EPD INCOME DISTRIBUTION PROJECT. Jong-goo Park. Division Working Paper No August 1980

Transcription

1 Public Disclosure Authorized DRAFT EPD INCOME DISTRIBUTION PROJECT Public Disclosure Authorized DATA ON INCOME DISTRIBUTION IN TRINIDAD AND TOBAGO Jong-goo Park Public Disclosure Authorized Division Working Paper No August 1980 Public Disclosure Authorized Economic and Social Data Division Economic Analysis and Projections Department Development Policy Staff The World Bank Division Working Papers report on work in progress and are circulated for Bank staff use to stimulate discussion and comment. The views and interpretations in a Working Paper are those of the author and may not be attributed to the World Bank or its affiliated organizations.

2 DATA ON INCOME DISTRIBUTION IN TRINIDAD AND TOBAGO This paper evaluates Trinidad and Tobago's 1975/76 survey of household budgets and estimates of income distribution derived therefrom. Based on the evaluation, it adjusts the survey data for under-reported income. The paper then derives a more representative distribution of household income for Trinidad and Tobago. ECONOMIC AND SOCIAL DATA DIVISION ECONOMIC ANALYSIS AND PROJECTIONS DEPARTMENT AUGUST 1980

3 TABLE OF CONTENTS Page I. Introduction....l... 1 II. The 1975/76 Household Budgetary Survey... 1 III. Evaluation of the 1975/76 Household Bugetary Survey.. 3 IV. Data Adjustment and Estimation of Income Distribution. 7 V. Concluding Remarks Appendix Tables

4 DRAFT JPark:prr August 21, 1980 DATA ON INCOME DISTRIBUTION IN TRINIDAD AND TOBAGO I. Introduction 1. This paper evaluates Trinidad and Tobago's 1975/76 survey of household budgets and the estimates for household income distribution derived therefrom. It also discusses the survey's various shortcomings - the under-reporting of income, the low rate of enumeration, the bias toward lower income households, and the insufficient disaggregation of the published data. 2. National household budget surveys have been conducted in Trinidad and Tobago for 1957/58, 1971/72, and 1975/76. Although designed primarily for obtaining information on household expenditure patterns, these surveys also provided data for estimating the size distribution of household income. A World Bank mission used the survey data to estimate the country's distribution of household income for the periods,1971/72 and 1975/76.-l Subsequently, these estimates were published in the Bank's 1980 Social Indicators Data Sheets, 3. These and other estimates were based, however, on data that excluded non-monetary income./ Moreover, there is no evidence that any attempt was made to evaluate the quality of the household budget survey data before using them to derive income distribution estimates. II. The 1975/76 Household Budgetary Survey (HBS) 4. The 1975/76 HBS was the third national sample survey of household budgets in Trinidad and Tobago. It was carried out by the country's Central Statistical Office. /1 Economic Position and Prospects of Trinidad and Tobago, World Bank Report No TR, Vol. II: Statistical Appendix, December 19, 1978, pp /2 Other estimates are found in Ahiram, E., "Distribution of Income in Trinidad and Tobago and Comparison with the Distribution of Income in Jamaica," Social and Economic Studies, Vol. 15, No. 2, 1966; and in Henry, R. M., "A Note on Income Distribution and Poverty in Trinidad and Tobago," ILO World Employment Programme Research Working Papers, WEP2-23/WP29, Geneva, Oct., 1975.

5 -2- a. Coverage 5. The survey covered all private households in the country. b. Definitions 6. Private Household consists of one or more persons, related or unrelated who occupy the same dwelling, sleep most n.ights in that dwelling, and share at least one daily meal. 7. Household Income includes the money income of all household members (before deductions for income tax and national insurance payments into pension funds, etc.). Also counted as income are such transfers as retirement pensions and gifts. Excluded are income in kind, imputed rents, and production consumed by the producer. c. Sample selection 8. The sample consisted of 2,992 households, or about 1.2% of all households in the country in The sample households were selected through a two-stage, stratified sampling process, with the probability of selection proportional to population size. The sampling frame was the same as that used in the 1970 census, but with some modifications made for population changes and new housing areas. The sample scheme was prepared so that each of the major administrative areas would be adequately covered. d. Survey methodology 9. The survey was conducted in two rounds. Each round lasted six months. The first ran from mid-march 1975 to mid-september 1975, the second from the beginning of October 1975 to the end of March Information was collected partly by interview and partly by questionnaires filled out by the individual household spenders. /1 Spenders are those household members who are 18 years or more of age or who are under 18 but working and not attending an educational institution fulltime. Excluded are those persons at least sixty-five years of age who are not heads of households.

