Poverty, Inequality and Growth in Jamaica, , and Beyond

Transcription

1 Poverty, Inequality and Growth in Jamaica, , and Beyond Anders Danielson Economics, University of Lund POB 7082 S Lund, Sweden An earlier version was presented at the III rd CEISAL conference, Amsterdam, July 3-7, I am grateful to the Geske Dijkstra, Andy Thorpe and David Hojman for comments and to the Statistical Institute of Jamaica, the Planning Institute of Jamaica and the LSMS Group at the World Bank for permission to use the data from three Jamaican Surveys of Living Conditions.

2 Poverty, Inequality and Growth in Jamaica, , and Beyond Anders Danielson Keywords: growth, poverty, Jamaica, growth decomposition JEL codes: O11, O49, O54 Abstract This paper looks into the relations between economic growth and poverty reduction in Jamaica. A poverty profile is constructed, based on household budget survey data, and per capita growth is decomposed into pure growth effects and changes in inequality effects. I find that growth has contributed to poverty reduction mainly because there has not been any large increases in inequality. However, simulations also show that in order to for growth to continue to be efficient for poverty reduction, it is necessary to improve the prospects for rurally based, particularly small-scale agricultural growth. If not, the possibility of growth being anti-poor cannot be ignored. 1

3 1. Introduction Market-oriented economic reforms, often with IMF support, have been pursued in the majority of the Caribbean countries since the early 1980s. The main objectives of these have been to increase economic growth, reduce vulnerability to external shocks and, ultimately, help alleviate poverty. Results are not particularly encouraging. For most of the period and for most of the countries growth has been uneven and slow, attempts to diversify away from mono-crop cultures have often failed, and poverty does not seem to have been consistently reduced. While the last remark should be qualified by adding that there is no reliable panel data for some of the countries, the data that exist point to no particular trend in poverty. Thus, for instance, in several countries in the region, per capita income growth was negative for most of the 1990s (ECLAC, 1999: 9). What is more revealing is how the countries ranking on the UNDP s Human Development Index (HDI) changed in the 1990s. Of the 15 countries showed in ECLAC (1999: 39), 8 had slipped down in their HDI ranking. Of these eight countries, no less than six (all except Bahamas for which no growth data were recorded, and Haiti) recorded positive per capita GDP growth rates during the same period. 1 It is notable that the list in footnote 1 includes most of the large countries (except Cuba) in the region. Even though the HDI originally was developed to reflect a multidimensionality of poverty that the per capita income measure cannot capture, it is well known that rankings according to the HDI and per capita incomes are positive and high. This is not very surprising. The HDI consists of weighted averages of three components: income, education and life expectancy. The first one is measured by per capita income 1 The six countries in question (with average annual GDP per capita growth in parentheses) are the Dominican Republic (3.2), Guyana (5.6), Jamaica (0.6), Trinidad and Tobago (2.1), Barbados (1.2), and Suriname (0.9). 2

4 and the other two are usually reflected in the governments ability to spend, which in turn depends very much on per capita income. In fact, the simple correlation coefficient between the rank of income and the rank of HDI in 1998 was Since the correlation between income levels and well-being is so strong on a global scale, it would be interesting to look closer into the question of why so many Caribbean countries simultaneously have experienced growth of income and a slide down the HDI-scale: why, apparently, has increased growth (possibly triggered by economic reforms) failed to increase welfare on a sustained basis? Available data do not allow for a comprehensive analysis of all aspects of this question, so I focus on the relations between poverty (particularly income poverty) and economic growth. In particular I will explore to what extent economic growth has contributed to poverty reduction Jamaica and to what extent the pattern of growth mattered for this. However, while the focus of the analysis is on Jamaica, the conclusions are possibly extendable to other small and resource-strapped countries as well. The structure of the paper is this. Section 2 discusses the relations between growth and poverty in general and shows how impact of growth on poverty can be formalized. Section 3 presents a poverty profile of Jamaica in 1998 in order to pave the way for the necessary assumptions to be made in Section 4 which decomposes growth effects on poverty for three periods: , , and simulations for Section 5 concludes. 2 I used the data available at formed a rank of the 173 countries according to PPP-adjusted per capita income and another rank according to their HDI index. I then calculated the simple correlation coefficient between the two rankings. 3

