More Evidence on the Effect of Higher Unemployment on the Canadian Size Distribution of Income

7
Canadian Public Policy More Evidence on the Effect of Higher Unemployment on the Canadian Size Distribution of Income Author(s): Susan Johnson Source: Canadian Public Policy / Analyse de Politiques, Vol. 21, No. 4 (Dec., 1995), pp. 423-428 Published by: University of Toronto Press on behalf of Canadian Public Policy Stable URL: http://www.jstor.org/stable/3551340 . Accessed: 14/06/2014 04:17 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. . University of Toronto Press and Canadian Public Policy are collaborating with JSTOR to digitize, preserve and extend access to Canadian Public Policy / Analyse de Politiques. http://www.jstor.org This content downloaded from 188.72.126.108 on Sat, 14 Jun 2014 04:17:14 AM All use subject to JSTOR Terms and Conditions

Transcript of More Evidence on the Effect of Higher Unemployment on the Canadian Size Distribution of Income

Canadian Public Policy

More Evidence on the Effect of Higher Unemployment on the Canadian Size Distribution ofIncomeAuthor(s): Susan JohnsonSource: Canadian Public Policy / Analyse de Politiques, Vol. 21, No. 4 (Dec., 1995), pp. 423-428Published by: University of Toronto Press on behalf of Canadian Public PolicyStable URL: http://www.jstor.org/stable/3551340 .

Accessed: 14/06/2014 04:17

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

University of Toronto Press and Canadian Public Policy are collaborating with JSTOR to digitize, preserveand extend access to Canadian Public Policy / Analyse de Politiques.

http://www.jstor.org

This content downloaded from 188.72.126.108 on Sat, 14 Jun 2014 04:17:14 AMAll use subject to JSTOR Terms and Conditions

More Evidence on the Effect of

Higher Unemployment on the

Canadian Size Distribution of Income* SUSAN JOHNSON Department of Economics Wilfrid Laurier University

Ce texte montre que le ch6mage cyclique est associ6 a un accroissement des inegalit6s de revenu au Canada. Nos r6sultats empiriques renforcent ceux obtenus par Erksoy qui furent publi6s dans cette revue en septembre 1994 dans un article intitul6: 'The Effects of Higher Unemployment on the Distribution of Income in Canada: 1981-1987'. Erksoy utilise un modble de micro-simulation dynamique avec lequel il modelise la p6riode 1981-1987. Mon texte, en contraste avec le sien, prend une perspective macroeconomique et utilise un modble de parts de revenu. J'examine la p6riode de 1981 a 1992 en utilisant des donnees de l'Enquete sur les finances des consommateurs. I1 me semble trbs r6vdlateur que deux approches aussi diff6rentes permettent d'en arriver a la meme conclusion.

This paper provides evidence that cyclical unemployment is associated with increased income inequality in Canada. The empirical results reinforce the recent findings of Erksoy, 'The Effects of Higher Unemployment on the Distribution of Income in Canada: 1981-1987,' that appeared in the September 1994 issue of this journal. Erksoy used a dynamic micro-simulation to model the period from 1981 to 1987. In contrast to his work, my paper takes a macroeconomic perspective and uses an income share model. I examine the period from 1981 to 1992 using data from the Survey of Consumer Finances. It is striking that two such different approaches reach the same conclusion.

I Introduction

This paper provides evidence that cycli- cal unemployment is associated with in-

creased income inequality in Canada. These empirical results reinforce the re- cent findings of Sadettin Erksoy (1994). Erksoy used a dynamic micro-simulation to model the behaviour of the Canadian in- come distribution in response to changes in the unemployment rate over the period from 1981 to 1987.

I present results in this paper using a different technique, a longer time period and a different data set. In contrast to Erk-

soy, my paper takes a macroeconomic per-

spective and uses an income share model. I examine the period from 1981 to 1992 using data from the Survey of Consumer Fi- nances. It is striking that two such differ- ent approaches reach the same conclusion.