6 -3- e. Period of Reference 10. Income data were for either the previous month or the most recent pay period. III. Evaluation of the 1975/76 Household Budgetary Survey a. Non-enumeration 11. Of the sample of 2,992 households, information was actually collected for only 2,493 households,equal to a 16.7% rate of non-enumeration. No directly relevant data are available for assessing with any precision the bias introduced by this relatively high rate of non-enumeration. However, data on the distribution of sample households and of average household income by major administrative areas do show that Port of Spain and St. George, which had relatively high average household income in 1971/72, were under-represented in the effective sample (i.e., the responding sample households), whereas Caroni and St. Patrick, which had relatively low average household income, were over-represented - (Table 1). This implies that the effective sample was biased toward the lower income households, if the non-enumerated households were randomly distributed throughout the different income classes in the various administrative areas. b. Under-reporting of Income 12. Survey data show that average expenditures exceeded average incomes for all but two income classes (Table 2). For the country as a whole, average household expenditures exceeded average household incomes. This would mean that the non-corporate private sector must have generated negative net savings, a dubious state of affairs. Because there is no evidence for any systematic upward bias in the expenditure data, the explanation for such an anomaly must be that most sample households under-reported their incomes in the survey much more than they overstated their expenditures.

7 -4- Table 1: DISTRIBUTION OF SAZIPLE HOUSE-HOLDS (1975/76) AND AVERAGE HOUSEHOLD MONETARY INCOME BY ADMINISTRATIVIE AREAS (1971/72) Distribution Distribution of Original of Enumer- Average Household S-ample Responding ation Monetary Income Administrative Househ?lds (A) House olcs (B) Biasl 2 in 197L/72 Areas 71 - (B-A) (TT 9/Month)/-! Port ot Spain ±Z San Fernando St. George Caroni Nariva/Mayaro St. Andrew/St. David Victoria St. Patrick Tobago Trinidad & Tobago /I Derived trom Household Budgetary Survey 1975/76, Report No.2, Republic of Trinidad and Tobago, Central Statistical Office, Port of Spain, November 1978,page xi, Table A. /2 The negative (positive) figure indicates that the particular areas was under-(over-) represented in the effective sample. /3 Derived from Household Budgetary Survey, Income and Expenditure Patterns, Rounds 17 and 18, Trinidad and Tobago, Central Statistical Office, Port of Spain, July 1974, Table 5, P Income data for 1971/72 (rather than for 1975/76) are used here because of their greater reliability, having been based on the recently completed 1970 Census and given the higher level of response to the 1971/72 survey.

8 -5 Table 2: AVERAGE MONTHLY HOUSEHOLD INCOME AND EXPENDITURE (April March 1976) Income Percentage Income (Y)J - Expendi-/a Class of ture (C) - C (TT$) households (TT$) (TT$) Y , , , , ,z 1,100-1, , , ,300-1, , , , , , Country /a Imputed rents and production consumed by the producer are included. Source: Household Budgetary Survey 1975/76, Bulletin No , Trinidad and Tobago, Central Statistical Office, Port of Spain, June 1977, Table 1B, p. 1; Household Budgetary Survey 1975/76, Report No. 2, Central Statistical Office, Fort of Spain, Nov. 1978, Table 26, pp

9 Unfortunately, it was not possible to check the degree of overall under-reporting of income by comparing total household income as derived from the survey data with total household income as derived from national accounts data. There was not sufficient information to derive the latter. However, we did find that total household income as derived from the survey data was equal to 54% of estimated private consumption expenditure in the national accounts./ Even when such non-monetary incomes as imputed rents and home-grown food are included, this proportion increases to only about 60%o./- This too implies that the income figures derived from the survey data are very much understated. 14. There is indirect evidence that the under-reporting of income is particularly pronounced for households in the lowest income class (Table 2). For these households, average expenditures were more than three times average monetary incomes. Such high budget defitits could not have been financed by these low income households. Once again, under-reporting of income would appear to be the problem. 15. Households with a monetary income of at least TT$1,300 per month accounted for 5% cf total sample households in 1975/76. On the other hand, in the 1971/72 survey, households with a comparable income in real terms (a monetary income of at least TT$700 per month in 1971/72 prices) accounted /1 The private consumption expenditures in the national accounts of Trinidad and Tobago include some changes in inventories, estimated at about 20% of the total. See Table A-1 in the Appendix for the original HBS data. /2 See Table 3 for information on both monetary and non-monetary income by income classes, and Table A-2 for the HBS data adjusted to include nonmonetary income.

