The anatomy of subjective poverty in Morocco

Transcription

1 The anatomy of subjective poverty in Morocco what is the role of local comparisons? Raphaël Cottin 1, Christophe Jalil Nordman 2 AFD International Conference Inequality and Social Cohesion Paris, December 6th, PSL, Université Paris Dauphine, UMR DIAL 2. IRD, UMR DIAL & Institut Français de Pondichéry 1 / 35

2 Outline 1. Introduction 2. Relevant literature 3. Country context 4. Data & Methodology 5. Descriptive statistics 6. Results 2 / 35

3 Introduction 3 / 35

4 poverty reduction strategies : based on idea that reduction in poverty will make people better-o Mission statements of aid agencies and international institutions make explicit link between poverty reduction and political stability but measurement of poverty problematic (monetary, absolute, relative, multidimensional?) `being poor' and `feeling poor' are two dierent things Old ideas in economics casting doubt on mechanic link between higher living standards and satisfaction Veblen : Conspicuous consumption Duesenberry : Hedonic threadmill are these ideas relevant in the context of developing countries? 4 / 35

5 This paper What we do : exploit nationally representative panel database of subjective poverty in LMIC context link it to living conditions examine the presence of `relative income' eects, at level of village/ neighbourhood and province look for channels explaining these comparison eects (local public goods, social capital) (link between intra-household inequality and subjective poverty) What we nd : comparison eects : positive at neighbourhood level, negative at province level mostly in cities not explained by public goods or social capital disappear when using xed eects 5 / 35

6 Literature 6 / 35

7 Subjective well-being and comparison eects `Happiness Economics' has treated Subjective-Well Being (SWB) as economic variable [Frey and Stutzer, 2002, Layard, 2005] taken as empirical counterpart to utility (experienced utility ) [Kahneman and Krueger, 2006] Stylized fact : SWB is relative : internal reference point : adaptation and expectations [Di Tella et al., 2010, Clark et al., 2008a] external reference point : comparison eects [Luttmer, 2005, Clark and Senik, 2010] Possible explanation for the `Easterlin Paradox' [Easterlin, 1995, Clark et al., 2008b] Comparison may depend of the geographical scale of the comparison (local vs. distant neighbours) [Brodeur and Flèche, 2017, Ifcher et al., 2018] 7 / 35

8 SWB in LMIC In developing countries context : mostly life satisfaction" or subjective poverty Comparison eects, relative utility : goes against poverty reduction as only focus poverty reduction may not be enough to increase welfare if what matters is relative position in society Most papers do not nd any comparison eect except for richest [Tao and Chiu, 2009, Bookwalter and Dalenberg, 2010, Castilla, 2012] Some nd positive comparison eect [Senik, 2004, Ravallion and Lokshin, 2010, Lentz, 2017] consistent with literature on informal risk-sharing [Fafchamps and Lund, 2003, Fafchamps and Gubert, 2007] some exceptions [Dedehouanou et al., 2013] Subjective poverty also in the `dimensions of poverty' debate [Razandrakoto and Roubaud, 2005, Herrera et al., 2006, Dat et al., 2015] 8 / 35

9 Context 9 / 35

10 Figure Monetary poverty, / 35

11 11 / 35

12 Introduction Literature Context Methodology Descriptive statistics Results 12 / 35

13 13 / 35

14 Morocco : context important reductions in (monetary) poverty since end of 1990s high inequalities, esp. between urban and rural political stability, but frequent protests esp. since 2011 (Arab Spring) mostly in cities, not correlated to (monetary) poverty grievances often linked to public services (hospital, university...) political will to put in place targeted social safety nets (in education, health, housing) targeting based on geography or monetary poverty 14 / 35

15 Methodology 15 / 35

16 Data Nationally representative panel household survey, 3 waves (2012, 2013, 2015) subjective poverty question : according to you, your household belongs to the poorest households 2. rather poor households 3. average households 4. rather well-o households 5. well-o or rich households one respondent per household also Minimum Income Question (MIQ), feeling of deterioration in past 12 months, main sources of concern for future 16 / 35