5 2. Growth, Inequality and Poverty In general, the change in poverty can be written as (1) Change in poverty = Change in mean income change in inequality where poverty is measured in terms of income and inequality may be measured by the Gini coefficient or any other metric for which an increase is an increase in inequality. Expression (1) says that growth will benefit the poor unless there is a sufficiently large increase in inequality. Dollar and Kraay (2000) regressed the share of income earned by the poorest quintile against growth of mean income and concluded that growth is good for the poor: when mean income increased by x percent, the income of the poorest quintile increased by approximately as much; in their data set, the change in inequality was on average zero, so that the percentage change in poverty was as large as the percentage change in mean income. Other studies (Timmer, 1997; Mellor, 2000) have qualified that conclusion by showing that there is considerable variation around the average: sometimes the poor gain more that the mean; sometimes less; and sometimes the poor experience falling incomes when median incomes grow (Danielson, 2001 is a selective survey of the literature). The only robust conclusion appears to be that there is no clear relation between income levels and income distribution; the Kuznets Curve, once celebrated and famous, should now be in the closet with other theories that did not work out. How does growth affect poverty? Economists usually refer to trickling down a process by which growth somehow and eventually reaches the poor but exactly how this process works is seldom specified. It is reasonable to think of at least three 4

6 possible channels from increased per capita GDP to reduced poverty. The first goes through the goods market. When economic growth occurs i.e., when the mean income increases the demand for goods increases. To the extent that this demand fully or partly is directed towards goods produced by the poor e.g., food their income will increase. The second effect is through the labor market. Increasing incomes implies increasing demand for labor that is likely to affect the poor either in the form of higher wages, or in the form of better employment opportunities. Finally, economic growth means that the government s income increases, so its ability to spend is enhanced. Poverty may be reduced if the government increases spending either on activities that have an impact on poor people s health or education, or on activities that increase the poor people s ability to increase earnings (e.g., infra structure). Consequently, the impact of any given increase in mean income on poverty depends, inter alia, on how the extra income will be spent, on how the demand for labor is affected, and the conditions on the labor market, and on how the government will use its additional tax revenue. It is thus to be expected that the impact of growth on poverty will differ between countries and over time. The key issue is, of course, to what extent the poor share the fruits of growth. Limiting ourselves to income poverty for the moment, the poverty impact of growth will be higher the more growth raises the demand for unskilled labor and the goods that the poor produce. Consequently, the pattern of growth matters (Mellor, 2000; Ravallion and Datt, 1999; White and Anderson, 2000). In particular, growth which takes place in sectors where the majority of the population is poor is likely to be more efficient that growth in other sectors. For most countries, then, rural growth tends to 5

7 be more efficient that urban growth (Danielson, 2001), but this does not mean that urban growth does not reduce poverty, nor that rural growth is necessarily efficient for poverty alleviation. To formalize and to see how the impact of growth on poverty can be studied in large data sets, we follow Datt and Ravallion (1992) and note that any poverty measure P t can be expressed as (2) P = P( z,µ, L ) t t t where z is the poverty line, µ t is mean income and L t is a set of parameters describing the Lorenz curve at time t. 3 Given a poverty line, z, the amount of poverty may change either because mean income changes or because the parameters in L t changes, i.e., because the relative income accruing to certain income classes changes. Using (2), we talk of a growth component in a change of poverty as being the change of mean income with inequality held constant and a redistribution component as being the change of L t holding mean income constant. A change in poverty between times t and t+n can then be decomposed as (3) P P = G( t, t + n; r) + D( t, t + n; r) + R( t, t n; r) t + n t + where G() is the growth component, D() the redistribution component and R() a residual. The variable r in the functional expressions denote the period used as a reference; Datt and Ravallion (1992:278) show that the size of the residual depends on r. In particular, the residual with r = t will be the negative of the residual with r = t + n. This need not concern us here, mainly because in the data presented later, the 3 Usually L t contains cumulative population and cumulative income for various income classes. 6