10 -7- for 9% of total households (Table A-7, Appendix). This would appear to imply that the high income households were substantially under-represented in the 1975/76 survey. IV. Data Adjustment and Estimation of Income Distribution 16. Relevant data, such as the functional distribution (wages, rent, interest, entrepreneurial income) of national income and household income distribution by household-head occupation were not available for us to adequately assess, and then properly adjust for, the extent of under-reporting of income. As a result, the assumptions we employed in our data adjustments are in fact open to some crit*ism. 17. Because the published survey data on household incomes exclude nonmonetary income we added imputed rent from owner-occupied dwellings and the value of home-grown, home-consumed food (both of which were reported as expenditure items) to reported monetary income (Table 3). In addition, we assumed that the average income of the lowest income class (i.e. that with an average monetary income of less than TT$50 per month) was equal to that of the second lowest income class. This was based on the fact that (i) the average expenditure of the lowest income class (TT$196 per month) was close to that of the second lowest income class (TT$205 per month) and that (ii) the average proportion of expenditures on food to total expenditures of the lowest class was lower than that of the second lowest (see Table A-3, Appendix). /1 The 1971/72 survey is considered a fair representation because it was based on the recently completed 1970 census and because its non-enumeration rate was very low.

11 8- Table 3 AVERAGE MONTHLY HOUSEHOLD INCOME, 1975/76 Adjusted Monthly Average Monetary Average Monthly Income Monetary Imputed/la Home-grown /La Income Class Income (A) Rents (B) Food (C) (A+B+C) , , ,100-1,299 1, , ,300-1,499 1, , , , ,412.6 National Average /a Derived from HBS data on household expenditures. - Source: Household Budgetary Survey 1975/76, Bulletin No.1, op. cit., Table 1B, p.1 and Household Budgetary Survey 1975/76, Report No.2, op. cit., Table 26, pp

12 Almost all sample households under-reported their incomes in the 1975/76 survey. As a result, many higher income households were incorrectly classified as lower income households in the published survey data. For those households which reported incomes that were lower than expenditures we assumed /1 that their actual incomes were in fact equal to their expenditures-/ and we reclassified these households into higher income classes. To do this, we used data on the distribution of households cross-classified by income and expenditure groups (Tables A-4 and A-5, Appendix). The share of the top income class increased to 8%, which is comparable to that of the top income class in 1971/72 (see para. 15; Table 4). 19. We also estimated average household expenditures for each income class by using the adjusted household distribution (by both incomes and expenditures). Comparison of these estimated average household expenditures with the average household incomes (as derived in para. 17) suggests that many households in higher income classes actually had savings and that the private non-corporate sector as a whole generated net savings in 1975/76 (Table 5). 20. Using the data as derived in paras. 17 and 18, we estimated the distribution of adjusted household incomes for 1975/76 in TrLnidad and Tobago /2 (Table A-6, App,ndix). - From this estimate, we then derived income distribution by decile household groups (Table 6). /1 In the cases of the lowest three income classes, reported expenditures were assumed to be equal to actual incomes only for those households with reported expenditures more than double their income. /2 Total household income as derived from the adjusted data used for this estimate was 55% higher than that estimated from the original HBS data.

13 -10- Table 4: DISTRIBUTION OF HOUSEHOLDS, 1975/76 Monthly Monetary Households as Percentage Income of Total Households Class (TT$) Original/a Adjusted l , ,100-1, ,300-1, , TOTAL /a Derived from Household Budgetary Survey 1975/76, Report No.2, op. cit., Table lla p. 11. /b Derived from Household Budgetary Survey 1975/76, Bulletin No , op. cit., Table 3A, p.3.

14 Table 5: AVERAGE MONTHLY HOUSEHOLD INCOME AND EXPENDITURE (ADJUSTED), 1975/76 Percentage of Avg. Monthly Avg. Monthly Total Household Household B Households Income (TT$) Expenditure(TT$) A (A) (1) , , , , , , , , /a Estimated from Table A -5, by assuming the mid-point of each expenditure range as the average expenditure and then computing the weighted average expenditure for each income class, the weight being the number of households distributed over different expenditure classes within the income class.