II The Model

In order to examine the short-run effects of cyclical unemployment on the income dis- tribution an income share model is used. Quantile income shares (Qi) are regressed on variables that are thought to determine the income distribution. In this case, unem- ployment (U), and inflation (rr) are included to capture short-run cyclical effects on the

Canadian Public Policy - Analyse de Politiques, XXI:4:423-428 1995 Printed in Canada/Imprim6 au Canada

This content downloaded from 188.72.126.108 on Sat, 14 Jun 2014 04:17:14 AMAll use subject to JSTOR Terms and Conditions

income distribution. A trend (T) is included to capture long-run factors that effect the income distribution.1'2

The equation estimated for the ith quan- tile, Qi is: Qi = ai + PiU + yTiT + iT + Ei (1)

This creates a model with 'i' seemingly unrelated regressions that potentially have cross-correlated error terms. It is impor- tant to note that when the independent variables in all equations in the system are the same the income share model automati- cally imposes adding-up constraints. The constants from all five equations sum to 100 (when percentages are used). The coeffi- cients on each variable sum to zero.3

If unemployment increases the degree of income inequality in the Canadian econo- my we would expect the coefficients on un- employment to be negative for lower in- come quantiles and positive for higher income quantiles.

The income share model has been used in many other studies that examine the im- pact of economic variables on the income distribution.4 As Erksoy correctly points out, the use of annual cross-sectional data (required by the income share approach) rather than longitudinal data (used in the dynamic micro-simulation) does not accu- rately measure the total impact of unem- ployment on the income distribution for two reasons. First, over time the gains and losses in individual earnings may cancel each other out and smooth the redistribu- tional impact of unemployment. Second, unemployment occurs within real time and observed spells of unemployment may not necessarily align with an annual time frame. This criticism must be kept in mind when interpreting the empirical results in this paper.

III Data

The Income Share Variable The income share variable is based on data from the Survey of Consumer Finances.5 The income concepts in this survey are un- changed since 1965.

Three income definitions are available from the Survey of Consumer Finances. Total Money Income includes wages and salaries (gross, before deductions for unem- ployment insurance etc.); net income from self employment (gross income minus ex- penses); investment income (bond interest, dividends, bank interest etc.); government transfer payments; pensions; and miscel- laneous income (scholarships, alimony etc.). It is important to note that Total Money Income includes transfer payments. A second measure, Income After Tax, ad- justs Total Money Income for the payment of income taxes. A third measure is Total Money Income Before Transfers which pro- vides a measure of income before govern- ment transfers are received or income taxes are paid. The regression model is estimated using each of these income concepts as the dependent variable. When these results are compared it is possible to isolate the impact of the transfer and income tax systems on the effect of unemployment on the income distribution.

Quintile share data are available for the above income concepts for the period from 1981 to 1992 for various groups in the popu- lation.6 The quintile shares are expressed in percentage terms.

The Survey of Consumer Finances is based on a sample designed to represent virtually all private households in Canada. All Families and Unattached Individuals is the most comprehensive group for which quintile income share data are available for all three income concepts over the period from 1981 to 1992. Therefore this is the group chosen for this study.7

Explanatory Variables Inflation is expressed as an annual percent- age and is measured as the rate of change in the Consumer Price Index. Unemploy- ment is the overall unemployment rate in Canada expressed as a percentage.8

The data used in my study are fun- damentally different from those used by Erksoy. First, the data used here are based on actual reported incomes collected in the

424 Susan Johnson

This content downloaded from 188.72.126.108 on Sat, 14 Jun 2014 04:17:14 AMAll use subject to JSTOR Terms and Conditions

Table 1 Total money income before transfers. Families and unattached individuals unrestricted joint multivariate analysisI

Quintile Constant U r T R2 2 D.W.

