Earnings inequality and gender in New Zealand, 1998-2008.
The decade from the mid-1980s to the mid-1990s was a period of enormous economic upheaval in New Zealand, as it was transformed from a highly protected economy to one of the most deregulated in the world. A number of studies have documented an increase in earnings inequality or income inequality over this same time period; however, less attention has been paid to the changes in inequality since the major reforms concluded. While inequality growth may have subsided in recent years, it is equally possible that it has remained high, affected by continuing adjustment of the labour market to the economic reforms. Important long-term trends in the attainment of university qualifications and the age and ethnic composition of the population have also been present during this time and it is unclear what effect these have had on inequality. Finally, despite considerable interest in the labour market outcomes of women, it remains unknown whether the observed convergence of male and female labour force participation rates has translated into a closing of the gender pay gap in recent years.
The objective of this study is to investigate what happened to earnings inequality in New Zealand between 1998 and 2008, both among men and women and between genders. Furthermore, it aims to provide an indication of which demographic and economic factors were important determinants of these changes in the earnings distribution. This is achieved by using Yun's (2006) 'unified' decomposition method, which not only allows estimates of how much each wage determinant contributed to the change in inequality, as measured by the variance of log hourly earnings, but also whether it was due to a change in the composition of the workforce or a change in coefficients over the period. The basic Yun method is extended to allow inequality between genders to be analysed too.
The next section reviews previous empirical research on income and earnings inequality in New Zealand, with Section 3 outlining the approach that will be used to analyse changes in inequality in this paper. After a discussion of the dataset that is used, Section 5 presents the results of estimating cross-sectional earnings equations for 1998 and 2008 and the conclusions that can be drawn regarding the source of changes in inequality between these years. Section 6 summarises the results of the paper and suggests some directions for future work to take.
2. Past research
A number of studies have analysed the determinants of changes in income and earnings inequality in New Zealand prior to 1998. (1) Martin (2002) used census data, which provides information on annual income only. Although he found that income became more equally distributed between 1976 and 1996 with respect to the entire population, due principally to employment and income growth among women, income inequality increased with respect to the labour force and the employed. Martin found that changes in the structure of employment accounted for a small part of this change, while unemployment did not have a significant effect. He concluded that the observed increase in income inequality is largely not captured by the standard variables used in labour market research.
The best data for studying income data in the 1980s are provided by the Household Economic Survey. This survey reveals that between 1983 and 1998, household gross income inequality, as measured by the Gini coefficient, interquartile range and 90-50 percentile ratio, increased; although the 50-10 percentile ratio indicated a slight decrease in inequality (Hyslop & Mare, 2005; Podder & Chatterjee, 2002). The growth in inequality was much smaller when income is adjusted for payments to and benefits from government (Crawford & Johnston, 2004). Much of the increase in inequality was concentrated between 1987/88 and 1990/91, which Bakker and Creedy (1999) suggested was due to a downturn in the business cycle during this period. Dixon (1998) noted that the increases in weekly earnings dispersion between 1984 and 1997 were substantially larger than the increases in hourly earnings dispersion, implying that at least part of the increased inequality could be attributed to changes in the distribution of hours worked.
Hyslop and Mare found that changes in household structure (particularly the declining proportion of two-parent families), attributes and employment outcomes each contributed to the observed increase in household income inequality between 1983/84 and 1997/98, while changes in returns were found to reduce the level of inequality. Collectively, these factors accounted for about half of the observed increase, depending on the measure of inequality used. Focusing on individual weekly earnings, Dixon found that, between 1984 and 1997, changes in observed characteristics accounted for 9% of the increase in inequality (measured by the standard deviation of log earnings) for men and 27% for women, changes in the returns to these characteristics accounted for 19% for men and 0% for women, with the remainder unexplained. The results for women, however, belie large opposing changes in the age and education distribution (which increased inequality) and the returns to age and education (which decreased inequality). For men, age and education together accounted for about half the effects of changes in both characteristics and returns.
