POLICY RESEARCH WORKING PAPER 2106 Subjective Economic Welfare As conventionally measured, current household income reiative to a poverty line can Martin Ravallion onlv partially explain how Michael Lokshin Russlan adults perceive their economic welfare. Other factors include past incomes, individual incomes, household consumption, current unemployment, risk of unemployment, health status, education, and relative income in the area of residence. The World Bank Development Research Group Poverty and Human Resources H April 1999 POLICY RESEARCH WORKING PAPER 2106 Summary findings Paradoxically, when economists analyze a policy's impact They find that Russian adults with higher family on welfare they typically assume that people are the best income per equivalent adult are less likely to place judges of their own welfare, yet resist directly asking themselves on the lowest rungs of the subjective ladder them if they are better off. Early ideas of "utility" were and more likely to put themselves on the upper rungs. explicitly subjective, but modern economists generally But current household income does not explain well ignore people's expressed views about their own welfare. self-reported assessments of whether someone is poor or Even using a broad set of conventional socioeconomic rich. Expanding the set of variables to include incomes at data may not reflect well people's subjective perceptions different dates, expenditures, educational attainment, of their poverty. health status, employment, and average income in the Ravallion and Lokshin examine the determinants of area of residence doubles explanatory power. subjective economic welfare in Russia, including its Healthier and better educated adults with jobs perceive relationship to conventional objective indicators. For themselves to be better off, controlling for income. data on subjective perceptions, they use survey responses The unemployed view their welfare as lower, even in which respondents rate their level of welfare from with full income replacement. "poor" to "rich" on a nine-point ladder. Individual income matters independent of per capita As an objective indicator of economic welfare, they use household income. the most common poverty indicator in Russia today, in Relative income also matters. Living in a richer area which household incomes are deflated by household- lowers perceived economic welfare, controlling for specific poverty lines. income and other factors. This paper- a product of Poverty and Human Resources, Development Research Group - is part of a larger effort in the group to better understand the relationship between "objective" and "subjective" economic welfare. The study was funded by the Bank's Research Support Budget under the research project "Policies for Poor Areas" (RPO 68 1-39). Copies of this paper are available free from the World Bank, 1818 H Street NW, Washington DC 20433. Please contact Patricia Sader, room MC3-632, telephone 202-473-3902, fax 202-522-1153, Internet address psader@worldbank.org. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/html/dec/Publications/Workpapers/home.html. The authors may be contacted at mravallion@worldbank.org or mlokshin@aworldbank.org. April 1999. (39 pages) The Policy Research Wiorking Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Produced by the Policy Research Dissemination Center Subjective Economic Welfare Martin Ravallion and Michael Lokshin' Development Research Group, World Bank. The financial support of the World Bank's Research Comnmittee (under RPO 681-39) is gratefully acknowledged. For their comments the authors thank Jeanine Braithwaite, Christine Jones, Branko Milanovic, Dominique van de Walle and Bernard Van Praag. 1. Introduction It is a paradox that when economists analyze the welfare impacts of policies, they typically assume that people are the best judges of their own welfare, yet they resist directly asking people themselves whether they are better off. It is assumed instead that the economist knows the answer on the basis of objective data on incomes and prices. While early ideas of "utility" were explicitly subjective,2 the modem approach in economics has generally ignored the expressed views of people themselves about their own welfare. However, the view that we can make interpersonal comparisons of welfare by looking solely at demand behavior is known to be untenable. Households differ in characteristics, such as size and demographic composition, which can influence their welfare in ways which are not evident in their behavior as consumers (Pollak and Wales, 1979). The problem stems from the restrictions economists place on the information which is brought to bear in measuring welfare. Responses to survey questions on perceived well-being may well provide the extra information needed for identifying welfare (van Praag, 1991; Kapteyn, 1994). There has been work by psychologists and economists on understanding self-perceptions of welfare. The most direct and common approach is to survey people's opinions about their own well-being, assuming that the answers are inter-personally comparable. The questions asked by subjective welfare surveys have typically related to a broad notion of "happiness" or "satisfaction with life". The evidence available does not suggest that the answers given can be predicted well by standard 2 Famnously, Jeremy Bentham defined utility in terms of psychological states (pleasure and pain). On the distinction between "objective" and "subjective" utility see Kahnman and Varey (1991). 