Executive summary

Annex 3.6: Estimation of the marginal propensity to consume

Poverty measures in Panama are based on consumption. Given that the Panama CCT program supposes a monetary benefit, to estimate the `contrafactual’ or `post-transfer’ consumption distribution, it is necessary to know the effect caused by the increase in the household income into its consumption.
To this aim, we estimate the total household consumption as a function of its income. An easy way to compute this relation is to assume a constant marginal propensity along the entire income distribution⁷⁹. This idea can be resumed in the following equation:
(4)
where and means the total household consumption and income respectively, represents the vector of observed characteristics that affects the consumption, represents the vector of non observed characteristics (errors) and is the marginal propensity to consume.
It is known that in equation (4) the income might be correlated with the error term (if the error term contains omitted variables that are correlated with the variables included, if contains measurement errors, or if are determined jointly with). This causes inconsistent OLS estimation.

To go beyond this problem, we also propose to estimate equation 4 using the method of Instrumental Variables (IV). This technique is the standard prescription for correcting such cases and it gives:

(5)

where is called the instrument for and it have the following two properties: (i) it is correlated with , , and (ii) it is uncorrelated with , .

Following Deaton (1997), we used as instrumental variable the average income of the village where the household is placed estimated without the income of the household (expresses in log.). Hence, z is determinate by the following expression:
con h = 1,…,75 (7)
Where the subindex i means household, the subindex h means the village where the household is placed and, N_h means the household population of each village.
Table A.3.6.1 shows four specifications using both IV and OLS methods. Model 1 is the simplest specification where total consumption (in log.) is estimated as a function of its total income (in log) and the household size. Model 2 adds regional controls. Model 3 included controls for demographics characteristics. Finally, model 4 incorporates educational controls.
Performing a Hausman test we confirm that the OLS estimation results to be inconsistent under these models (Table A.3.6.2). As we mention before, the income coefficient denotes the marginal propensity to consume. This coefficient results to be statically significant in all IV models. Moreover, these estimations show that the value of the marginal propensity in Panama is between [0.7460, 0.8243].⁸⁰

Table A.3.6.1: Estimation of the Marginal Propensity to Consume

Source: Own estimation based on ENV 2003 data. Note: OLS means Ordinay Least Square estimation, IV means Instrumental Variables estimation

Table A.3.6.2: Hausman Test

Source: Own estimation based on ENV 2003 data.

1 The extreme poverty line, which equals the cost of an adequate minimum food basked, for 1997 was estimated at B.\519 and for 2003 at B.\534.

2 B.\ 1= US$1.

3 More precisely, a program that does not affect the poverty rate but that reduces the poverty gap significantly by bringing those far below the poverty line closer to it, may be preferable to a program that reduces the poverty rate, but leaves many of the neediest unaffected (e.g., by lifting those close to the poverty line to just above it).

4 The poverty severity index measures the contribution of the very poor to the poverty gap.

5 Social Assistance programs are programs aimed at maintaining households out of poverty, are usually targeted to the poor, and are not linked to previous contribution to an insurance pool. Social Insurance programs, on the other hand, are programs designed to mitigate the impact of unexpected income shortfalls due to unemployment, health problems, disability and old age.

6 For any measure, the growth rate per capita is equal to the total growth rate minus the population growth rate.

7 For discussion of the reasons for these differences, see Deaton (2005) and Ravallion (2003).

8 For 1997, the poverty and the extreme poverty lines were set to B.\905 and B.\519. For 2003, there were set to B.\953 and B.\534, respectively.

9 Tests for statistical significance of the changes in poverty and inequality are given in Annex 1.3. The changes are generally significant, with the following exceptions: changes in the Gini coefficient for urban areas, the national headcount poverty rate using the moderate poverty line, and the poverty gap and poverty severity index for indigenous areas using the extreme poverty line.

10 The approach followed here is that off Datt and Ravallion (1992).

11 See the study prepared by Barros, Carvalho, Franco and Mendonca (2006) in Annex 3 for further details.

12 This analysis follows Huppi and Ravallion (1991).

13 The simulations follow the procedure of Datt and Walker (2003).

14 See Annex 1.1, Figure A1.1.4 for the poverty impact of three different growth scenarios with an associated increase in inequality of 1 percent. The inequality reductions referred to here are percentage reductions in the Gini coefficient. In the simulations, inequality is reduced through a proportional shift in the Lorenz curve which adjusts the consumption level of each household relative to its deviation from the mean. The adjustment effectively redistributes consumption from households with consumption above the mean to those with consumption below the mean. The precise formula can be found in Datt and Walker (2003).

15 Article 35 of Law 34 of 1995 made obligatory and free (in public schools) the two years of pre-primary (ages 4-5 years); it also made free in public schools the three years of upper secondary.