In contrast to the plethora of studies of inequality during the 1980s and early 1990s, less work has focused on inequality over the last decade. This is partly because of a perception that the income distribution has changed little since the major economic reforms concluded in the mid-1990s, but also because the replacement of the Household Economic Survey with the New Zealand Income Survey in 1997 means that it is impossible to compare incomes since then with those of the earlier period. Two papers have studied changes in inequality since 1998 using the Income Survey. Hyslop and Yahanpath (2005) studied changes in the income distribution from 1998 to 2004, focusing on working-age individuals' earnings and income as well as household income. They found that individual earnings inequality (measured by the Gini coefficient) fell slightly, individual income inequality was unchanged and household income inequality increased slightly. Dixon and Mare (2007) focused on the distribution of weekly income among Maori between 1997 and 2003. Using a number of measures of inequality, they found that income inequality declined among working-age Maori but was relatively stable for employed Maori, reflecting the fact that employment rates among Maori rose significantly during this period so that far fewer received only income from benefits in 2003. Furthermore, the average income gap between Maori and Pakeha declined between 1997 and 2003.
Both Hyslop and Yahanpath, and Dixon and Mare performed decompositions of the mean and various percentiles of income, however, neither pair decomposed earnings and neither attempted to isolate the effects of changes in the distribution of and returns to each income determinant. In particular, Hyslop and Yahanpath found that changes in observed attributes explained a lot of the shifts in the lower income percentiles, but that changes in coefficients were more important further up the income distribution.
Taken together, the previous research indicates that both income and earnings inequality grew substantially in New Zealand in the decade from the mid-1980s but both have eased in recent years. While a lot of the changes in inequality are due to unexplained factors, changes in the composition of the workforce and the returns to measured characteristics appear to have contributed. There is little evidence of what specific factors had the largest effects on the growth in earnings inequality, although age and education appeared to play a role during the 1980s and 1990s.
A common approach to studying the determinants of changes in inequality is based on the tradition of Blinder (1973) and Oaxaca (1973), who were interested in explaining wage differentials in terms of differences in individual characteristics and differences in the coefficients of the wage-generating function. Juhn et al. (1993) provided a method for applying the so-called Blinder--Oaxaca decomposition to changes in income inequality between two points in time. This procedure allows the overall change in inequality to be decomposed into the part due to changes in the measured and unmeasured characteristics of the population and the part due to changes in the returns to these characteristics.
More recently, Fields (2003) proposed an alternative methodology that decomposes changes in income inequality over time into the portion explained by each determinant of income. In this approach, so-called relative factor inequality weights are calculated for each explanatory variable, measuring the proportion of the variation in income that is explained by the variable in a given year. The contribution of this variable to the overall change in inequality is then given by the change over time in the product of the factor weight and the chosen measure of inequality.
Yun (2006) noted that Fields' method does not decompose the changes in wage inequality in terms of characteristics, coefficients and residuals effects. On the other hand, the approach of Juhn et al. is able to explain changes in wage inequality in terms of these three effects but does not identify the separate contribution of each variable. Yun therefore suggested the following method that 'unifies' these two approaches, if the variance of the logarithm of incomes is accepted as the measure of inequality. (2)
First, consider two regression equations generating income in logarithmic form, [y.sub.t] = ln[Y.sub.t], in two periods, 0 and 1:
[y.sub.t] = [[??].sub.0t] + [[summation].sup.K-1.sub.k=1][[??].sub.k1] + [e.sub.t], t + [e.sub.t] t = 0, 1. (1)
Defining [y.sup.*] = [[??].sub.01] + [[summation].sup.K-1.sub.k=1][[??].sub.k1] [X.sub.k0] + [e.sub.0], Yun proposed using the following decomposition of the change in the variance of log income:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
where [[sigma].sup.2.sub.i] is the variance of i and [[sigma].sub.i]. j is the covariance of i and j.
The first, second and third terms on the right-hand side of this equation can be thought of as representing the characteristics effect, coefficients effect and residuals effect, respectively, and the first two of these are expressed as the sum of separate effects due to each determinant of income.