2 objective measures, such as income.3 However, for many purposes, including policy analysis, one is typically interested in a more narrow concept of "economic welfare"; is one person "poorer" than another? Possibly subjective assessments of this more narrow concept will accord better with objective data on incomes. A better understanding of the determinants of self-rated economic welfare may also help understand the political economy of economic policy making, such as why some sub-groups in society appear to be more opposed to policy change than a conventional calculation of income gains and loses would suggest. This paper examines the systematic determinants of subjective economic welfare in Russia, including its relationship to conventional objective indicators. Our data on subjective perceptions use survey responses to a question in which respondents say what their level of welfare from "poor" to "rich" is on a nine-point ladder. We make the standard identifying assumption in analyzing attitudinal questions that there is inter-personal comparability of the interpretations given to the survey question; in our case, a given rung of the ladder is taken to mean the same thing to each person in terms of a continuous latent measure of economic welfare. Of course, there are still systematic differences in where people place themselves on the ladder, but these are interpreted as arising solely from differences in their economic welfare. Our objective indicator of economic welfare is the most common poverty indicator in Russia today, in which household incomes are deflated by household-specific poverty lines. 3 Simon (1974) found a weak association between income and subjective welfare in the U.S., and survey evidence since has generally suggested a significant but low correlation; for a survey see Fumham and Argyle (1998, Chapter 11). Also see Scitovsky (1978), Easterlin (1995) and Oswald (1997). 3 For a number of reasons, Russia is an interesting setting for this inquiry. The economic and political reforms of the last decade had a profound impact on the well-being of Russian households. A sharp drop of GNP was accompanied by a high level of inflation in the early 1990s, an increase in unemployment, and income inequality. The poverty rate rose sharply (Lokshin and Popkin, 1997). The problem of identifying those living in poverty is of increasing importance in Russia. Usually the problem is addressed by comparing household income with a poverty line which varies according to the prices faced and household size and demographic composition. In practice, this method has tended to show that it is larger and younger families that have higher incidence of poverty, and this finding has had considerable influence on anti- poverty policies. However, there appears to be a perception amongst many in Russia today that poverty is more acute in older (particularly pensioner) and smaller households. This suggests that there may well be disagreement between objective and subjective welfare indicators. We aim to compare the two types of indicators and to better understand the source of any divergence. The following section discusses alternative approaches to welfare measurement, and our approach and data. Section 3 gives our results comparing the objective and subjective measures. Section 4 tries to explain the differences. Section 5 concludes. 2. Measuring economic welfare The most widely used measure of a person's "economic welfare" is the real income of the household to which the person belongs, adjusted for differences in family size and demographic composition (relative to some reference, such as a single adult). This can be defined as the household's total income divided by a poverty line giving the cost of some reference utility level 4 at the prevailing prices and household demographics. Under certain conditions, this ratio can be interpreted as an exact money metric of utility defined over consumptions.4 Standard practice is to calibrate the cost function from consumer demand behavior. It is known, however, that there the parameters of the cost function are in general under-identified from demand behavior when household attributes vary (Pollak and Wales, 1979).5 This problem has plagued applied work, and the policy interpretations of data on economic welfare including "poverty profiles" aiming to give consistent measures of poverty across sub-groups of society. For