16 Note here that we refer to school services as opposed to schools. Many of the pre-school programs are physically located within a primary school. Also, many schools have one set of classes while others have separate morning and afternoon programs. The data are presented at the level of school services. Thus a primary school that has both a morning and an afternoon pre-school program would be considered to be three school services, even though all of this activity is taking place in one physical space.

17 This rise appears in several data sources but it is not clear what the explanation is for such a startling increase.

18 Obviously, the first consideration was to ensure that the calculations themselves were correctly done: several checks have been done to ensure that the results reported in the table reflect the reality of malnutrition and are not simply measurement errors. The calculations of the various indicators (z scores and standard deviations) have been done using the Anthro software and the construction has been done several times by different individual. Additionally, tests of significance of the differences have been carried out to ensure that the results are statistically significant and not simply due to the fact that malnutrition is a relative rare event and the number of children in the analysis (zero to five year olds) is not large. As is shown in the previous table, the differences in chronic malnutrition are significant for all geographic levels although none of the other indicators show a significant change.

19 Clearly this is not exactly comparable as the age group measured is different and, to the extent that there is not global primary school attendance, the universe of children is itself different. It may be that the children who do not attend primary school are the very poor and the most likely to be malnourished. Thus a school census could understate the problem of malnutrition.

20 One could assume, just to see the effect, that chronic malnutrition in indigenous areas was measured badly in 1997. For both rural and urban areas, the 6-11 year olds show, on average an increase of 8.9 percent over the 1997 figures for 0 to 5 year olds. For this same relationship to exist within indigenous areas, the 1997 figure would have to be 53.9 percent chronic malnutrition. If that were the case, then the overall malnutrition rate for 1997 would have been slightly higher, but not enough higher to contradict the finding of a large increase in chronic malnutrition. So this does not appear to be a solution either.

21 There has been no change between 1997 and 2003 for these quality indicators

22 This does not mean that that there is no public health care provided in the corregimiento, the Ministry of Health has several modalities of mobile care.

23 Social Assistance programs are programs aimed at maintaining households out of poverty, are usually targeted to the poor, and are not linked to previous contribution to an insurance pool. Social Insurance programs, on the other hand, are programs designed to mitigate the impact of unexpected income shortfalls due to unemployment, health problems, disability and old age.

24 For a discussion of some of these difficulties see Marques, José Silvério “Central America, Cross-Country Evaluation of Social Safety Net Assessments (SSNAs)- Issues Paper”, paper prepared for the World Bank, November 2002.

25 Various subsidies (water, electricity, etc) are directed at seniors but these subsidies are generally not targeted on the poor.

26 Mesa, Alberto Arenas “ Alternativas de Políticas para la Reforma y Modernización del Sistema de Pensiones en la CSS en Panamá”, Septiembre 2004, p. 53.

27 B/ 60.00 is equivalent to one-half the existing minimum wage of B/ 119 for domestic work in the capital or to one third of the minimum pension of B/ 175

28 Cuevas, Fernando “Precios Combustibles en América Central”, Presentation to the Fuel Pricing Policies in Latin America and their Economic and Environmental Implications” December, 2002, Chile.

29 Some have argued that in many poor and isolated communities this is not possible because there is not local supply of the needed foodstuffs. Although this could be true in a few instances, it should not hold in most cases. Indeed, the new SENAPAN pilot program which targets the poorest communities includes among the eligible products that the beneficiary families can buy locally, the very same products that are distributed in kind to the schools.

30 A positive development in 2006 was the elimination of MEDUCA school lunch program, which duplicated the SIF program.

31 If only milk is replaced by a cheaper alternative, about US$ 4 million could be saved annually.

32 The water and electricity subsidies for seniors are also not targeted; only 16% of the seniors above 62 years are poor. If existing subsidies are targeted only on poor senior additional savings could be realized. However, since these subsidies are cross financed in principle the saving would accrue to other consumers.

33 Argentina’s Jefes de Hogares program is a bit different in that the “conditionalities” involve work-related and labor training actions on behalf of beneficiaries rather than school attendance and health care. Argentina also operates a smaller CCT, called the Income for Human Development Program (IDH), which conditions cash transfers on schooling and health care.

34 See: Maluccio (2004), Olinto (2004), Rawlings and Rubio (2004) and Rawlings (2004) for summaries of the impacts of CCTs.

35 Assuming a concave welfare function, or aversion to inequality, it can be shown that the welfare loss from erroneously excluding a extreme poor household is greater than the welfare gain from the reduction in program cost coming from excluding a non poor household.