Q1 2.217 -.0850 .0143 -.0195 .84 2.59 (.000) (.001) (.243) (.075) (.545)

Q2 11.76 -.1877 .0575 -.0857 .92 2.80 (.000) (.000) (.041) (.002) (.707)

Q3 19.931 -.0843 .0070 -.0990 .92 2.14 (.000) (.006) (.687) (.000) (2.13)

Q4 26.743 .0348 -.0225 -.0249 .24 2.93 (.000) (.239) (.295) (.178) (.800)

Q5 39.74 .3108 -.0669 .222 .95 2.29 (.000) (.000) (.092) (.000) (.311)

1 The numbers in parentheses are P-values for the null hypothesis that the parameter is equal to zero. In the case of the Durbin-Watson statistic the P-value is for the null hypothesis that there is no autocorrelation. 2 This is the adjusted R2. SystemR2 = .9966.

Cross equation tests on the coefficients The test of the joint hypothesis that the parameters on the unemployment variable equal zero is rejected. Ho: p1= 02= P3= P4= P5= 0 F: 13.32 P: .000. The test of the joint hypothesis that the parameters on the inflation variable equal zero is rejected. Ho: Tl=y2=T3=Ty4=?y5=0 F:3.77 P:.007.

Table 2 Total money income after transfers and before tax. Families and unattached individuals unrestricted multivariate regression analysis'

Quintile Constant U R T R2 D.W.

Q1 4.419 -.0384 .0111 .0224 .47 2.48 (.000) (.081) (.453) (.095) (.453)

Q2 10.575 -.0247 .0492 -.0068 .81 2.55 (.000) (.198) (.005) (.546) (.508)

Q3 18.407 -.0211 .02351 -.058 .88 2.75 (.000) (.318) (.145) (.001) (.669)

Q4 25.68 .0036 -.0093 -.0354 .52 2.85 (.000) (.860) (.540) (.019) (.743)

Q5 40.437 .0927 -.0644 .0929 .84 2.00 (.000) (.073) (.087) (.010) (.137)

1 The numbers in parentheses are P-values for the null hypothesis that the value of the parameter is zero. In the case of the Durbin-Watson statistic the P-value is for the null hypothesis that there is no autocorrelation. 2 This is the adjusted R2. System R = .9943.

Cross equation tests on the coefficients The test of the joint hypothesis that the parameter on the unemployment rate is zero is rejected. Ho: P01= P2= P3= 4= 05= 0 F: 3.05 P: .0200. The test of the joint hypothesis that the parameter on the inflation rate is zero is rejected. Ho: 1=y2=y3=T4=y5=0 F:6.81 P:.001.

Survey of Consumer Finances rather than on a simulation. Erksoy notes that repre- sentative longitudinal microdata are not

available in Canada. Therefore he was forced to estimate the dynamic effects using only the two year (1986-87) panel of micro-

High Unemployment and Income Distribution 425

This content downloaded from 188.72.126.108 on Sat, 14 Jun 2014 04:17:14 AMAll use subject to JSTOR Terms and Conditions

Table 3 Income after tax and transfers. Families and unattached individuals unrestricted multivariate analysis1

Quintile Constant U r T R2 2 D.W.

Q1 4.735 -.0373 .01278 .0431 .86 2.42 (.000) (.019) (.211) (.001) (.407)

Q2 10.81 -.0066 .0527 .0204 .70 1.90 (.000) (.692) (.002) (.075) (.094)

Q3 18.319 -.0012 .0187 -.0364 .70 2.20 (.000) (.959) (.298) (.032) (.247)

Q4 25.317 .0113 -.0108 -.0285 .36 2.89 (.000) (.592) (.488) (.051) (.78)

Q5 40.207 .0648 -.0580 .0097 .56 2.47 (.000) (.147) (.087) (.707) (.444)

1 The numbers in parentheses are P-values for the null hypothesis that the parameter is equal to zero. In the case of the Durbin-Watson statistic the P-value is for the null hypothesis that there is no autocorrelation.