example, consider one property of the cost function, its elasticity to household size when evaluated at the reference utility level used to set the poverty line. An elasticity of unity is equivalent to dividing income by household size, while an elasticity of zero implies that aggregate household income is the relevant indicator of individual welfare. There is evidence to support the intuition that at some critical value of that elasticity somewhere between zero and one, measured poverty will tend to be uncorrelated with household size, while at elasticities above (below) this critical value larger (smaller) households will be deemed poorer (Lanjouw and Ravallion, 1995). The same evidence indicates that the range of approaches to determining the size elasticity of the cost function found in practice spans both sides of this critical elasticity, 4 See Blackorby and Donaldson (1987) who refer to consumption normalized by a poverty line as the "welfare ratio". The main assumption required for this to be an exact money metric of utility is that the consumer's preferences are homothetic. 5 Also see Pollak (1991), Blundell and Lewbel (1991), Browning (1992), and Kapteyn (1994). To understand the problem, suppose that we find that an indirect utility function v(p, y, x) (depending on prices p, income y and other household characteristics x) supports observed demands q(p, y, x) as an optimum. The indirect utility function then implies a cost function c(p, x, u) for utility u, such that the objective welfare indicator is y/c(p, x, ur) for the reference utility ui (interpretable as the poverty line in utility space). However, if v(p, y, x) implies the demands q(p, y, x) then so does every other indirect utility function V(v(p, y, x), x]. 5 so that one could go from saying that larger households are poorer to the opposite depending on the way one identifies the elasticity. This indeterminacy has bearing on policy. The demographic poverty profile is important information for a number of questions in social policy, such as the allocation of public spending between family allowances and pensions. The issue has been prominent in discussions of social policy in Eastern Europe. Yet sensitivity tests on past objective welfare indicators for Russia and other countries in Eastern Europe and Central Asia suggest that the demographic profile of poverty can change appreciably with even seemingly small changes in the allowance made for scale economies in consumption within the household (Lanjouw, Milanovic and Paternostro, 1998). Some economists have turned to data on self-perceptions of welfare as a source of the extra information needed for identification. There are various approaches. Van Praag (1968, 197 1) introduced the Income Evaluation Question (IEQ) which asks what income is considered "very bad", "bad", "not good", not bad", "good", "very good". Another method is based on the Minimum Income Question (MIQ) which asks what income is needed to "make ends meet". Subjective poverty lines can be calibrated to the answers (see, for example, Kapteyn et al., 1988).6 By this approach the welfare indicator is still taken to be objectively measured income or expenditure normalized by the (subjective) poverty line. A common use of subjective welfare measures is to calibrate an objective welfare measure, such as setting equivalence scales. Survey-based subjective indicators are used to 6 Qualitative data on consumption adequacy can also be used, without the minimum income question (which is unlikely to give sensible answers in some settings) (Pradhan and Ravallion, 1998). 6 identify the consumer's cost fumction (Van Praag and Van der Sar, 1988; Kapteyn, 1994). But what variables should be included? Even within a given approach to measuring subjective welfare, the set of individual characteristics deemed relevant to the corresponding objective welfare measure can differ. An objective welfare indicator chosen to have best fit in explaining a subjective indicator may still leave worryingly large residuals. Arguably the IEQ and MIQ are both motivated by a rather narrow, income-based, characterization of welfare. This has been recognized explicitly in the literature. For example, in estimating the Leyden poverty line using Russian data, Frijters and van Praag (1997) recognize that "..income is only one factor among others influencing individual life satisfaction levels Nevertheless, being economists, we'll assume that absolute and relative material circumstances define poverty" (p.6). They go on to calibrate their welfare metric to only a few variables - income, household size, and age. A more open-ended approach can be found in the psychological literature on subjective perceptions of welfare. While the psychological literature has naturally tended to focus on mental health, a strand of the literature has attempted to understand people's self-rated welfare.7 Respondents are asked to place themselves on a ladder - sometimes referred to as a Cantril ladder, following Cantril (1965) - according to their "happiness" or "satisfaction with life a whole."8 7 See, for example, Argyle (1987), Diener (1994), Diener, Suh and Oishi (1997) and Furnham and Argyle (1998). A strand of the economics literature has drawn on subjective assessments of welfare, following Van Praag (1968, 1971). (We discuss this approach in section 2 below.) There is remarkably little cross-referencing of the literatures in psychology and economics. 