17 Role of relative income S it = βx it + γy it + δy rel it + ɛ it Y is own income, Y rel is relative income δ < 0 : when others income go up, I feel poorer/ less happy (case in dev'd countries) δ > 0 : I benet from other's income construct reference income based on geography using sample design dene Y rel as leave-out median of income : at PU level (village, neighbourhood) : approx. 30 households in sample at province level (40 to 260 hh) 17 / 35

18 Controls estimation is done at the household level (provisionally) Characteristics of household head : age, age squared, education level, marital status Household structure : number of children, youth, working age and elderly health : number of chronically and acute sick in past 4 weeks ; number of underweight/overweight (BMI) `social capital' : dummies for participation in community- level activities and membership in collective organisation wealth index based on household durables `Minimum Income Question' (MIQ) : How much would a household similar to yours need per month in order not to be considered as poor 18 / 35

19 Methodology : ordinal variables in panel setting subjective poverty is an ordinal variable we can assume that `very poor' is worse than `poor' and that `poor' is worse than `middle class' but we cannot assume that there is the same distance from `very poor' to `poor' as from `poor' to `middle class' (this last assumption is often made in SWB research) use ordinal models (non-linear) doesn't lend itself easily to xed-eects modeling 19 / 35

20 Ordinal models and the incidental parameters problem (1) What we want to estimate (ideally) S it = βx it + α i + ɛ it (1) with : Sit a latent variable (unobserved) linked to subjective poverty X it a vector of covariates α i is individual xed eect (personality...) ɛ i t follows a logistic distribution with mean zero and variance 1 (wlog) Observation rule S = 1 (`very poor') when S µ 1 S = 2 (`somewhat poor') when µ 1 S µ 2... S = 4 (`well-o') when S µ 3 If J categories, J 1 thresholds, estimated jointly with β parameters through maximum likelihood 20 / 35

21 Ordinal models and the incidental parameters problem (2) What to do with unobserved heterogeneity term α i? make assumptions about joint distribution of α i and X it => random eect ordinal logit treat α i as parameter to be estimated (xed eect) probability of observing outcome k for indiv i at time t : P (S it = k X it, α i ) = Λ (µ i,k+1 βx it α i ) Λ (µ i,k βx it α i ) only µ ik α i identied : not possible to distinguish thresholds from xed eect with xed T, only small number of observations used to estimate α i, but total number of observations is NT α i cannot be estimated consistently the inconsistency `spills over' to estimation of β 21 / 35

22 Ways around the incidental parameters problem 1. use random eects and cross ngers renement : use `correlated random eect' model (parametrize α i as function of the time- varying covariates) 2. dichotomize the dependent variable into dummy `poor'/ `nonpoor' use linear probability model with xed eects use conditional logit 3. combine conditional logit estimates from all possible dichotomizations into one through minimum distance estimation [Das and Van Soest, 1999] through `Blow-up and Cluster' (SUR estimation on duplicated data) [Baetschmann et al., 2015] 22 / 35

23 Descriptive statistics 23 / 35

24 Figure Subjective poverty by consumption quintile 24 / 35

25 Table Subjective poverty by year (national) Total very poor rather poor middle class well-o/rich Total (7 853) (7 370) (6 976) (22 199) percentages, frequencies in parentheses 25 / 35

26 Table Transition between subjective poverty status, \ 2013 very poor poor middle class well-o very poor poor middle class well-o Table Transition between subjective poverty status, very poor poor middle class well-o very poor poor middle class well-o / 35

