food consumption in the first instance (many form 3 to 2 meals per day, but some from 2 to 1 meal per day) prior to selling assets. It was common to find younger household heads to prefer asset smoothing by obtaining an MFI loan, and not reducing food consumption because they felt that they would be able to earn money to repay such loans in the future. Gender We have already seen how the quantitative data indicates that households headed by females who are ill face a higher probability of moving into poverty, compared with households headed by ill men. We now try and take this forward and see how the quantitative data can provide insights regarding issues such as asset inequality, across gender. For example, descriptive data in Table 5 would suggest that Women Headed Households (WHHs) are more likely than Male Headed Households (MHH), to report smaller areas of cultivatable land among their assets, as at 1992 (Columns 5 and 10), 1999 and Table 5 : Cultivated Land Assets Disaggregated by Marital Status and Gender of Household Head MHH WHH Married/ Married/ Unmarried Cohabiting Divorced Widowed All Unmarried Cohabiting Divorced Widowed All All (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11l Source: UNHS1999 (for 1992&1999 data) and UNHS

14 In addition, if we look at gender inequality issues, for example for health care access, we can see from Table 6 that within households adult women are more likely to be sick (31.8%) than men (24.3%). If we consider health care demand relative to the income of a household, using the 2002/3 data, we also find women in the poorest and richest households receive disproportionately less treatment, when sick, compared to men. 9 Once again, however this tells us little regarding the process that underpins and explains WHY? Table 6: Proportion of Men, Women, Boys and Girls Within Households Sick (not) and Having Treatment (not) Sick Not sick Treatment No Treatment 10 Men 24.3% 75.7% 89.7% 10.3% Women 31.8% 68.2% 90.0% 10.0% Boys 26.0% 74.0% 93.8% 6.2% Girls 26.7% 73.3% 93.9% 6.1% Source: UNHS 2002/3 Although due to limiting sample sizes we were unable to further test whether WHH s are were statistically associated with being chronically poor a number of gender associated findings where observed. In particular, the descriptive data suggested: That there appeared to be an element of gender inequality, in favour of men, when seeking health care. The life histories provide only limited support for this. For example, only in one instance (chronically poor household) was it highlighted that women were discriminated against when medical care was being sought. Even in consistently poor households there appeared to be little gender bias in relation to who received medical care. However, in several household the social safety nets of WHH s appeared substantially less developed than those of MHH s. There were several instances of this however one example of a never poor household highlights this better than most. The household was relatively comfortably off, but then her husband s HIV/AIDS began to make him weak and he could not work much and she had to spend more and more time taking care of him. The wife became the main provider. As he got weaker and could not go out (and 8 months before his death he became bed ridden).friends visited and gave him a little money or gift - they never had a problem trying to get him sugar, milk, eggs, fruit, fish etc. because they were commonly given as gifts. He had many friends because he used to drink beer in the evening and when they did not see him they would call around. However after the man died she found out that she had HIV/AIDS but she did not receive similar support because being a woman, she spent more of her 9 Health care demand refers to the whether health care has been sought when an individual is sick. 10 Unless otherwise stated No treatment refers to no medical care of any type (i.e. not even home treatment) 14

15 time at home and did not make such wide social networks, therefore limiting empowerment.. The above partly explained the very common finding that few women, particularly in rural areas, were partaking in formal labour market activities. This supports econometric findings that suggest potential gender bias in formal labour market participation. When asked explicitly about this issue, several individuals commented on their limited contacts, further supporting the social network finding (above). However, several of females also commented on the need to sleep with employers to get a job. Other Findings Of the other finding being part of organisations, even if it was attending church on a regular basis appears to have increased peoples coping strategy, though increased assistance networks. Of particular note of the other highlighted findings the issue of beiland eviction/lack of land access was mentioned regularly Limited Access to Land, Loss/Eviction from Land was noted as a growing concern by several, mostly chronically poor, households but also in community group participatory meetings. For example, one household had bought 5 acres of land in 2001, and although they possesses a letter of receipt the title holder sold off a very big piece of land that included their 5 acres households on this land were all chased off. Some households noted this to be a direct result of recent land reforms. Boxes 2 & 3 provide an example of how this has further impacted on a chronically poor household. BOX 2: (Abbreviated) Life History 11.of a Chronically Poor Household A was born in 1947 in the village that he currently resides. He is the oldest of 4 siblings (2 boys, 2 girls one of which died at a very young age). When A left school he looked after his father s cattle (there were 2 cows when he started doing this, but this increased to 40, by 1983). However, the by the end of the war, in 1986, all of our cows had been stolen or had died...we then tried to accumulate assets once again and in 1992 we bought 2 cows, by 1995 we had 6, but since then 4 of them died and 2 were sold when the old man was sick (and died in 1996). The entire family has recently been evicted from the land that they had occupied, without a land title, since This has de-stabilised their lives and created great uncertainty. Despite the household being headed by a hard working couple they are now unable to make any agriculturally related medium term planing decisions. Despite some good times prevailing in the mid-1990 s (e.g. the number of cows/assets increasing until the death of household members forced sales), tt is clear that psychologically at least things are worse and the families long term income prospects are diminished due to land insecurity (see box 3). 11 See Separate Time Line Information Box for further life history information on this individuals and household. 15