2 This is the adjusted R2 System R2 = .9930.

Cross equation tests on the coefficients The test of the joint hypothesis that the parameters on the unemployment rate is equal to zero is rejected. Ho: P1= P2= 03= 04= 05= 0 F: 2.961 P: .0229. The test of the joint hypothesis that the paramer on the inflation rate is equal zero is rejected. Ho: y1=y2=y3=y4=y5=0 F:8.50 P:.0001

data available in the Labour Market Activ- ity Survey to simulate for the period from 1981 to 1987. The time series aggregates I employ are available for the years 1981 to 1992. The additional five years of data con- tain a second recession and therefore more information about the influence of unem- ployment on the income distribution. Sec- ond, Erksoy uses a narrow income concept based on employment earnings and unem- ployment insurance. My paper uses three income concepts and the basic definition of income is much broader. Finally, the cover- age in my paper is more comprehensive. Erksoy's simulation is performed only for labour market participants with at least one week of employment and omits 15 per cent of household heads and wives in the Assets and Debts Survey. The results pre- sented here are based on a sample designed to represent all households and unattached individuals in Canada.

IV Empirical Results

The results from the empirical estimation of the income share model for each of the

three income concepts are presented in Ta- bles 1,2, and 3.9

Table 1 presents the results for the model based on income before taxes and transfers. The coefficients on unemploy- ment are negative for the bottom three quintiles and positive for the top two quin- tiles. Except for the fourth quintile all of these coefficients are significant. This clearly indicates that higher unemploy- ment is associated with increased income inequality before the government adjusts income through the transfer and income tax systems.

Examination of Table 2 and a compari- son of these results to those in Table 1 re- veals the influence of the transfer system on the effect of unemployment on the in- come distribution. The coefficients on un- employment are negative for the bottom three quintiles and positive for the top two quintiles. All of these coefficients are smaller than those in Table 1. Only the coefficients on the first and fifth quintiles are significant. These results suggest that unemployment worsens the income dis- tribution but its effect is not nearly as

426 Susan Johnson

This content downloaded from 188.72.126.108 on Sat, 14 Jun 2014 04:17:14 AMAll use subject to JSTOR Terms and Conditions

strong when the transfer system is taken into account.

When the results in Table 3 are com- pared to those in Table 2 it is possible to iso- late the impact of the income tax system on the effect of unemployment on the income distribution. Again, the coefficients on un- employment are negative for the bottom three quintiles and positive for the top two quintiles. Generally, the coefficients are slightly smaller than in the income after transfers model. Only the coefficient on the first quintile is significant. These results show that the income tax system has little additional affect over the transfer system in redressing the increased inequality re- lated to unemployment.

V Conclusions

The evidence presented in this paper sup- ports the view that cyclical unemployment increases the degree of income inequality in Canada. The transfer system appears to be an important mechanism for redressing the increased inequality associated with higher unemployment. The income tax system is less important than the transfer system.

This paper provides strong support for

Erksoy's results because it reaches the same conclusion - namely that higher un-

employment increases income inequality in Canada - using a different perspective, technique, data and time period.

Notes

* The author would like to thank Frank Denton, David Johnson and Mike Veall for helpful com- ments on this paper.

1 For a more detailed discussion of the theory be- hind the inclusion of these explanatory variables see Johnson (1994).

2 Since the goal of this study is not to explain the factors that have caused long-run changes in the income distribution it is hoped that the inclusion of the time trend will be sufficient to capture the effect of these variables. This may result in some omitted variable bias in the regressions. Since there are only 12 observations it is hoped that the increased efficiency from having more degrees of freedom will compensate for any bias that is intro-

duced. 3 As the explanatory variables are the same in each

equation, the model is estimated efficiently by OLS.

4 Other studies that have used this approach in- clude Beach (1976;1977), Blank and Blinder (1986), Blinder and Esaki (1978), Buse (1982), McWatters and Beach (1990) and Nolan (1989).

5 The data from the Survey of Consumer Finances are available in two Statistics Canada publica- tions, Income Distributions by Size in Canada (cat.no. 13-207) and Income After Tax, Distribu- tions by Size in Canada (cat. no. 13-209).

6 Quintile share data for Total Money Income are available on a limited basis (i.e. only for some population groups) biennially from 1965 to 1971 and annually from 1971 to 1992. Quintile share data for Total Money Income Before Transfers are available on a limited basis annually from 1981 to 1992. Quintile share data for Income After Tax are available on a limited basis annually from 1971 to 1992. One of the reasons the sample size was limited to the period from 1981 to 1992 is that this is the only period for which quintile share data are readily available for all three income concepts.