While Yun's method is suitable for evaluating the contribution of each determinant of changes in inequality within genders, it is unable to explain changes in inequality between genders. However, this can be addressed with a simple extension of the original approach, since it is possible to divide the total level of wage inequality in any period into within-gender inequality, [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], and between-gender inequality, [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3)
Within-gender inequality can be expressed as follows, where p is the proportion of males in the sample and [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] and [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] are the variances of [y.sub.t] among men and women, respectively: (3)
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (4)
Consequently, the change in between-gender inequality between periods 0 and 1 can be found if p is time invariant:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (5)
Since each of the three terms on the right-hand side of equation (5) can be decomposed as in equation (2), the change in between-gender inequality may also be expressed as the sum of a characteristics effect, coefficients effect and a residuals effect. These measure the contribution of each source of inequality to the change in the average income difference between men and women.
This paper uses individual-level data drawn from the Household Labour Force Survey and the New Zealand Income Survey. The Household Labour Force Survey is a quarterly survey that provides a range of statistics on the employed, the unemployed and those not in the labour force, who collectively comprise the working-age population. The target population for the survey is the civilian non-institutionalised usually resident New Zealand population aged 15 and over. Each quarter, a representative sample of approximately 15,000 households and 30,000 individuals is surveyed.
Household questionnaires and individual questionnaires for all working-age household members collect information on labour force status, hours worked and educational status, along with basic demographic information, but no wage or income information. Since 1997, however, the June quarter Household Labour Force Survey has included a supplemental questionnaire known as the New Zealand Income Survey, which collects information on pre-tax income from self-employment, wages and salaries, government transfers and other sources. This study will examine usual hourly earnings, defined as total earnings usually received from all wage and salary jobs.
As an indication of changes in the dispersion of earnings in New Zealand, Figure 1 plots the variance of log hourly earnings among those with wage and salary income for each year of the Income Survey. Separate series for males and females are also presented. For comparison, corresponding values from the Household Economic Survey for 1984 to 1997 are included, as taken from Dixon (1998). (4) Surprisingly, inequality rose at approximately the same rate between 1997 to 2003 as during the earlier reform period, although male inequality in particular is highly volatile from year to year. Since 2003, inequality has fallen, driven largely by a decrease in inequality among women. The dispersion of earnings among men has been consistently greater than that among women. The gap in inequality between the sexes reached a maximum in the early 1990s and has been roughly constant during the period covered by the Income Survey. The inequality series for all employees in Figure 1 reflects the fact that women comprised a steadily increasing fraction of the workforce between 1984 and 2008.
For the analysis in this paper, data are drawn from the June 1998 and June 2008 Household Labour Force Surveys and Income Surveys for all wage and salary earners aged 25 to 59. (5) The final sample has 20,853 observations, which exclude the self-employed, the unemployed and those not in the labour force. (6) All summary statistics for the sample and the regression results in the next section are adjusted using the survey weights created by Statistics New Zealand to increase the representativeness of the samples given the realities of non-random survey response.
Table 1 presents summary statistics for some of the key demographic characteristics used in the analysis, as well as the earnings variables. The first two columns describe the characteristics of the male sample in 1998 and 2008, respectively. On average, male wage and salary earners had a usual wage of approximately $18.00 in 1998, which rose to $20.80 in 2008, after adjusting for inflation. Over this period, female hourly earnings also rose from $14.60 to $17.30, as seen in the last two columns of Table 1. Among both men and women, there were increases in the average age of individuals in the sample, the proportion with a post-school qualification and the proportion of non-Pakeha. (7)
[FIGURE 1 OMITTED]
Figure 2 depicts kernel density estimates for the distribution of usual hourly earnings among wage and salary earners in 1998 and 2008, to illustrate the shape of the earnings distribution. Overall, the distribution was somewhat more positively skewed in 2008 than a decade earlier. There was a reduction in the mass around the median wage rate but little change in the lower tail of the distribution, which is consistent with rising mean hourly earnings.
It should be stressed that the decomposition method outlined in the previous section relies on the use of the variance of log wages as the sole measure of inequality. Previous studies in New Zealand and elsewhere have suggested that this indicator may be overly restrictive, as it provides information about only one moment of the wage distribution and may miss changes that are concentrated in certain segments of the distribution (see Hyslop & Yahanpath, 2005; Dixon & Mare, 2007). A weakness of Yun's method is that it does not allow changes in alternative measures such as the 90-10 percentile ratio or Gini coefficient to be decomposed, unlike the approach taken by DiNardo et al. (1996), for example, which involves generating an entire counterfactual distribution to compare with the actual distribution.