27 Table household characteristics by perceived poverty status very poor rather poor middle class well-o/rich Total household size (1.961) (2.178) (2.238) (2.636) (2.197) head education : none (0.471) (0.492) (0.496) (0.495) (0.499) head education : primary (0.423) (0.433) (0.419) (0.358) (0.423) head education : lower secondary (0.239) (0.278) (0.313) (0.274) (0.290) head education : upper secondary (0.162) (0.219) (0.323) (0.310) (0.273) head education : higer ed (0.102) (0.149) (0.312) (0.423) (0.254) age of hh head (11.01) (11.29) (11.08) (11.92) (11.23) # of employed in hh (0.784) (0.872) (0.889) (0.982) (0.876) # of elderly (>65yo) (0.448) (0.502) (0.510) (0.594) (0.503) school enrollment rate (0.297) (0.266) (0.228) (0.214) (0.253) employment rate (0.242) (0.220) (0.221) (0.237) (0.225) # inactive (1.240) (1.255) (1.340) (1.376) (1.302) log hh income (0.742) (0.692) (0.761) (1.173) (0.849) log expenditure/pers (0.521) (0.518) (0.628) (0.862) (0.633) food share (14.38) (14.00) (14.23) (15.72) (14.48) 27 / 35

28 Results 28 / 35

29 Table Baseline results dep. var : subjective poverty (1=well-o, 4=very poor) ologit-re LPM-fe (1) (2) (3) (4) (5) (6) total rural urban total rural urban female hh head (0/1) age of hh head square age of hh head marital status (b. single) : - married widowed divorced and others hh head education (b. none) - primary lower secondary upper secondary higher ed # children (<6 y) # women # youth (6-15 y) # working-age (16-65 y) # elderly (>65 y) urban (0/1) ref. ref. # of chronically ill # temporary ill community participation (0/1) member in coll. org (0/1) total hh revenue (ln) durable and housing index MIQ (ln) Constant region Yes Yes Yes No No No year Yes Yes Yes Yes Yes Yes N vce cluster cluster cluster cluster cluster cluster p < 0.1, p < 0.05, p < / 35

30 Table Relative income eects on perceived poverty dep. var : subjective poverty (1=well-o, 4=very poor) ologit-re LPM-FE (1) (2) (3) (4) (5) (6) b/se b/se b/se b/se b/se b/se total hh revenue (ln) (0.031) (0.031) (0.031) (0.009) (0.009) (0.009) durable and housing index (0.036) (0.036) (0.036) (0.014) (0.014) (0.014) MIQ (ln) (0.042) (0.041) (0.042) (0.012) (0.012) (0.012) cluster median (log) income (0.047) (0.053) (0.020) (0.025) province median (log) income (0.076) (0.087) (0.026) (0.033) Constant (0.270) (0.316) (0.318) cut cut cut sigma2_u region Yes Yes Yes No No No year Yes Yes Yes Yes Yes Yes N vce cluster cluster cluster cluster cluster cluster 30 / 35

31 Table comparison eects by urban/rural dep. var : subjective poverty (1=well-o, 4=very poor) rural urban (1) (2) (3) (4) ologit-re LPM-FE ologit-re LPM-FE b/se b/se b/se b/se total hh revenue (ln) (0.037) (0.009) (0.042) (0.010) cluster median (log) income (0.089) (0.027) (0.069) (0.017) province median (log) income (0.118) (0.034) (0.115) (0.032) Constant (0.288) (0.356) cut cut cut sigma2_u region Yes No Yes No year Yes Yes Yes Yes N vce cluster cluster cluster cluster p < 0.1, p < 0.05, p < / 35

32 Table Channels (1) : local public goods dep. var : subjective poverty (1=well-o, 4=very poor) total rural urban (1) (2) (3) (4) (5) (6) ologit-re LPM-FE ologit-re LPM-FE est5 est6 b/se b/se b/se b/se b/se b/se total hh revenue (ln) (0.028) (0.007) (0.038) (0.009) (0.041) (0.010) cluster median (log) income (0.053) (0.015) (0.090) (0.027) (0.069) (0.017) province median (log) income (0.080) (0.022) (0.120) (0.034) (0.114) (0.032) middle schools (density) (1.545) (2.587) (2.703) (4.971) (1.737) (2.945) postal oces (density) (0.571) (1.178) (0.806) (1.673) (0.773) (1.610) specialized health centres (density) (7.561) (19.577) (11.786) (30.417) (7.682) (22.861) generalist hospitals (beds) (0.000) (0.000) (0.000) (0.001) (0.000) (0.001) specialist hospitals (beds) (0.000) (0.000) (0.000) (0.001) (0.000) (0.000) Constant (0.302) (0.387) (0.387) cut cut cut sigma2_u region Yes Yes Yes No No No year Yes Yes Yes Yes Yes Yes N vce cluster cluster cluster cluster cluster cluster p < 0.1, p < 0.05, p < / 35