7 A family is defined as a group of individuals shar- ing a common dwelling that are related by blood, marriage or adoption. Family income is the sum of the income of all family members. Unattached individuals are people who live alone or in house- holds where they are not related to the other mem- bers of the household.

8 These data came from the Department of Finance (1993).

9 Restricted versions of these models were run in an attempt to increase efficiency. These results are presented in an Appendix, available from the author.

10 In the discussion of the empirical results a coeffi- cient is considered significant if it is significant at at least the 10% level.

References

Beach, Charles M. (1976) 'Cyclical Impacts on the Personal Distribution of Income,' Annals of Economic and Social Measurement, 5:1.

- (1977) 'Cyclical Sensitivity of Aggregate In- come Inequality,' Review of Economics and Statistics, 59:56-66.

Blank, Rebecca and A. Blinder (1986) 'Macroe- conomics, Income Distribution, and Poverty.' In Sheldon Danziger and Daniel Weinberg (eds.), Fighting Poverty, What Works and What Doesn't (Cambridge, MA: Harvard University Press).

Blinder, A. and H. Esaki (1978) 'Macroeconomic Activity and Income Distribution in the Post-

High Unemployment and Income Distribution 427

This content downloaded from 188.72.126.108 on Sat, 14 Jun 2014 04:17:14 AMAll use subject to JSTOR Terms and Conditions

war United States,' Review of Economics and Statistics, November:604-8.

Buse, Adolf (1982) 'The cyclical behaviour of the size distribution of income in Canada: 1947- 78,' Canadian Journal of Economics, May:189-204.

Department of Finance (1993) Economic and Fiscal Reference Tables. August.

Erksoy, Sadettin (1994) 'The Effects of Higher Unemployment on the Distribution of In- come in Canada: 1981-1987,' Canadian Pub- lic Policy - Analyse de Politiques, XX:3:318-28.

Johnson, S. (1994) 'The Effect of Anti-Inflation Policies on the Size Distribution of Income: A Case Study of Canada 1981-1992.' Wilfrid Laurier University Working Paper 94-0018.

Malik, M.H. (1983) 'Macroeconomic influences on the Distribution of Income: A Case Study

of Canada.' Unpublished Ph.D thesis, McMaster University, Hamilton, ON, April.

McWatters, C.J. and C.M. Beach (1990) 'Factors Behind the Changes in Canada's Family In- come Distribution and the Share of the Middle Class,' Relations Industrielles/In- dustrial Relations, 45:1:118-35.

Nolan, Brian (1989) 'Macroeconomic Conditions and the Size Distribution of Income: Evi- dence from the U.K.' Pp. 115-37 in Paul Davidson and Jan Kregel (eds.), Macroe- conomic Problems and Policies of Income Distribution (Brookfield, VT: Gower Publish- ing Co.).

Statistics Canada. Income Distributions by Size in Canada. Catalogue No. 13-207, 1981-1992. - Income After Tax, Distributions by Size in

Canada. Catalogue No. 13-209, 1981-1992.

The Centre for Research on Economic and Social Policy, UBC announces recent discussion papers:

DP-26 Jobs, Labour Standards and Promoting Competitive Advantage: Canada's Policy Challenge, by Morley Gunderson and W. Craig Riddell

DP-27 Taxation Policies for British Columbia in the New Economy, by Jonathan R. Kesselman DP-28 Interprovincial Migration and the Efficacy of Provincial Employment Creation, by

Gideon Rosenbluth DP-29 Payroll Taxes Around the World: Concepts and Practice, by Jonathan R. Kesselman DP-30 Unemployment Insurance and Job Durations: Seasonal and Non-Seasonal Jobs, by

David A. Green and Timothy C. Sargent

Single copies are available upon request by writing: CRESP, c/o Department of Economics, University of British Columbia, Vancouver, B.C. V6T 1Z1

428 Susan Johnson

This content downloaded from 188.72.126.108 on Sat, 14 Jun 2014 04:17:14 AMAll use subject to JSTOR Terms and Conditions