Nevertheless, comparing the variance of log earnings with the 90-10 (or 90-50) ratio between 1998 and 2008 suggests that the former is a reasonably representative measure of inequality in New Zealand. Both measures follow a similar trend, with the spike in 2003 due to an increase in earnings at the top end of the earnings distribution. The variance of log earnings does not follow the trend in the 50-10 ratio as closely, with the latter remaining relatively constant over the sample period. Furthermore, although Figure 2 indicates that taking logs moves the earnings distribution closer to a normal distribution, log earnings are still positively skewed. Therefore, the results presented in the next section will be driven most by changes in the earnings of higher-paid workers between 1998 and 2008 and may do a poorer job explaining changes in inequality among lower-paid workers over this period.
[FIGURE 2 OMITTED]
Table 2 presents the results of estimating earnings equations separately for each gender and for 1998 and 2008 using feasible generalised least squares. The explanatory variables consist of age and age squared, eight dummy variables for educational attainment, eight dummy variables for ethnic group, 11 dummy variables for region, eight dummy variables for major occupational group and 16 dummy variables for major industrial group, along with dummy variables for New Zealand-born and full-time workers. (8)
All explanatory variables have coefficients with the expected signs and most regressors, except some of the industry and region variables and the dummy variables for New Zealand-born, are found to be significant at the 1% level. The effect of age on earnings is found to be stronger for men than for women, presumably reflecting the fact that age is a better proxy for actual labour market experience among men. The full-time dummy variable is found to have a significant positive effect in all cases, although it is much larger for men. The wage premium from having a bachelor degree was lower for women than for men in both years; however, the gap narrowed substantially over the intervening period. Furthermore, the wage-education profile is generally steeper in 2008 than in 1998, suggesting a reversal of the trend identified by Maani (1999) during the first half of the 1990s. (9)
Between 1998 and 2008, inequality, as measured by the variance of log earnings, increased from 0.184 to 0.225 among men and from 0.155 to 0.182 among women. Using Yun's procedure, these changes were decomposed into a characteristics effect and a coefficients effect for each independent variable, as reported in the first four columns of Table 3. The table also presents the sum of the characteristics and coefficients effects across all variables and the residuals effect. For men, changes in the returns to the measured characteristics of the workforce accounted for 0.0083 (21%) of the overall rise in inequality of 0.0400, while changes in the distribution of these characteristics explained 0.0073 (18%) of the rise in inequality. The remaining 0.0244 (61%) is attributable to a rise in the variance of the residuals of the earnings equation, i.e. the unexplained component of earnings. Changes in the returns to occupations (especially low-skilled occupations) was the largest contributing factor to the increase in inequality, along with changes in both the ethnic distribution (mainly the increase in the Asian share) and the coefficients on the ethnicity variables. The last factor is due mainly to a fall in the coefficient on Maori in the earnings equation. This would appear to conflict with the findings of Dixon and Mare (2007), who found a narrowing of the income gap between Maori and Pakeha between 1997 and 2003. However, analysing data from 2003 reveals that the decline in the Maori coefficient only occurred between 2003 and 2008.
Among women, exactly the same share of the increase in inequality was left unexplained by the observed wage determinants as for men. However, of the remainder, the characteristics effect accounted for a greater fraction of the inequality change than for men (27%), while the coefficients effect accounted for a smaller fraction (12%). As with men, increases in the number of Asians contributed to the increase in inequality; however, other important factors were the shifts towards more education and high-skilled occupations.
For both men and women, the fractions of the overall inequality change attributed to the characteristics, coefficients and residuals effects are remarkably similar to those Dixon (1998) found for the 1984-1997 period, even though Dixon used weekly rather than hourly earnings and decomposed the standard deviation, rather than the variance, of log earnings. In both cases, changes in coefficients affect inequality among men more than women, while changes in characteristics have a larger effect on female inequality.