15 Table 6.. HOUSEHOLD INCOME DISTRIBUTION IN THE TRINIDAD AND TOBAGO, 1975/76 (ADJUSTED) Cumulative % Cumulative % of Households Household Income/a (95) (83.1) Gini Coefficient /a Derived from Table A-3, Appendix,by linear interpolation between two adjacent income classes.

16 Our estimate showed slightly less income inequality (Table 7)- than did the 1975/76 estimate of income distribution appearing in the Bank Social 1/1 Indicators Data Sheets. This can be explained partly by the fact that (i) the income data used for our estimate included a non-monetary component, which accounts for the somewhat higher proportion of total income received by low income households and (ii) the reclassification of households reporting incomes below expenditures shifted relatively more households in low income than high income classes to higher income classes. Table 7: INCOME INEQUALITY MEASURES FOR 1975/76 Income Income Share (%) Share (%) Data Source Gini Coefficient of Lowest 20% of Highest 5% Income Group Income Group Adjusted HBS Data Original EBS Data (Social Indicators) 22. We reviewed briefly the national household survey conducted in 1971/72, in order to examine the trends in income distribution pattern between then and 1975/76./2 We found that income in that earlier survey was also generally /1 This comparison is not valid in the strict sense because of different income concepts employed for the two estimates (total income vs. monetary income). /2 The sample for the 1971/72 survey consisted of 2,826 households. Adequate information was obtained for 2,744 households, a 97% rate of enumeration. The sampling method, concepts, and definitions were the same as those in the 1975/76 survey (see Household Budgetary Survey, Income and Expenditure Patterns, HBS 1-2, Central Statistical Office, Port of Spain, July 1974).

17 under-reported.l- Using assumptions similar to those we used in adjusting the 1975/76 data (see paras. 17 and 18), we adjusted the published 1971/72 data on average household income and household distribution by income and expenditure groups. Then we derived estimates of 1971/72 household income distribution (Tables A-8 and A-9 Appendix). 23. A comparison of the adjusted income distribution estimates for 1971/72 and 1975/76 shows that the income distribution pattern in the country had remained virtually unchanged between the two periods (Table 8). Table 8: INC0OE INEQUALITY MEASURES IN 1971/72 & 1975/76 (ADJUSTED) Year Income Share Income Share (%) of Lowest (%)of Highest 20% Income 20% Income Gini Group Group Coefficient 1971/ / V. Concluding Remarks 24. Despite its national coverage, the 1975/76 household survey, with its emphasis on the pattern of household expenditures, did not provide a good data base for deriving estimates of income distribution. In addition to having a large under-reporting of income and a high non-enumeration of sample households, the published survey data did not include such relevant disaggregated /1 Total household income as derived from the 1971/72 survey data amounted to about 70% of total private consumption as estimated from national account data. See Table A-7, Appendix, for the originally published data on the 1971/72 survey.

18 data as household income distribution by urban and rural areas or income distr-bution by household members. Therefore, it was not possible to check the internal consistency of the survey data with respect to income distribution. Also, published data on household income did not show net income (after deductions of income tax, national insurance, and pension contributions). Furthermore, other relevant independent data sources (e.g., census and national accounts data) do not provide the information needed for evaluating the survey data. We, therefore, consider the income distribution estimates presented in this.paper to be subject to an unknown margin of error.

19 -16- Appendix Table A-1: DISTRIBUTION OF HOUSEHOLD MONETARY INCOkDE, 1975/76 (UNADJUSTED) Monthly /a Incom L Households Household Income - (TT$) No. % Amount (TT$) % - 4Y ,b74 u , , , , , , , , ,100-1, , ,300-1, , ,500-1, , ,gQo , TOTAL 2, ,140, /a Monetary income only. Source: Household Budgetary Survey, 1975/76, Bulletin No , Central Statistical Office, Port of Spain, June 1977, Table 1B, p.l.

