A Risk and Vulnerability Assessment

Transcription

1 a This research has been possible by the financial support of the International Development Research Center (IDRC), provided through the Community Based Monitoring System (CBMS) initiative of the Partnership for Economic Policy (PEP) network. A Risk and Vulnerability Assessment Using CBMS a Werner Hernani-Limarino Manila - June, 2016 Fundación ARU

2 Table of contents 1. Motivation 2. Objectives 3. Measuring Vulnerability Intuition: A Disney economy Vulnerability to future poverty 4. An application to Bolivia and Concepcion 5. Risks and Management Stategies Sources and Impact of Risks Social Risk Management Strategies 6. Conclusions 2

3 Motivation

4 How should we do health policies? Figure 1: status Population types according to their current and future health 4

5 How should we do health policies? Public health interventions typically have three elements: 1. treatment for those who are sick 2. preventive measures to reduce the risks of contracting (or re-contracting) the disease 3. programs to halt the transmission of the disease 5

6 How should we do social policies? Figure 2: Population types according to their current poverty and future vulnerability 6

7 How should we do social policies? Social interventions typically have three elements: 1. poverty alleviation programs that mitigate (treat) the adverse effects of poverty (i.e. social assistance) 2. poverty prevention programs that reduce the risks of becoming poor (i.e. social protection) 3. programs to prevent the inter-generational transmission of poverty (e.g. CCTs) 7

8 Objectives

9 Objectives Our main objective is to understand the gaps in social protection in Bolivia. In order to do this we: 1. measure the ex-ante vulnerability to future poverty 9

10 Objectives Our main objective is to understand the gaps in social protection in Bolivia. In order to do this we: 1. measure the ex-ante vulnerability to future poverty 2. explore the sources of risk, their correlation and their impact on vulnerability to future poverty 9

11 Objectives Our main objective is to understand the gaps in social protection in Bolivia. In order to do this we: 1. measure the ex-ante vulnerability to future poverty 2. explore the sources of risk, their correlation and their impact on vulnerability to future poverty 3. analyze the importance of different social risk management strategies in protecting households from different risks 9

12 Research questions What is the (average) level and intensity of vulnerability to future poverty of a given population and different population subgroups? e.g. Are self-employed (a.k.a entrepreneurs) more vulnerable than salaried workers 10

13 Research questions What is the (average) level and intensity of vulnerability to future poverty of a given population and different population subgroups? e.g. Are self-employed (a.k.a entrepreneurs) more vulnerable than salaried workers What are the sources of risk, their correlation and their (average) impact on the vulnerability of households? e.g. Are health shock more common than economic shocks, are they correlated, which one affect more the households disposable income 10

14 Research questions What is the (average) level and intensity of vulnerability to future poverty of a given population and different population subgroups? e.g. Are self-employed (a.k.a entrepreneurs) more vulnerable than salaried workers What are the sources of risk, their correlation and their (average) impact on the vulnerability of households? e.g. Are health shock more common than economic shocks, are they correlated, which one affect more the households disposable income What is the importance of the different social risk management mechanisms? e.g. Are formal credit markets better than informal social insurance?, Are social protection policies enough? 10

15 Measuring Vulnerability

16 A Disney economy Table 1: Disney s Economy Households Consum. P(c, 600) g(x t+1, Kt) Value Name Mickey Unexpected death Minnie Expected death Goofy No shock Donald Child birth expenses Daisy Gift from marriage 12

17 A Disney s economy Table 2: Mice s Vulnerability to Poverty TFP Scenario h(ϵ) Consum. Et[P(c t+1, 600)] Et[P(c t+1, 600)] = ( ) 5 =

18 A Disney s economy Table 3: Goofy s Vulnerability to Poverty TFP Scenario h(ϵ) Consum. Et[P(c t+1, 600)] Et[P(c t+1, 600)] = ( ) 5 =

19 A Disney s economy Table 4: Ducks Vulnerability to Poverty TFP Scenario h(ϵ) Consum. P(c t+1, 600) Et[P(c t+1, 600)] = ( ) 5 =

20 Vulnerability: Definitions Vulnerability to Poverty. Ex-ante household s probability of being poor. Includes both: probability of remain into poverty (if currently poor) probability of move into poverty (if currently non-poor) Formally, V h,t = E t [P(c h,t+1, z)] where V h,t is household h probability of being poor in period t + 1 -calculated at time t, c h,t+1 consumption of household h at time t + 1, z poverty line. 16