In contrast to earnings inequality within each gender, between-gender inequality, as defined in equation (5), fell from 0.010 to 0.007 between 1998 and 2008. (10) The final two columns of Table 3 present the results of decomposing this change into characteristics and coefficients effects. Changes in estimated returns, particularly on occupation and industry, explained most of the reduced inequality between men and women. The overall conclusion is that the female earnings distribution is moving closer to the male distribution and that both are becoming more dispersed.
The Blinder-Oaxaca decomposition is often used to examine differences in wages between men and women and for comparison purposes, this decomposition can be used on the pooled sample for 1998 and 2008. This reveals that of the total difference in log wages between men and women, 78% is due to differences in coefficients between men and women and 31% is due to the interaction between coefficients and characteristics, while differences in characteristics alone predict a 9% lower wage gap. This is consistent with the findings in Table 3, which indicate that the coefficients effect dominates, while the characteristics effect has a negative sign.
Yun's method does not include the calculation of standard errors; however, to provide some rough evidence of the significance of each contributor to the total change in inequality, bootstrapped standard errors were calculated. Table 3 indicates which effects were significant at the 10% level using these. Changes in the ethnic distribution had a significant effect on inequality for both men and women, while changes in the coefficients on ethnicity significantly increased inequality both within each gender and between genders. Variation over time in the occupation coefficients had a significant effect on male inequality, as did the coefficients on age and age squared on female inequality. Changes in the age composition and the returns to age also had a significant effect on between-gender inequality.
As noted earlier, the overall increase in inequality between 1998 and 2008 reflects a large increase in inequality between 1998 and 2003 followed by a decrease in the next half-decade. To examine the sources of this fluctuation in inequality, Table 4 reports the results of decompositions for the two 5-year intervals. Most of the increase in inequality between 1998 and 2003 was unexplained, especially for women. However, changes in coefficients also contributed to the inequality growth among both sexes and most of the overall coefficients effect for 1998-2008 appears to reflect changes before 2003. Most of the overall decrease in between-gender inequality over the decade also seems to have taken place between 1998 and 2003.
This paper has used individual-level observations from the New Zealand Income Survey to evaluate the change in earnings inequality between 1998 and 2008, both among men and women and between genders. A straightforward method recently proposed by Yun (2006) was used to decompose the total change in inequality into the portion explained by changes in the measured characteristics of the population, the portion explained by changes in the returns to these characteristics and the portion that is unexplained by these attributes. A simple way to extend Yun's approach to provide estimates of the contributions to changes in inequality between groups, such as gender, was also provided.
Among males, inequality, as measured by the variance of log earnings, was found to have risen by 22% over the sample period, while among females inequality rose by 18%. In both cases, around 60% of this change was due to unobserved factors rather than shifts in demographic characteristics or returns. Of the remainder, changes in observed characteristics accounted for around 18% of the increase in inequality for men and 27% of the increase for women. Shifts in the distributions of ethnicity, education, occupation and industry were found to be the most important factors.
Changes in the returns to observed characteristics accounted for around 21% of the increase in earnings inequality among men and 12% of the increase among women. For men, changing returns to ethnicity and occupation were found to have the largest effect, whereas for women, changes in the wage effects of education were most important and returns to occupation reduced the level of inequality. Between-gender inequality fell between 1998 and 2008, largely due to shifts in the returns to industry and occupation, while changes in the returns to ethnicity worked in the opposite direction.
Repeating the decomposition for 1998-2003 and 2003-2008 reveals that inequality peaked in 2003, a fact that was largely unexplained by either changes in characteristics or returns. Most of the increases in inequality due to changing coefficients occurred during the first half-decade, while shifts in characteristics contributed to inequality during both periods.
Appendix. Categories for the demographic variables
Educational attainment: No qualification; School certificate; Sixth form certificate; Higher school certificate; Other school qualification; Vocation or trade qualification; Bachelor or higher degree; Other post-school qualification; Unspecified qualification.
Ethnic group: Only European/Pakeha; only Maori; only Pacific; only Asian; only other; Maori/Pacific; Maori/non-Pacific; Pacific/non-Maori; non-Pacific/non-Maori.