8 residual is rather small. In Section 4, I use this to trace the impact of growth on poverty in Jamaica and beyond. 3. A Poverty Profile of Jamaica, 1998 Approximately 16 percent of the population had in 1998 incomes below the officially recognized poverty line of J$36,235 per annum (about US$950), down from over 30 percent in However, Jamaica, as most Caribbean countries, shows a consistently higher rank for the UNDP s Human Development Index than for per capita income, suggesting that these countries allocate a larger share of income to health and education expenditures than do other countries in the same income rank. Table 1 shows some social indicators for Jamaica and other Caribbean countries in the lower middle-income group. Table 1 Socio-Economic Indicators. Most Recent Estimate Belize D.R. a Guyana Jamaica T.T. b Per capita income (US$; PPP) 4,367 4,337 3,139 3,344 7,208 Primary school enrolment (%) Poverty (Head Count) Poverty (depth) c 12.5 n.a Public expenditure on health Life expectancy (years) Infant mortality d Gini coefficient n.a Sources: World Bank (1996); WDI (2000); Deininger and Squire (1996); STATIN (1998) a Dominican Republic b Trinidad and Tobago c Percentage of poverty line that is needed to eradicate poverty d Per 1,000 live births If we take as the point of departure the idea that there is a (negative) correlation between per capita income and poverty indicators, Jamaica scores quite well, even when compared to other Caribbean countries. Thus for instance, Jamaica has a marginally lower level of poverty (as reflected in a head count) than Belize, even 7

9 though the latter has a per capita income which is 30 percent above that of Jamaica s. Also, infant mortality in Jamaica is considerably lower than that in Belize and the Dominican Republic and not much higher than that in Trinidad, even though Trinidad s per capita income is more than twice that of Jamaica s. This suggests that poverty is a matter not only of growth and a high level of income but also of government policies. Indeed, distributional mobility and poverty is closely linked (Danielson, 1999). Not in the sense that low mobility inevitably causes high poverty or vice versa, but in the sense that the same government policies that affect distributional mobility also affect poverty. A major task is to identify these policies and their impact. To be able to devise a reasonable poverty alleviation strategy one must know why the poor are poor. We know that the major causes differ between continents: people are poor in many African countries because of low-fertile land; in Asia, because of lack of land; and in part of Latin America because the available land exhibits a very skewed distribution. Inasmuch as the causes seem to differ, however, there are also a number of common denominators: poor households are more often than not rurally based, poorly educated and without access to credit. The available data for Jamaica do not contradict that, although the source for Table 2 does not report on land distribution or the location of the poor. 4 We know from other sources, however (such as Boyd, 1988) that the poor are predominantly rural based. Table 4 shows that the mean size of the household is also larger for the poorer quintiles and, consequently, that the number of children is larger in poorer families. Contrary to the situation in 4 What we know (from World Bank, 1996: p.166) is that approximately 70 percent of the poor live in rural areas while approximately 55 percent of the total population is rurally based. 8

10 many other countries, however, female-headed households are not significantly poorer in Jamaica. Table 4 Characteristics of Poor and Rich Households, mid 1990s Q1 Q2 Q3 Q4 Q5 Household Characteristics Female headed households (%) Household Size (mean) Number of Children (< 15 yrs) Highest Education a Primary Secondary Tertiary Employment Unemployment Labor force participation rate Source: World Bank (1996) Note: Q1 is the poorest quintile and Q5 the richest. a Head of household As for education, the situation in Jamaica resembles the one in other countries: almost sixty percent of the household heads in the poorest quintile has, at most, primary education, while well over half of the household heads in the two richest quintiles have at least secondary education. To the extent that it is possible to draw policy conclusions from these meager data, one would be that the private returns to education seem quite high and that universal access to education of reasonable quality would be one avenue out of poverty. The data for unemployment and labor force participation confirm the suggestion that poor households are based in rural areas: unemployment in the poorest quintile is as low as in the richest quintile, and labor force participation in the poorest quintile is significantly higher than in the middle-income quintiles. The poor, then, have some land and often complement that income by working for wages on plantations. People in Quintiles 2 and 3, by contrast, are often based in urban areas (at least partly), 9