8 For a useful cross-country compendium of the questions asked, and a summary of the answers, see Veenhoven et al., (1993). 7 However, this is arguably too broad a concept for measuring economic welfare, and assessing conventional income-based measures. When one says that someone is "poor" one typically does not mean that they are unhappy. It cannot be too surprising that income is not all that matters to "happiness" or "satisfaction with life". The more interesting question is how income, and other objective characteristics of people, relate to self-perceptions of economic welfare. Here we adopt an alternative approach, in which the Cantril-type question is asked about economic welfare. In particular, we use the following question: Please imagine a 9-step ladder where on the bottom, the first step, stand the poorest people, and on the highest step, the ninth, stand the rich. On which step are you today? We will call this the Economic Ladder Question (ELQ).9 The question serves the purpose of this paper well. It does not presume that "income" is the relevant variable for defining who is "poor" and who is not, but leaves that up to the respondent. At the same time, by using the words "poor" and "rich" the question focuses on a more narrow concept of economic welfare than the "ladder of life" questions often used in psychometric and other surveys. It does not appear plausible to us that discrepancies between answers to the ELQ, as posed above, and an objective measure of real income reflect the fact that they are aiming to measure different things. Nonetheless, there are still ways in which the answers to the ELQ could deviate from conventional real income metrics. The following reasons for divergence can be identified: (i) There might well be systematic differences in the values attached to specific 9 An antecedent is found in Mangahas (1995) who asked respondents in regular surveys for the Philippines whether they are "poor", "borderline" or "non-poor". 8 household characteristics in assessing differences in "'needs", i.e., differences in the structure of the equivalence scales underlying the real income measure (as typically built into the poverty lines used as deflators) versus those which affect perceptions of welfare. (ii) The ladder question is individual specific, and there may well be inequality within households not captured by aggregate household income or consumption. (iii) There may be differences in the time period over which income is measured versus the time period on which perceptions of economic welfare are based. Past incomes may matter, as may expected future incomes, or determinants of these, such as education. (iv) There could also be differences arising from the influence of relative incomes on perceptions of personal affluence. It has been argued that the circumstances of the individual, relative to others in some reference group, influence perceptions of well-being at any given level of individual command over commodities.10 (v) From psychological research it is known that subjective welfare is affected by both transient and fixed idiosyncratic factors. Intrinsic aspects of temperament, such as extroversion and neuroticism, influence self-rated welfare (see for example Costa and McCrae, 1980). Transient effects also matter such as short-lived peaks of happiness and how a recent experience ended (Fredrickson and Kahneman, 1993). For the purpose of assessing a person's "typical" welfare, there is clearly noise in subjective welfare data. This presumably also be the case for subjective economic welfare as assessed by the ELQ. We call this noise "mood effects". To have any hope of understanding the systematic determinants of subjective economic 10 Runciman (1966) provided an influential exposition, and supportive evidence. Also see the discussions in van de Stadt et al., (1985), Easterlin (1995), Frank (1997) and Oswald (1997). 