21 Measurement of vulnerability V h,t = E t [P(c h,t+1, z)] = P(c h,t+1, z)df(c h,t+1 ) where: df(c h,t+1 ) is the distribution of future consumption -i.e. the set of all possible consumptions Challenge: How to estimate the distribution of future consumption? Case 1: Static setting Case 2: Dynamic setting 17

22 Case 1: Static setting Consider the simplest static setting: c ht = y ht c h per-capita household consumption, and y h per-capita household income, there is no need for time subscripts. 18

23 Case 1: Static setting By definition, per-capita household income is given by : y h = 1 (y l,i + y n,i ) n h i h y l labor income, y n non-labor income, and n number of household members, Objective What are the determinants of per-capita household income? What are housholds income generating capabilities? 19

24 Case 1: Static setting Per-capita household labor income, can be decomposed into: y lh = m h n h p h m h e h p h i h y li e h m h the number of household adults - i.e. [15, 65] year olds; p h the number of household adults that participate into the labor market, e h the number of household adults that participate into de labor market and actually get a job. the labor income per worker. 20

25 Case 1: Static setting Per-capita household labor income, can be decomposed into: y lh = m h p h e h y li n h m h p h e h }{{} d 1 h }{{} p h }{{} e h i h }{{} ȳ lh 21

26 Case 1: Static setting Per-capita household labor income, can be decomposed into: y lh = m h p h e h y li n h m h p h e h }{{} d 1 h }{{} p h }{{} e h i h }{{} ȳ lh y lh = 1 d p he h ȳ lh n m p m e p the dependency rate (i.e. the ratio of household members to the number of adults); the participation rate (i.e. the ratio of household members participating into the labor market to the number of adults); the employment rate (i.e. the ratio of employed to participating household members), and ȳ lh the labor income per worker. 21

27 Case 1: Static setting household labor income per-worker can be decompose into: where: ȳ l,h = h h w h }{{} e α+β x h +ϵ h h h the household average number of working hours per-worker, w h the household average of hourly wages per-worker, which can be further decomposed into: ln( w h ) = α + β x h + ϵ h x h average observable productive endowments (mainly age and education), β prices of observable productive endowments, α overall (economy wide) wage level, and ϵ a residual related to unobserved heterogeneity component (labor income shocks) and measurement error. 22

28 Case 1: Static setting Therefore, y l,h = 1 d h p h e h ȳ lh 23

29 Case 1: Static setting Therefore, y l,h = 1 d h p h e h ȳ lh = 1 d h p h e h h i e α+β x h+ϵ h 23

30 Case 1: Static setting Therefore, y l,h = 1 d h p h e h ȳ lh = 1 d h p h e h h i e α+β x h+ϵ h Aplying logs: ln(y lh ) = ln(d h ) + ln(p h ) + ln(e h ) + ln( h h ) + α + β x h + ϵ h 23

31 Case 1: Static setting Non-labor income can be decomposed into:. y nt = p t + T ht + T gt + T st p t asset income, including income from financial assets (i.e. interests and dividends) and income from real wealth (rental income), T ht private transfers (inter-family assistance and remittances) T gt public transfers (non-contributory pensions and conditional cash transfers) T st social security transfers 24

32 Case 1: Static income specification So that, ln(y nh ) = θ p 1(p t 0) + θ Th 1(T ht 0) + θ Tg 1(T gt 0) +θ Ts 1(T st 0) + η where: 1() indicator function that defines dummies for household that receive asset income private transfers, public transfers, and social security transfers. θ p, θ Th, θ Tg and θ Ts are average amounts of money, and η measurement error 25

33 Case 1: Static income specification So that, our final specification of household income generating capabilities is given by: ln(y h ) = ln(d h ) + ln(p h ) + ln(e h ) + ln( h h ) + α + β x h + ϵ h +θ p 1(p t 0) + θ Th 1(T ht 0) +θ Tg 1(T gt 0) + θ Ts 1(T st 0) + η 26