11 earning an income in the informal sector as street vendors, higglers, or in the urban, formal sector. (The large proportion of labor working in the informal sector may account for the low official labor force participation ratio.) The data in Table 2 is complemented by some economic data in Table 3, also from the mid 1990s. There are several noticeable features here. First, the share of food in the consumption basket is higher for the poorer quintiles: while the poorest households spend around two-thirds of their income on food, the richest quintile spend over half of its income on non-food items. Second, the traded goods sector is much more important for poorer households: the poorest quintile earn over ten percent of total income from the traded sector. Since the poorest quintile earn about 6.5 percent of total income, it relies disproportionately on the tradeable sector. By contrast, the richest quintile which earns about 45 percent of all income spends less than a third of all income on tradeables. Table 3 Economic Characteristics of Households, 1994 Q1 Q2 Q3 Q4 Q5 Consumption Food Non-Food Income Traded Nontraded Expenditure Traded Nontraded Total IE-ratio traded goods a IE-ratio nontraded goods a Nutritional Status Mildly malnourished (%) Severely malnourished (%) Source: World Bank (1996) Ratio of income to expenditure 10

12 Another way to see the dependence of the poorer strata on the traded sectors is to examine the rows labeled IE-ratio. Here, the ratio of income to expenditure is shown for traded and non-traded goods. The three poorest quintiles have IE-ratios for traded goods which exceed the IE-ratio for nontraded goods, while the reverse it true for the two upper quintiles. Consequently, an change in the real exchange rate will have immediate consequences for income distribution. In particular, a real devaluation will tend to favor the poorer segments of the population as this tend to increase the relative price of tradeables and the poorer strata earn more from tradeables than they spend on it. Finally, it is quite remarkable that there is virtually no sign barring the proportion of mildly malnourished in the poorest quintile of nutritional status varying with income: 5-6 percent of children in all income categories are mildly malnourished and the cases of severe malnourishment does not appear to be related to income level, either. To summarize: the poor in Jamaica have several distinguishing features. First, they tend to live in rural areas and have access to land or wage labor opportunities. Consequently, unemployment is relatively low among them. Second, they are relatively dependent on tradeables both for production and consumption, but more so for the former. Third, about two-thirds of the budget of the poor is devoted to food, so changes is food prices by, for instance, subsidies or taxes have a direct and significant impact on the poor. Fourth, while the poor tend to have larger families than the nationwide average, female-headed households do not tend to be poorer than the average. This is, perhaps, an indication that women have relatively good access to the labor market. Finally, poor household heads tend to have a significantly lower level of education than their richer counterparts. From the available data, this would seem to 11

13 be a major lesson as it points to high private returns from primary and perhaps secondary education. A poverty profile based on the data collected in the 1998 Household Budget Survey confirms the situation defined above. Table 4 shows head count poverty by region for most of the 1990s and shows clearly, first, that most of the poor live in rural areas and, second, that poverty increased in the early 1990s and then declined, with the decline being more pronounced in urban areas. I shall return to this observation in the next section. Table 4 Headcount Poverty by Region, KMA Other Towns Rural Jamaica Source: STATIN (1998), Table 7.2 Note: KMA = Kingston Metropolitan Area A final element in the poverty profile is to see to what extent individual characteristics can explain income and poverty. I approach this issue from two angles. The first is to formulate and run a regression analysis in which (the log of) annual per capita expenditures is used as the dependent variable and (the log of) a number of individual characteristics are used as independent variables. The results are in Table 5. I have run seven regressions. In all of these, the variables AGE and SEX are included. In general, coefficients can be interpreted as elasticities around the mean (with the natural exception of the binary variables SEX, MARRIED and DISABLED). With the exception of MARRIED in regression (2), all coefficients are significant at the five percent (two-tailed) test level, and most are significant at the one percent test level. 12