9 welfare one needs to have the above question (or something similar) asked in the context of a comprehensive objective socio-economic survey. However, the resistance to subjective welfare assessments in economics has been reflected in the nature of much of the socio-economic survey data available." The lack of integration of subjective methods into comprehensive survey instruments has meant that it has not often been possible to examine the socio-economic determinants of self-rated welfare, and the relationship with the more conventional welfare measures favored by economists. We shall use data from the Russian Longitudinal Monitoring Survey (RLMS) for 1996. The RLMS is a comprehensive survey of all aspects of levels of living, based on the first nationally representative sample of several thousand households across the Russian Federation.'2 In addition to a wide range of more conventional socio-economic data, all adults were asked the Economic Ladder Question discussed above. As our main objective welfare indicator we use the "welfare ratio" given by total household income (y) as a proportion of the poverty line (z),13 The distribution of such welfare ratios determines the level of absolute poverty. (Almost all measures of poverty are homogeneous of degree zero in incomes and the poverty line.) We use established poverty lines I For example, until recently, the household surveys done for the World Bank's Living Standards Measurement Study almost never asked for subjective assessments of welfare, even though welfare measurement was the main aim of the surveys. 12 A range of issues related to the sample design and collection of these data are explained in the documents found in the home page of the RLMS, where the data sets can also be obtained free; see http://www.cpc.unc.edu/projects/rlms/rlms_home.html. 13 We use income rather than consumption because income appears to have been more popular in past work on poverty in Russia; we leave aside the issue of which is preferable. 10 for Russia."4 These used linear programming to find the food baskets which minimized the cost of reaching predetermined age- and gender-specific nutritional norms, subject to the constraint that the quantities obtained were no lower than certain positive bounds given by the averages for those with the lowest 30% of consumption. The food basket was created separately for children aged 0-6, 7-17, adult males and females, female pensioners aged 55 and older, and male pensioners aged 60 and older. Region-specific food prices were then used to cost these food baskets. Age- and gender-specific Engel coefficients were then used to obtain allowances for non-food spending. Thus, each age and gender grouping has its specific poverty line which is used to construct a household-specific poverty line according to the demographic composition of the household. Total real monthly disposable household income (in June 1992 prices) includes wages and salaries, social security, private transfers, income in-kind and from home production. 3. Comparing objective and subjective indicators Table 1 summarizes the joint distribution of the objective and subjective indicators of economic welfare. We assign individuals to categories of welfare ratios (ylz) in such a way that the number of respondents in each category is equal to the number of respondents in the corresponding subjective welfare group. If there was a complete agreement between the two then the number of respondents in the non-diagonal cells of Table 1 would be zero. We decided to condense the highest 7th, 8th, and gth rungs of the ELQ into one due to a small number of 14 The poverty line are from Popkin et al., (1995). These were accepted as the guideline for all official Russian poverty line calculations. They are modified versions of those in Popkin et al. (1992) which were accepted as a law in Russian Federation in 1992 both on the regional and on the all Russia levels. The main modification is that the new poverty lines allow for economies of scale in consumption. 11 respondents who assigned themselves to these rungs (only 28 of the 7405 respondents put themselves in rung 8 and only 3 put themselves on rung 9). The matching of objective and subjective rankings is clearly weak. For example, of the 993 adults who said they were on the lowest rung of the ladder, only 224 were amongst the poorest 993 adults in terms of y/z. The matrix is not even dominant diagonal, though it is not far from it. The value of Cramer's V statistic is under 0. 1, though the association between the two variables is still highly significant."5 Naturally then there is only a weak matching in terms of poverty; while 29.4% of adults placed themselves in the lowest two rungs, less than half (43.0%) were also amongst the 32.7% of adults living in households with incomes below the poverty line. Figure 1 gives the mean proportion of the sample on the lowest two rungs against ln(y/z). The curve is downward sloping, but it is clearly quite flat, even near the poverty line. For example, going from 0.5 standard deviation below the poverty line to 0.5 above reduces the probability of being on the lowest two rungs from 0.34 to 0.25 (the objective poverty measure falls from 1.0 to 0.0); going " Let n,. i=1,...j=J...,J, be the number of observations in the i'th row andj'th column. The Pearson / statistics with (I-l) (J-1) degrees of freedom is defined as: x2=S EE(n, -mj) min where J I my =nnn n; n = j n; n,= n..; n. n- . 'i 'I 1 J J=l '=1 Cramer's V is a measure of association given by: V =[(X2/n)/min(I-,J_-1j)]112 for which O