36 These impacts are the short run immediate impacts assuming an average income elasticity of consumption of 0.8. We also take into account the targeting errors simulated above. In the next simulation exercise we look into long run impacts that take the effects of human capital accumulation into account. But if w assume that all three designs tested have the same impact on the demand for education and health services, then in order to chose the best design all we need is to simulate the short run immediate impact on welfare.

37 This recommendation was made in the 2000 World Bank Poverty Assessment and corroborated in the recent SENAPAN nutrition study. It may require a modification to the law.

38 Ibid.

39 For any measure, the growth rate per capita is equal to the total growth rate minus the population growth rate.

40 For discussion of the reasons for these differences, see Deaton (2005) and Ravallion (2003).

41 See Székely and Hilgert (1999). Among the 18 countries examined is a 1997 income survey in Panama.

42 Labor income accounted for 62.6% of total income in the 1997 survey and 64.4% of total income in the 2003 survey.

43 These are equivalent to annual growth rates of 3.5 percent for GDP and 2.7 percent for labor income. Note that labor income grew more rapidly than income overall, which grew at a rate of 2.2 percent.

44 See Moran (2005) and Gasparini and Sosa Escudero (1999) for references related to this topic.

45 One could assume, just to see the effect, that chronic malnutrition in indigenous areas was measured badly in 1997. For both rural and urban areas, the 6-11 year olds show, on average an increase of 8.9 percent over the 1997 figures for 0 to 5 year olds. For this same relationship to exist within indigenous areas, the 1997 figure would have to be 53.9 percent chronic malnutrition. If that were the case, then the overall malnutrition rate for 1997 would have been slightly higher, but not enough higher to contradict the finding of a large increase in chronic malnutrition. So this does not appear to be a solution either.

46 Atalah, Eduardo and Rosario Ramos “Evaluación de Programas Sociales Con Componentes Alimentarios y/o Nutricionales en Panamá”, Informe de Consultaría, SENAPAN, October 2005.

47Secondary education is organized in two cycles of three years each. The lower secondary education (grades 7-9) is mandatory and covers general subjects. The second cycle or upper secondary (grades 10-12) is voluntary and has two tracks: academic and professional-technical.

48 The estimate of population not covered at secondary level includes 35,194 overage students (12-14 years) that attend primary school. Estimates from MEDUCA’s Planning Department.

49 “Objetivos de Desarrollo del Milenio, Segundo Informe”,Gabinete Social de la Republica de Panama, 2005, Box 9, p. 75,

50 Total newborns in 2003 were 61,743. “Panamá en Cifras 2000-04”. Dirección de Estadística y Censos, Noviembre 2005, p. 42.

51 Data for 2003. “Objetivos de Desarrollo del Milenio, Segundo Informe”, Gabinete Social de la Republica de Panamá, 2005.

52 Contributors relative to the labor force. See Caroline Crabble, editor, “A Quarter Century of Pension Reform in Latin American and Caribbean: Lessons Learned and Next Steps”, IDB, 2005, (Table 1.12, p. 36).

53“Situación de la Seguridad Social,” Dirección de Estadística y Censo, 2003 (Cuadro 421-03). “Panama-Selected Issues and Statistical Appendix”, IMF, February 2006, Table 7, p.34.

54 Nutricereal provides 350 calories, 12 grams of proteins, 220 micrograms of vitamin A, 5.7 milligrams of iron, and 250 grams of calcium per 100 grams.

55 In 2005, there was another school lunch program managed by MEDUCA which was discontinued in 2006. This program transferred funds to the schools to finance the purchase of foodstuffs. In 2005 it transferred B/ 786,000 and benefitted 14,415 students.

56 Law No. 8 of March 2006 changed the name of INAFORF to INADHE or National Institute for Human Resource Development. (Instituto Nacional de Formación Profesional y Capacitación para el Desarrollo Humano). The Law contemplates that the resources that were previously assigned to the Ministry of Labor, Panama Institute of Tourism, and the Authority for the Medium, Small and Micro Enterprises for training will be reassigned to INADEH, which will train the personnel needed for the expansion of the Canal if the project materializes. INADEH budget could reach B/ 37 million.

57 Information on INSAFORT is based on its web page. No detailed information on the programs (cost or coverage) is available.