34 Rural gradient Figure 3: % of rural households by per-capita income percentile 27

35 Gradient of the hh dependency ratio Figure 4: dependency ratio by per-capita income percentile 28

36 Gradient of the hh employment rate Figure 5: employment rate by per-capita income percentile 29

37 Hours gradient Figure 6: percentile average hours of work among working adults by per-capita income 30

38 Age gradient Figure 7: average age among working adult by per-capita income percentile 31

39 Schooling gradient Figure 8: percentile Average schooling among working adults by per-capita income 32

40 Female gradient Figure 9: % female among working adults by per-capita income percentile 33

41 Property income gradient Figure 10: % of hh with property income by per-capita income percentile 34

42 HH transfers gradient Figure 11: % of hh with family trasnfers by per-capita income percentile 35

43 Government transfers gradient Figure 12: % hh with government transfers by per-capita income percentile 36

44 Social security income gradient Figure 13: % of hh with s.s. income by per-capita income percentile 37

45 Other income gradient Figure 14: % of hh with other income by per-capita income percentile 38

46 Income shocks gradient Figure 15: % of rural households by per-capita income percentile 39

47 Vulnerability to future poverty Table 5: Income Generating Determinants Base Rural premium ln(d) *** *** ln(e) 0.74 *** 0.21 *** ln(h) 0.27 *** α 5.60 *** *** β age 0.02 *** *** β age 2-2.1e-4 *** 0.7e-4 ** β school 0.04 *** 0.04 *** β female *** 0.27 *** ρ *** *** θ Th *** θ Tg ** *** θ Ts *** *** θ Other ** 2 40

48 Dynamic setting Now consider, a more complex and more dynamic setting: c h,t + [a h,t+1 a h,t ] + [e h,t+1 e h,t ] = y h,t c h,t + [a h,t+1 a h,t ] + [e h,t+1 e h,t ] = y ( γ, Z h,t ) where c t consumption expenditure a t+1 net asset position of financial wealth at end of t e t+1 net asset position of real wealth at end of t y h,t household income Z h,t income generating capabilities 41

49 Dynamic setting Consumption decision c t = f(y ( γ, Z h,t ), a t, e t, h(ϵ t )) c t+1 = f(y ( γ, Z h,t+1 ), a t+1, e t+1, h(ϵ t+1 )) where Z t+1 productive characteristics, K t wealth, ϵ t+1 shocks (positive and negative). 42

50 An application to Bolivia and Concepcion

51 Vulnerability to future poverty Figure 16: Cumulative distribution of vulnerability to future poverty 44

52 Are the non-poor non-vulnerable to future poverty? Figure 17: Cumulative distribution of vulnerability to future poverty by poverty status 45

53 Are rural areas more vulnerable? Figure 18: Cumulative distribution of vulnerability to future poverty by area 46

54 Are female more vulnerable? Figure 19: Cumulative distribution of vulnerability to future poverty by sex 47

55 Are the youth more vulnerable? Figure 20: Cumulative distribution of vulnerability to future poverty by groups of age 48

56 Are unskilled more vulnerable? Figure 21: level Cumulative distribution of vulnerability to future poverty by skill 49

57 Are unemployed more vulnerable? Figure 22: Cumulative distribution of vulnerability to future poverty by employment status 50

58 Are self-employed (a.k.a entrepreneurs ) more vulnerable? Figure 23: Cumulative distribution of vulnerability to future poverty by type of employment 51

59 Vulnerability to future poverty Figure 24: Poverty and Vulnerability. Concepción,

60 Vulnerability to future poverty Table 6: Poverty and Vulnerability. Concepción, 2014 TOTAL Non-poor Moderate poor Extreme poor Extreme vulnerable Moderate vulnerable Low vulnerable TOTAL 53

61 Risks and Management Stategies

62 Shocks Pregnancy, birth. Death. Divorce, separation. Socioeconomic standing is captured by different measures unemployment (fired, quit), individual labor earnings, and occupational status, etc. 55

63 Simulating Shocks Poverty and inequality measures (W) are a function of the distribution of income - which is determined by the equation above, we can relate the changes in poverty and inequality to changes in each of the components. ( m W = W n, p m, e ) p, ȳl i, { yn ik } K k=1 56

64 Risks:Sources and Types The sources of risk: natural (e.g. floods) or man-made (health, life cycle, economic and social). their types according to: 1. correlation: idiosyncratic - when they are not correlated among individuals or households in a community, or covariant - when they are correlated among individuals or households; 2. dependence isolated or bunched 3. frequency: sporadic or infrequent or recurrent or frequent. 4. intensity: non-catastrophic - those who have low or moderate welfare effects, and catastrophic - those who have severe welfare effects. 57

65 Risks: Average effect on disposable income What is the risks average effect on disposable income? Two alternative ways of measuring where 1. ˆϵ h,t = j γ js h,t,j + u h,t 2. c h,t = j γjs h,t,j + u h,t ˆϵ are the estimated shock values, c h,t are the self-declared changes in consumption due to shocks, γ j magnitud of shock S j. 58