33 Table Channels (2) : social capital dep. var : subjective poverty (1=well-o, 4=very poor) ologit-re LPM-FE (1) (2) (3) (4) (5) (6) subjpov subjpov subjpov povsubj2 povsubj2 povsubj2 b/se b/se b/se b/se b/se b/se total hh revenue (ln) (0.028) (0.028) (0.028) (0.007) (0.007) (0.007) cluster median (log) income (0.045) (0.056) (0.013) (0.015) orgaxmedlnrevup (0.113) (0.144) (0.024) (0.033) province median (log) income (0.065) (0.081) (0.018) (0.022) orgaxmedlnrevprov (0.189) (0.243) (0.041) (0.056) Constant (0.179) (0.222) (0.222) cut cut cut sigma2_u region Yes Yes Yes No No No year Yes Yes Yes Yes Yes Yes N vce cluster cluster cluster cluster cluster cluster p < 0.1, p < 0.05, p < / 35

34 Table Fixed Eects ordered logit (1) : Minimum Distance dep. var : subjective poverty (1=well-o, 4=very poor) (1) (2) (3) (4) (5) (6) age of hh head (0.0214) (0.0210) (0.0213) (0.0216) (0.0215) (0.0215) square age of hh head ( ) ( ) ( ) ( ) ( ) ( ) # children (<6 y) (0.0542) (0.0545) (0.0545) (0.0541) (0.0542) (0.0542) # women (0.0603) (0.0606) (0.0606) (0.0636) (0.0637) (0.0637) # youth (6-15 y) (0.0537) (0.0539) (0.0539) (0.0547) (0.0548) (0.0548) # working-age (16-65 y) (0.0420) (0.0422) (0.0422) (0.0434) (0.0435) (0.0435) # elderly (>65 y) (0.0809) (0.0810) (0.0810) (0.0838) (0.0839) (0.0839) cluster median (log) income (0.0624) (0.0858) (0.0676) (0.0892) province median (log) income (0.0825) (0.113) (0.0879) (0.116) total hh revenue (ln) (0.0352) (0.0352) (0.0355) Observations Standard errors in parentheses p < 0.1, p < 0.05, p < / 35

35 Table Fixed Eects ordered logit (2) : Blow Up and Cluster dep. var : subjective poverty (1=well-o, 4=very poor) (1) (2) (3) (4) (5) (6) subjpov age of hh head (0.0212) (0.0209) (0.0211) (0.0214) (0.0214) (0.0214) square age of hh head ( ) ( ) ( ) ( ) ( ) ( ) # children (<6 y) (0.0543) (0.0545) (0.0545) (0.0547) (0.0549) (0.0549) # women (0.0605) (0.0606) (0.0606) (0.0642) (0.0643) (0.0643) # youth (6-15 y) (0.0538) (0.0539) (0.0539) (0.0557) (0.0558) (0.0558) # working-age (16-65 y) (0.0418) (0.0420) (0.0419) (0.0438) (0.0440) (0.0440) # elderly (>65 y) (0.0812) (0.0811) (0.0812) (0.0846) (0.0847) (0.0847) cluster median (log) income (0.0627) (0.0861) (0.0679) (0.0896) province median (log) income (0.0830) (0.114) (0.0883) (0.116) total hh revenue (ln) (0.0356) (0.0356) (0.0359) N N_clust Standard errors in parentheses p < 0.1, p < 0.05, p < / 35