20 -17- Appendix Table A-2: DISTRIBUTION OF HOUSEHOLD INCOME, 1975/76 (PARTIALLY ADJUSTED)/a Monthly Monetary Households Household Income Income Class (TT$) No. % Amount (TT$) % , , , , , , , , , ,100-1, , ,300-1, , , , TOTAL 2, ,365, Average Household Income (TT$) Average Household Size (persons) 4.6 /a Includesalso imputed rents and the value of home-grown food. Source: Household Budgetary Survey 1975/76, Bulletin No.1, op. cit. Table 1B, p.l., Household Budgetary Survey 1975/76, Report No.2, op. cit., Table 26, pp

21 Appendix Table A-3: AVERAGE HOUSEHOLD EXPENDITURE PATTERT, 1975/76 Percentage of Ex- Percentage of Monthly Monetary Percentage of Ex- penditures on Drink Expenditures on Income Class (TTS) penditures on food and Tobacco Accommodation , ,100-1, ,300-1, , Source: Household Budgetary Survey 1975/76, Report No.2, Central Statistical Office, Port of Spain, Nov. 1978, Table 12, p. 16 and Table 26, pp

22 -19- Table A-4: DISTRIBUTION OF HOUSEHOLDS BY imonthly HOUSEHOLD INCOME AND EXPENITURE GROUPS (UNADJUSTED) 1975/76 All Monthly incomat/ Monthly household income Undcr S50- S100-1s20o- $ $700- i$900- fi,100-1 S1,300- S1:500-, 1,700-f S1,900.. I. I. I - expenditures goups SSO $ $5299 $499 ;5699 $899 IS ,299 $1, ,899 land more ",umber of households (1) (2) (3) (4) (5' (6) (7) (8) (9) (10) (I 1) (12) (13) (14) A Expenditre Groups 2,493 ' Under 36 $ t 50- S I , $ S j S S ) * S 500-S S 700-t S S1, S1,100- S ,300- S1, Sl,500-51, S1,700- S1,899. S S S1,900 and more... j /a Monetary income Source: Household Budgetary Survey 1975/76, Bulletin No.l , Port of Spain, June 1977, p. 3.

23 Table A-5: DISTRIBUTION OF SAMPLE HOUSEHOLDS BY HOUSEHOLD INCOME AND EXPENDITURE GROUPS (ADJUSTED)'A, 1975/76 Monthly La Monthly Household Income (monetary) Groups (TT$)/- Household ,000-1,300-1,500+ Expenditure ,099 1,299 1,499 Groups (TT$) (60.5) (115.2) (191.9) (307.5) (459.9) (675.4) (915.9) (1162.0) (1382.7) (1610.7) (2,412.6) , ,100-1, ,300-1, ,500-1, , TOTAL /a Derived from Table A-4,. /b Figures in parentheses indicate average household income (both monetary and non-monetary) in each income class.

24 -21- Appendix Table A-6: DISTRIBUTION OF HOUSEHOLD INCOME, 1975/76 (ADJUSTED) Monthly Average Households Household Income /a Income (TT$) No. % Amount (TT$) % , , , , , , , , , , , , , , TOTALS 2, ,769, Average Household Income (TT$) Average Household Size (persons) 4.6 /a Includes also imputed rents and value of home-grown food. Source: See Tables 3 and 4 in the text.

25 -22- Appendix Table A-7: DISTRIT7UTION OF HOUSEHOLD INCOME, 1971/72 (UNADJUSTED) Monthly Monetary. Households Household Income/a Income Class (TT$) No. % Amount (TT$) % , , , , , , , , , , ,100-1, , ,300-1, , , , TOTAL 2, , Average Housenold Income (TT$/Month): Average Household Size (persons) 4.6 /a Monetary income only. Source: Household Budgetary Survey, Continuous Sample Survey of Population. Publication No. 22, Central Statistical Office, Port of Spain, July 1974, Table 6.

26 -23- Appendix Table A-8: DISTRIBUTION OF HOUSEHOLD INCOME, 1971/72 (ADJUSTED) Monthly Average Households Household Income/a Income (TT$) No. % Amount (TT$) % , , , , , , , , , , , , , , , TOTAL 2, ,244, Average Household Income (TT$/Month): Average Household Size (persons) 4.6 /a Includes imputed rents and home-grown food. Source: Household Budgetary Survey, Continuous Sample Survey of Population. Publication No.22, op. cit.

27 -24- Appendix Table A- 9: DISTRIBJTION OF HOUSEHOLD INCOME BY DECILE HOUSEHOLD GROUPS, 1971/72 (ADJUSTED) Cumulative Percent of Households Cumulative Percent of Househ 7 ld Income Q Q 50.2 XC 67.1 (95).. (79.8) 100Q Giri Coefficient: /a Derived from Table A- 8 Ty linear interpolation of two adjacent income classes.