66 Sources and Impact of Risks Table 7: Sources and Impact of Risks. Concepcion, 2014 Source Correlation Dependence Frequency Intensity Health (.,0,+,0) isolated frequent very high Economic (.,.,0,+) bunched often high Life cycle (.,.,.,0) isolated often medium Social (.,.,.,.) bunched seldom low 59

67 Assessing Social Risk Management Strategies Table 8: 2014 Assessing Ex-ante Health Risk Management Strategies. Concepcion, Strategy Effectiveness (Loss in dis.inc.) Mitigation of socioeconomic shocks Private health insurance (formal w.) -.05 Public health insurance -.13 Coping with health shocks Private formal loans.-15 Private informal loans -.25** Public (CCT and UP).05 60

68 Risk Management Strategies: Types Risk prevention. Ex-ante reduction of the occurrence Private formal. e.g human and physical capital accumulation. Private informal. e.g. community work to improve public infrastructure. Public infrastructure programs. e.g. better health and education facilities and inputs, better infraestructure. Risk mitigation. Ex-ante reduction of the impact Private formal. e.g savings in banks. Private informal. e.g. savings in cattle, assets or improving dwelling (better colateral). Public insurance programs. e.g. state contingent risk sharing mechanisms. Risk copying. Ex-post relief of the impact Private formal. e.g borrowing from banks. Informal borrowing. e.g. borrowing from family and neighbors. Transfer programs. e.g. CCT and non-contributory pensions. 61

69 Risk Management Strategies What is the efficiency of risks management strategies? An counterfactual comparison under unconfoundedness where θ s = E[ˆϵ h,t S h,t,j = 1, RMS s = 1, X h,t = x, K h,t 1 = k] E[ˆϵ h,t S h,t,j = 0, RMS s = 1, X h,t = x, K h,t 1 = k] ˆθ s is the estimated efficiency of risk management strategy s, RMS s are s alternative risk management strategies. 62

70 Conclusions

71 Contribution This paper... 64

72 Contribution This paper... proposes a new methodological approach to: the measurement of vulnerability to future poverty, the measurement of the impact of different types of risks, the measurement of the importance of alternative social risk management strategies. 64

73 Contribution This paper... proposes a new methodological approach to: the measurement of vulnerability to future poverty, the measurement of the impact of different types of risks, the measurement of the importance of alternative social risk management strategies. applies our method to the analysis of vulnerability to future poverty in the municipality of Concepcion using CBMS data. 64

74 Conclusions I: Vulnerability (Obvious) Poverty status of today might not be the same as the poverty status of tomorrow 21% of the hh are structurally non-poor (not poor un the past, unlikely to be poor in the future) 11% of the hh are poor but not vulnerable (poor because they got unlucky. i.e. suffered an idiosyncratic negative shock) 28% of the hh are non-poor but vulnerable (non-poor because they got lucky i.e. suffered an idiosyncratic positive shock) 40% of the hh are structurally poor (poor in the past, very likely to be poor in the future) Therefore, social policies should recognize differences in objectives and design between social assistance (safety nets) and social protection (safety ropes). 65

75 Conclusions II: Risks Regarding different sources of risks in our case-study: Natural risks. Can not say much. Need to improve the questionary for better analysis. Health risks. Very likely and very high impact. Economic risks. Not unlikely to occur during economic expansions and high impact. Lyfe cycle risks. Not unlikely to occur and medium impact. Social risks. Unlikely to occur during economic expansions and low impact. 66

76 Conclusions III: Social Risk Management Don t overlook the importance of risk prevention mechanisms Don t overlook the importance of privately provided formal and informal risk coping mechanisms e.g. ceteris paribus, the occurrence of socioeconomic shocks is statistically higher for salaried workers than for self-employment workers. Government cash transfers are not always effective risk protection or coping mechanisms e.g. ceteris paribus, the occurrence and average effect on disposable income are not statistically different between recipients and non-recipients of BJP who suffer a health shock 67

77 Policy Implications Concepcion s case-study illustrates the need to: Move from the conventional ex-post measurement of past poverty to the more useful ex-ante assessment of future poverty i.e. from poverty to vulnerability to future poverty Complement social assistance policies (safety nets) with (TRUE) social protection policies (safety ropes). Make some social transfers contingent of the occurrence of certain shocks (e.g. non-contributory pensions), Improve prevention strategies, in particular for natural and health risks. Better dwelling and public infrastructure. Better health facilities and inputs, 68

78 Questions? 69

79 References I 70