14 Table 5 OLS Regression. All variables in logs. Dependent Variable: Annual per capita expenditure. (1) (2) (3) (4) (5) (6) (7) CONSTANT 4.7** 4.8** 4.6** 5.09** 4.8** 4.7** 5.17** AGE 0.07** ** -0.01** 0.07** 0.07** -0.13** SEX (M = 1) 0.04* 0.06* 0.05* 0.11** 0.04* 0.04** 0.14** MARRIED (Y = 0) ** DISABLED (Y = 0) 0.46** 0.48** HH Size -0.41** -0.41** AREA a -0.16** -0.08** EDUCATION 0.46** 0.30** R2 (adj) # useable obs 16,039 10,794 16,008 16,039 16,039 16,039 10,790 Data from the Jamaica Survey of Living Conditions (SLC) 1998 * Significant at the 5 percent level; ** significant at the 1 percent level a KMA =3; Other towns = 2; rural = 1 However, the predictive value of these independent variables is surprisingly weak. Including in regression (6) a person s education in the regression (a variable which is significantly and positively correlated to annual expenditure), can only explain about three percent of the variation of annual per capita expenditure in the sample (which consists of over 16,000 observations). The other explanatory variables, when added one at a time, give similarly low predictive values. In regression (7) when all independent variables are used at the same time the explanatory and therefore predictive value increases to about 19 percent. All coefficients are significant at the one percent test level. Interpreting coefficients as elasticities suggests that the most important variables for determining annual expenditures are DISABLED (which reflect productivity), HH size, and EDUCATION. However, the area variable produces a relatively low elasticity. This may appear surprising given the data in Table 4, but may be explained by the fact that the variation within each area is substantial. To conclude with a generalization, then, the poor are likely to have low productivity, live in large households, have little 13

15 education and possibly live in rural areas. In addition, women spend less than men and the result also suggests that spending decreases with age. The second approach to a poverty profile is to examine the probability that a person with certain characteristics is poor, i.e., have annual expenditures below the poverty line. To examine this issue I used a smaller sample from the Jamaica SLC and ran a Probit with a binary dependent variable (0 = poor; 1 = nonpoor). Selected results are in Table 6. Table 6 Probability of being poor. Probit regression Coefficient t-statistic Constant HH size EDUCATION DURABLES PARISH DISTRICT Loglikelihood: -3553,9 Pseudo-R2: 0,204 # useable obs: 7,375 Data from Jamaica SLC A few new variables have been included. DURABLES measures the fraction of total expenditures devotes to durables, and PARISH and DISTRICT are regional divisions (DISTICT being finer than PARISH). The conclusion from the OLS exercise is confirmed here: household size and education are important variables for characterizing the poor. In particular, increasing the household size by one person increases the probability of being poor by 0.21 percent. A unsurprising conclusion is also the fraction of expenditures devoted to the purchase of durables is higher for the non poor. The two final variables suggest that poverty is not evenly spread over parishes some are poorer than others but that the probability of being poor is 14

16 given the other variables more or less the same irrespective of which parish or district a person lives in. 4. Growth, Poverty and Inequality, , and Beyond I here apply the ideas from section 2 to decompose poverty changes in Jamaica into changes of inequality and changes of mean income. The first is called the inequality effect and the latter the growth effect. As explained before, there is also a residual; together these three variables add up to the total change in poverty. The data used are the Jamaica SLCs for 1988, 1994 and In order to keep calculations simple, rural incomes have been scaled with the urban-rural price ratio, so that one poverty could be used. Other towns are assumed to have the same price changes as Kingston. 5 Table 7 Poverty and Income Distribution. 1988, 1994 and Urban Rural KMA Other towns Jamaica 1988 Headcount Poverty gap Gini Headcount Poverty gap Gini Headcount Poverty gap Gini Sources: Jamaica SLC 1988, 1994 and Table 7 presents basic data on poverty and income distribution for the areas in the three years under scrutiny. I present two poverty measures. The headcount is simply the proportion of the population with an annual income below the poverty line. The poverty gap measures the income (as percentage of the poverty line) needed to bring 15