16 A common finding from many interviews was that several marked events/depressions appeared to have compounded poverty/welfare decreases in households. This seems to be in direct support of (Bird and Shinyekwa 2003) who found that a series of shocks/problems often compounded each other. Using a multiline analysis of a chronically poor household (same as in box 2) we are able to highlight through the recent economic, occupation, education, social and psychological, migration history of a chronically poor household head. BOX 3:Detailed Life History ( ) + Timeline Information 12 of a Chronically Poor Household Economic (and Occupation) History In 1992 the household bought, with his father, 2 cows and by 1995 they had 6 cows. I remember that my father bought a bicycle that improved our well being (+3) and the cows also (+2). During this period things were good and in 1993 we built an additional room onto his house. This put me/us at a good level of well being (3 or 4). However 4 cows then died and 2 were sold when the old man was sick (he died in 1996). The crops grew well during the mid-late 1990 s. (+1) but not since then (-1). The household size has increased from 6 (1999) to 9 (2005) this has created extra economic pressure on us (-2). In 2003, 3 goats died (-2) and in late 2003 one of my sons died after being in hospital for some time, we had to borrow 100,000 Ug Sh (30gbp) for hospital fees. (-3) In 2004 we were evicted from the land/home that we had stayed on since the 1960 s, by the owner who put herdsman and cattle on the land. This resulted in a very large decline in our wellbeing as we had to move and I had no job (-5). Since then the man across the road from where we now live (next door to the plot of land from which the household was evicted) said he could stay on his land (increase of +1) so we now stay here. In addition I also rent 1 additional acre (increase of +1) to crop food. However, I would like to plant cash crops on this additional land but can t as the owner would charge me (at the moment I pay rent in kind by giving the landlord crops). From an occupational perspective several of the economic history events overlap, but as a herdsman for his father and as an agriculturist for most of his life, the household heads occupation was relatively stable until the deterioration in crop output and loss of land. Social, psychological and cultural In 1996 my father was sick the family still ate 3 meals per day and we didn t borrow money - at the time of the funeral none of the family, including his father owed money to anybody. When my father died, this left a gap at the head of the family but I ve tried to fill this (-1) In 1998 and December 2003, my two sons died. The latter of these died in hospital, but we had to borrow money for hospital fees that we have still not repaid (-3). This 12 The number in brackets represents the level of severity attributed to an event by the household respondent. 16

17 resulted in a very large decline in well being. Shortly after this, in 2004, we were evicted from our house and land and I moved the entire family to town (for 6 months) but I could not find a job this was a very bad time for all of us. We didn t have enough money and we could not afford to send the penultimate child to school (-4). Late in 2004 a former neighbour told the head that he could stay on his land so they moved to a plot of land and house that is adjacent to the old one. However, he can t tell how long he will be at this current household and plot of land, and has not asked the owner as he does not want the landlord to consider this. Therefore, although things have improved (+1 increasing overall welfare to 6): There is a feeling of great uncertainty regarding the future, we don t know how long we have here, and things are very uncertain and I dare not ask the owner how long we can stay as this might be tempting fate. He says that there have been no good events recently, other than a child having been born (+1). Education history The head of this chronically poor household has 2 years of primary education and has not completed any subsequent training, formal or otherwise, and as such feels that his educational status/history has remained relatively stagnant over the period in question and has not improved or decreased his well being. NOTE: indicate quotations from the interviewee Combining the above details reveals the overall combined impact of economic, social, education and occupation history of this Chronically Poor household. Figure 2: Consolidated Time Line Analysis for (Chronically Poor Household) Welfare Occupation Economic Social, Psychological Education Year 17

18 Firstly, the death of the father in 1996 caused a negative short term psychological and economic impact. However, this period also coincided with the loss of 4 cows due to disease therefore substantially depleting reserves to call upon in emergencies. Combining such events with the lower crop production that followed, this marked a period of generally lower levels of welfare partly explaining why the poverty gap, over the period , increased for this household. 13 Secondly, although the occupation remains the same in the late 1990 s there is further economic and social downturn as deteriorating rainfall and crop sales reduce crop productivity and sales, combined with the death of one son. These events are shortly followed by a third series of shocks in 2003 when assets are further reduced (death of goats through disease) and one son dies and land eviction (2004) (economic and social impacts). Overall therefore, and regardless of however complex the quantitative analysis, we are sometimes restricted by what, even national representative, panel and cross section data can inform us of (particularly when there are only 2 waves of panel data). For example, we found the quantitative data to indicate that ill health was associated with chronic and transient poverty but this did not reveal to us the potentially more policy relevant stories/understandings relating to issues such as if distress sales of assets do occur then at what point do families resort to selling the assets? or if a household asset smooth before selling assets? When we combined the life history information we were able to provide a far richer understanding regarding a few of the main PROPOGATORS, MAINTAINERS AND INTERRPUTORS of poverty. Before concluding with a summary of the findings and some brief policy issues, it is however worth providing a more substantive methodological comment. 13 The shortfall of the welfare of this poor household from the poverty line, expressed as a proportion of the poverty line, increased from 0.67 (1992) to 0.75 (1999). 18