58 Other electricity subsidies are detailed below.

59 An IDB supported program which also involves direct subsidies has been recently initiated. This program for US$ 12.6 million (including US$ 2.6 million in counterpart funding) seeks to develop new instruments that will expand access to suitable housing for poor population. It includes four components: (i) improvement of neighborhoods (Integral Improvement of Neighborhood Program, PROMEBA); (ii) Construction of basic units of 36 m2 at an approximated cost of US$3,000 for families with monthly incomes of less than US$ 300 (Improved PARVIS). The component will benefit approximately 500 families in rural and indigenous areas considered high-priority by the country’s SPS. Units will be delivered fully terminated (turn key solution) with no cost to the beneficiary. Preference will be given to poor single mother with children with disabilities and senior. This component is expected to disburse B/ 1.2 million in 2006. (iii) Urbanization of macro lots to be sold to private developers that will build 600 low income houses below US$ 13,000 (Program of Provision of Basic Infrastructure, PROBIDA); (iv) Provision of a subsidy up to US$ 2,000 per family to facilitate the acquisition of low income housing (below US$ 16,000). Eligible families must have a monthly income of less than US$ 300 and demonstrate that they can obtain a commercial mortgage to complete the purchase. The Program will finance 1,000 subsidies at a cost of US$ 2 million (Housing Solidarity Program, PROVISOL).

60 Law No. 3 of May 24, 1985.

61 This section is based on Chapter VIII of the Panama- Public Expenditure Review, World Bank, 2006.

62 Law No. 15 of 2001.

63 The estimate is base on the following calculations. Number of consumers of less than 100 kwh: 252,016 (COPE); the average tariff in the second semester of 2005 (source ERSP) was: first 10 kwh : B/ 1.67; next 90 kwh: B/ 0.144 kwh; the monthly bill would be: B/14.63; the discount (20%): B/ 2.926; the monthly subsidy 252,016 x 2.926= B/ 737,398; and the annual subsidy: B/ 737,398 x 12= B/ 8.8 million.

64 Law No. 37 of 2001.

65 The estimate is base on the following calculations. Persons in retiring age represent 8.7 percent of the population (Departament of Statistics and Census). If same proportion applies to electricity consumers the number of seniors that are consumers of electricity would be: 685,711x0.087= 59,656; assuming that there are two senior for each house, the number of beneficiaries are: 59,656/2= 29,828. The average tariff (ERSP) in the second semester of 2005 was as follows: first 10 kwh : B/ 1.67; next 490 kwh: B/ 0.144 kwh; other 100 kwn B/ 0.150 kwh. Monthly cost would be: B/87.23; the discount (25%), B/ 21.8; the monthly subsidy 29,828 x 21.8= B/ 650,250; and the annual subsidy: B/ 650,250 x 12= B/ 7.8 millions.

66 The subsidy was created by Executive Decree No. 13 of April 7, 1993, and modified by the Executive Decree No. 30 of March 25, 1998.

67 Executive Decree No. 36 of September 17, 2003 (Article 78)

68 Instrumenting earnings with a selection bias procedure requires instruments that are not readily available. The standard correction using standard two stage procedure is weak in the case of more than two category choices.

69 Martha Rodríguez led the fieldwork for the qualitative análisis with Cecibel Flor de Liz Arias, Deici Guainara Guainara, Lineth Marilin Herrera, Enelía Mezúa Apochita, Filomena Miranda, Maritza Rodríguez. We thank the Ministry of Social Development (MIDES) and in particular the Social Protection Secretary (SPS) and the Direction for Indigenous People for their support. We also acknowledge the support of the Social Policy Division of the Ministry of Economy and Finance. We are grateful for the collaboration of the Kuna General Congress, the Embera comarca governor and community leaders in the comarca Ngobe. We are indebted to the Kuna Yala communities of Carti Yandup and Carti Sugdup, Ngobe communities of Soloy/Jebay, Oma and Quebrada de Guabo, and the Embera communities of Unión Choco y Puente for their support and participation in focus groups.

70 This section draws largely on Vakis and Lindert (2000), University of Maryland Minorities At Risk Project (2004) and US Department of State (2006) and from the 2003 LSMS.

71 Traditionally, the Kuna leadership fixes the price of coconuts and severely enforces it so as to maintain a predictable income for the islands.

72 In March 2005, authorities detained tour high-ranking Panamenian National Police stationed in Darién Province on charges of sexual abuse of minors. At year’s end, one officer was dismissed and the other three remained under investigation (US State Department, 2006)

73 Interestingly, no group commented on the possibility of using traditional dress in school.

74 Campuria means non-indigenous people.

75 This growth refers to puberty.

76 Nose rings are traditional gold jewels which Kuna wear with their traditional dress.

77 Men go by boat every other day to collect water in 20l containers

78 En ASMUN participan mujeres y hombres

79 It is possible to suppose a non lineal relation (cuadratic, cubic) between the consumption and income. This supposition implies that marginal propensity consumption can varies when we move along the income distribution.

80It is important to issue that we also estimate models which contemplates non-lineal specifications assuming quadratic and cubic relations between the consumption and income. We not take into account these models given the fact that the propensity to consume resulted to rise along the income distribution.