17 the poor up to the poverty line. The Gini coefficient is a standard metric for measuring income distribution. It ranges from zero to one, where zero indicates that everyone has the same income and one that a single individual earns all income in the economy. Table 7 shows that income distribution has improved substantially over the period studied, particularly in urban areas (even though Kingston shows a slight increase in inequality during the latter half). Poverty is as was noted in the previous section most prevalent in rural areas. Even though rural areas have experienced a dramatic decline in poverty over the period, it still in 1998 had a headcount poverty of more than twice that of urban areas. The data also indicate that the very poor have enjoyed increases in income: the poverty gap at the national level is down from 15 percent in 1988 to only 3.5 percent in Consequently, those that are classified as poor have moved closer to the poverty line. This, of course, is likely to have a favorable impact on the distribution of income. What are the relative sizes of the growth and inequality effects? In Table 8 I break down the poverty effect according to the method outlined in Section 2 into an effect due to an increase of mean income and an increase in the distribution of income. The data should be interpreted in the following way. I first calculate the percentage change in poverty (for headcount poverty only; other results are available on request) and then break this down into the percentage due to increases in mean income (i.e., the impact of growth on poverty holding income distribution constant) and the percentage due to the change in inequality (holding mean income constant). This procedure gives a residual which is also listed in the table. Together the three figures add up to the total change in poverty. 5 No separate price index for other towns is available, but it is likely that prices in Kingston are 16

18 Table 8 Decomposition of Headcount Poverty Changes and (percent) Growth Effect Inequality Effect Residual Poverty Effect Jamaica Urban Rural KMA Other towns Jamaica Urban Rural KMA Other towns Sources: Ramadas et al. (2002). Data from Jamaica SLC 1988, 1994 and The Excel program used for calculations SimSip_Poverty is available at Note: A minus sign signifies a decrease of poverty For Jamaica as a whole, poverty (measured as headcount poverty) decreased by 8.5 percent between 1988 and 1994 and by 7.2 percent between 1994 and In the first period, poverty reduction was most substantial in urban areas particularly in Kingston and in the latter period in rural areas. In the latter period, moreover, inequality increased in urban areas particularly in Kingston so the impact of substantial growth was somewhat diminished. In the period , this was not the case: growth obviously gained the poor more, so inequality decreased as well. One possible picture that emerges from this and that squares quite well with other analyses of the Jamaican economy, cf. IMF (2000: Ch. III) is the following. Economic growth has taken place mainly in agriculture and related activities (where the bulk of the poor get their income). At the same time rapid migration to urban areas has led to increasing income disparities in towns which has skewed the urban income distribution and thereby prevented many urban dwellers from enjoying the fruits of the growth that has taken place. Consequently, although Jamaica in the past has higher than in other towns that in turn have higher prices than rural areas. Consequently, incomes in 17

19 enjoyed a relatively even distribution of income and thus modest poverty, it is quite clear that the future pattern of growth will determine the extent to which this is likely to continue. To investigate the possibilities, I ran two simulations for the growth pattern from 1998 to Focus was on the impact of growth on inequality and poverty. The first simulation assumes that GDP growth will average 5 percent with 6 percent in rural areas (i.e., agriculture) and the second assumes that GDP growth is 5 percent, but rural incomes grow by only 2 percent. With a rural population of around 60 percent, this implies that urban incomes grow by 3.5 in scenario 1 and 9.5 percent in scenario 2. I do not distinguish between Kingston and other urban areas. These assumptions are not based on informed guesses, but constructed to make the point that the pattern of growth matters. The results are in Table 9. Table 9 Simulation of Headcount Poverty Changes Two Scenarios Growth Effect Inequality Effect Residual Poverty Effect Scenario 1 Jamaica Urban Rural Scenario 2 Jamaica Urban Rural Sources and notes: See Table 8. Population is assumed to grow at one percent per annum. Assumptions are described in text Given that the majority of the poor reside in rural areas, a percentage decrease in rural poverty affects more people than in urban areas. Consequently, the poverty impact of these two growth patterns are dramatically different. While the scenario that assumes that agriculture grows fast will generate very rapid reduction in poverty, the urban growth scenario does little to change rural poverty. 18