36 Références References I G. Baetschmann, K. E. Staub, and R. Winkelmann. Consistent estimation of the xed eects ordered logit model. Journal of the Royal Statistical Society : Series A (Statistics in Society), 178 (3) :685703, J. T. Bookwalter and D. R. Dalenberg. Relative to what or whom? the importance of norms and relative standing to well-being in south africa. World Development, 38(3) :345355, A. Brodeur and S. Flèche. Neighbors' income, public goods, and well-being. Review of Income and Wealth, C. Castilla. Subjective well-being and reference-dependence : Insights from mexico. The Journal of Economic Inequality, 10 (2) :219238, / 6

37 Références References II A. E. Clark and C. Senik. Who compares to whom? the anatomy of income comparisons in europe. The Economic Journal, 120 (544) :573594, A. E. Clark, E. Diener, Y. Georgellis, and R. E. Lucas. Lags and leads in life satisfaction : A test of the baseline hypothesis. The Economic Journal, 118(529), 2008a. A. E. Clark, P. Frijters, and M. A. Shields. Relative income, happiness, and utility : An explanation for the easterlin paradox and other puzzles. Journal of Economic literature, 46(1) : 95144, 2008b. M. Das and A. Van Soest. A panel data model for subjective information on household income growth. Journal of Economic Behavior & Organization, 40(4) :409426, / 6

38 Références References III V. H. Dat, P. M. Thai, L. Pasquier-Doumer, and N. Thang. Weighting deprivations using subjective well-being. Technical report, Working Paper 142, Young Lives, Oxford Department of International Development, Oxford, S. F. Dedehouanou, J. Swinnen, and M. Maertens. Does contracting make farmers happy? evidence from s enegal. Review of Income and Wealth, 59 :S138S160, R. Di Tella, J. Haisken-De New, and R. MacCulloch. Happiness adaptation to income and to status in an individual panel. Journal of Economic Behavior & Organization, 76(3) :834852, R. A. Easterlin. Will raising the incomes of all increase the happiness of all? Journal of Economic Behavior & Organization, 27(1) :3547, / 6

39 Références References IV M. Fafchamps and F. Gubert. The formation of risk sharing networks. Journal of development Economics, 83(2) :326350, M. Fafchamps and S. Lund. Risk-sharing networks in rural philippines. Journal of development Economics, 71(2) :261287, B. S. Frey and A. Stutzer. Happiness and Economics : How the Economy and Institutions Aect Human Well-Being. Princeton University Press, J. Herrera, M. Razandrakoto, and F. Roubaud. The determinants of subjective poverty : A comparative analysis in madagascar and peru. Working Papers DT/2006/01, DIAL (Développement, Institutions et Mondialisation), URL https: //EconPapers.repec.org/RePEc:dia:wpaper:dt / 6

40 Références References V J. Ifcher, H. Zarghamee, and C. Graham. Local neighbors as positives, regional neighbors as negatives : Competing channels in the relationship between others' income, health, and happiness. Journal of health economics, 57 :263276, D. Kahneman and A. B. Krueger. Developments in the measurement of subjective well-being. Journal of Economic perspectives, 20(1) :324, R. Layard. Happiness : Lessons from a new science. Penguin UK, E. C. Lentz. Keeping up with the neighbors? reference groups in ghana. Economic Development and Cultural Change, 66(1) : 91112, E. F. Luttmer. Neighbors as negatives : Relative earnings and well-being. The Quarterly journal of economics, 120(3) : , / 6

41 Références References VI M. Ravallion and M. Lokshin. Who cares about relative deprivation? Journal of Economic Behavior and Organization, 2 (73) :171185, M. Razandrakoto and F. Roubaud. Les multiples facettes de la pauvreté dans un pays en développement. le cas de la capitale malgache. Economie et statistique, 383(1) :131155, C. Senik. When information dominates comparison : Learning from russian subjective panel data. Journal of Public Economics, 88 (9) : , H.-L. Tao and S.-Y. Chiu. The eects of relative income and absolute income on happiness. Review of Development Economics, 13(1) :164174, doi : /j x. URL x. 6 / 6