19 5. Summary The research represents a relatively rare attempt at obtaining qualitative information from the same households for which multiple wave multiplex income and expenditure surveys have been completed. Many previous attempts at combining quantitative and qualitative research have rarely been genuinely integrated in such a way. In this instance the sampling frame of the quantitative panel was adopted for the life history interviews, and although the sample was too small to be nationally representative, the idea behind the work was to produce one of the first integrated works to highlight how such methods could be used to explain some of the Propagators, Maintainers, and Interrupters of Poverty. In this case, and considering the very limited funding, we were able to focus on three main issues of ill health, assets and gender. From a thematic perspective despite focusing on 3 main areas, we have (hopefully) been able to show how the combined Q2 approach can provide a rounded understanding regarding some of the Propagators, Maintainers and Interrupters of Poverty. Although it is recognised that the sample is fairly small and themes restricted, it would appear fairly clear that what the likes of Appleton and Booth advised in terms of combining research methods really can provide a deeper and very fuller understanding of important poverty/inequality/dynamics issues. In particular the Q2 approach adopted here, has helped further explain a number of issues that lay behind some of the themed research areas highlighted by The Government of Uganda. For example: There appears to be more than just an association between sickness/ill health and poverty dynamics. At least in the processes highlighted by the interviewed, there appears to be direct causality between sickness and movements into poverty. There appears to be clear preference ordering in relation to the types of assets sold in times of crisis (e.g. luxury goods such as radios were commonly sold first), but the willingness to sell any assets was dependant upon the age and geographical location of the household/head (e.g. older household heads were substantially less willing to sell any livestock firmly believing that they were looking after the assets on behalf of future generations). Such households recognised their limited potential in regenerating such assets and selling such assets would substantially reduce future coping strategies. Asset Smoothing seems to be very common e.g. in order to pay for medical bills/transport etc., food consumption of the chronic and transiently poor households was virtually always reduced for a period of time, before selling assets (from a pure food poverty consumption measure perspective, this may mean that they immediately fall below the poverty line). However, asset smoothing was found to be highly dependent upon the age and geographical location of the household. In some younger headed households both Asset and Consumption Smoothing occurred at the cost of obtaining a microfinance loan. 19

20 Hence, we found households avoided moving into (deeper) poverty by reducing food consumption, selling assets, and taking loans, but the order and extent to which these occurred varied by socio-economic characteristics. Gender inequality through social networks may be present and may at least partly explain some of the reasons for higher proportions of WHH s being chronically poor. Limited contacts in the job market also appear to partly explain why gender inequalities exist in formal labour market participation. Of the other findings, not covered in detail above, we also found that: The national, panel and cross section, quantitative household data suggest that an increase in household size is statistically associated with a movement into poverty. This was also supported by the life histories. A very common finding was for households to experience increased wealth (e.g. mid 1990 s coffee boom) and for household sizes to increase. But it is only 5-6 years later (possibly when incomes have fallen) that delayed costs of school fees etc. start causing a monetary impact/problems - some of which are solved by selling assets. The panel data analysis does not clearly pick up potential vulnerability households i.e. the extent that households may be about to have to adopt or cover for sick brothers and sisters who have siblings and who are likely to be adopted into the family. The qualitative interviews allow both current and future vulnerability to be assessed to a greater extent. Disputed land was mentioned by several households, and in group participatory meetings, as a growing concern in communities. For example, one household had bought land of 5 acres in 2001 although he possesses a letter of receipt the title holder sold off a very big piece of land that included his 5 acres households on this land were all chased off. Limited access to land was particularly prevalent amongst poorer households From a policy perspective, perhaps the most topical, and controversial, finding relates to that of land access, as these preliminary findings seem to indicate at the micro level there could be a limiting of agricultural productivity, due to lack of access/ability to plan long term. This was particularly the case for the poorer households. 14 Further to this, and perhaps equally as sensitive, but building on a steadily growing Uganda specific population growth literature, it is apparent that having a larger family, and particularly larger dependency ratios, is associated with being chronic and transiently poor. 15 With such effects often greatly accentuated by illness and death, such finding perhaps point to the importance of social protection as a having key role to avoid persistent poverty and inequality. 14 Further microeconometric evidence focusing on productivity would assist this claim. 15 This is consistent with an array of cross country poverty dynamics evidence e.g. Baulch and McCulloch for Pakistan etc. 20