20 Moreover and what this simulation exercise does not take into account rapid urban growth is likely to trigger rural-urban migration. It is clear from many studies that those that leave rural areas in search of the bright lights of the city are the young and able, so relatively rapid urban growth is likely to deplete rural areas of the resources available and may also add to urban poverty. In contrast, relatively rapid rural growth will slow down migration, and thereby help to reduce not only rural poverty (which is reduced through income increases), but also urban poverty through a slower inflow of migrants. The conclusion, then, is that Jamaica will remain an economy based on agriculture for some time to come. An efficient poverty-reduction strategy should then focus on this sector, for several reasons: agricultural growth reduces inequality and is thereby poverty efficient ; it decreases poverty by offering higher incomes directly to the poor without the detour through trickling down ; and it slows down rural-urban migration and may therefore help reduce urban poverty as well. 5.Concluding remarks This paper looks into the relations between economic growth and poverty reduction in Jamaica. A poverty profile is constructed, based on household budget survey data, and per capita growth is decomposed into pure growth effects and changes in inequality effects. I find that growth has contributed to poverty reduction mainly because there has not been any large increases in inequality. However, simulations also show that in order to for growth to continue to be efficient for poverty reduction, it is necessary to improve the prospects for rurally based, particularly small-scale agricultural growth. If not, the possibility of growth being anti-poor cannot be ignored. other towns are likely to be overestimated and poverty underestimated in what follows. 19

21 References Danielson, A. (1999) Poverty, Growth and Distributional Mobility in Jamaica: Implications for Public Policy in M. Mamalakis et al.: Distributional Mobility in Latin America: Evidence and Implications for Public Policy, Latin American Program Working Paper #242, Washington DC: Woodrow Wilson International Center for Scolars Danielson, A. (2001): When Do the Poor Benefit From Growth, and Why? Department of Economics Working Paper 2001: 12, Lund: University of Lund Deininger, K. and L. Squire (1996) A New Data Set Measuring Income Inequality World Bank Economic Review, Vol. 10 Dollar, D. and A. Kraay (2000): Growth Is Good For the Poor, Washington DC: World Bank, mimeo Duku Osei, P. (2002) A Critical Assessment of Jamaica s National Poverty Eradication Programme Journal of International Development, Vol. 14 ECLAC (1999): Review of Caribbean Economic and Social Performance in the 1980s and 1990s, Port-of-Spain: ECLAC/CDCC, mimeo Foster, J., J. Green and E. Thorbecke (1984) A Class of Decomposable Poverty Measures Econometrica, Vol. 52 IMF (2000): Jamaica: Selected Issues and Statistical Appendix, IMF Staff Country Report #00/19, Washington DC: IMF Mellor, J. (2000) Faster More Equitable Growth: The Relation Between Growth in Agriculture and Poverty Reduction CAER II Discussion Paper #70, Cambridge MA: HIID Ramadas, K., D. van der Mensbrugghe and Q Woodon (2002): SimSip_Poverty: Poverty and Inequality Comparisons Using Group Data, mimeo, Washington DC: World Bank 20

22 Ravallion, M. and G. Datt (1999) When is Growth Pro-Poor? Evidence from the Diverse Experiences of India s States Policy Research Working Paper #2263, Washington DC: World Bank STATIN (1998): Jamaica Survey of Living Conditions, Kingston: Statistical Institute Timmer, P. (1997) How Well Do the Poor Connect to Growth? CAER II Discussion Paper #178, Cambridge, MA: HIID WDI (2000): World Development Indicators on CD-ROM, Washington DC: World Bank White, H. and E. Anderson (2000): Growth Versus Distribution: Does the Pattern of Growth Matter? mimeo, Brighton, UK: IDS, University of Sussex World Bank (1996a): Caribbean Economic Overview, prepared for the CGCED meeting, Washington DC: World Bank 21