Industry: Agriculture, forestry and fishing; Mining; Manufacturing; Electricity, gas and water supply; Construction; Wholesale trade; Retail trade; Accommodation, cafes and restaurants; Transport and storage; Communication services; Finance and insurance; Property and business services; Government administration and defence; Education; Health and community services; Cultural and recreational services; Personal and other services.
Occupation: Legislators, administrators and managers; Professionals; Technicians and associate professionals; Clerks; Service and sales workers; Agricultural and fishery workers; Trades workers; Plant and machine operators; Elementary occupations.
Region: Northland; Auckland; Waikato; Bay of Plenty; Gisborne/Hawke's Bay; Taranaki; Manawatu-Wanganui; Wellington; Nelson/Tasman/Marlborough/West Coast; Canterbury; Otago; Southland.
The author would like to thank the New Zealand Department of Labour for providing access to the Income Survey data used and, in particular, Steve Stillman and Sylvia Dixon for their advice and assistance with coding. Helpful comments were also received from Gary Fields, Larry Kahn and Dave Mare. Access to the official data used in this study was granted by Statistics New Zealand under conditions designed to give effect to the confidentiality provisions of the Statistics Act 1975. The results presented in the paper are the work of the author, not Statistics New Zealand.
Bakker, A., & Creedy, J. (1999). Macroeconomic variables and income inequality in New Zealand: An exploration using conditional mixture distribution. New Zealand Economic Papers, 33, 59 79.
Blinder, A.S. (1973). Wage discrimination: Reduced form and structural estimates. Journal of Human Resources, 8, 436455.
Crawford, R., & Johnston, G. (2004). Household incomes in New Zealand." The impact of market, taxes' and government spending, 1987/88-1997/98. Treasury Working Paper 04/20, New Zealand Treasury, Wellington.
DiNardo, J., Fortin, N., & Lemieux, T. (1996). Labor market institutions and the distribution of wages, 1973 1992: A semiparametric approach. Econometrica, 64, 1001-1044.
Dixon, S. (1998). Growth in the dispersion of earnings: 1984-97. Labour Market Bulletin, 1998, 71-107.
Dixon, S., & Mare, D. (2007). Understanding changes in Maori incomes and income inequality 1997 2003. Journal of Population Economies, 20, 571-598.
Fields, G.S. (2003). Accounting for income inequality and its change: A new method, with application to the distribution of earnings in the United States. In S.W. Polachek (Ed.), Research in labor economics, Volume 22. Greenwich, CT: Elsevier.
Gang, I.N., & Yun, M.-S. (2003). Decomposing male inequality change in East Germany during transition. Schmollers Jahrbuch (Journal of Applied Social Science Studies), 123, 43-54.
Hyslop, D.R., & Mare, D.C. (2005). Understanding New Zealand's changing income distribution, 1983-1998: A semi-parametric analysis. Economica, 72, 469-496.
Hyslop, D., & Yahanpath, S. (2005). Income growth and earnings variations in New Zealand, 1998-2004. Treasury Working Paper 05/11, New Zealand Treasury, Wellington.
Juhn, C., Murphy, K.M., & Pierce, B. (1993). Wage inequality and the rise in returns to skill. Journal of Political Economy, 101, 410-442.
Maani, S. (1999). Private and public returns to investments in secondary and higher education in New Zealand over time: 1981-1996. Treasury Working Paper 99/2, New Zealand Treasury, Wellington.
Martin, B. (2002). Change in the distribution of income of the working-age population, 1976-1996. Labour Market Bulletin, 2000-02 Special Edition, 132-156.
Oaxaca, R. (1973). Male-female wage differentials in urban labor markets. International Economic Review, 14, 693-709.
O'Dea, D. (2000). The changes in New Zealand's income distribution. Treasury Working Paper 00/13, New Zealand Treasury, Wellington.
Podder, N., & Chatterjee, S. (2002). Sharing the national cake in post-reform New Zealand: Income inequality trends in terms of income sources. Journal of Public Economics, 86, 1-27.
Yun, M.-S. (2006). Earnings inequality in USA, 1969-1999: Comparing inequality using earnings equations. Review of Income and Wealth, 52, 127 144.
(1.) O'Dea (2000) provides an exhaustive summary of studies of the changes in New Zealand's income distribution up to that time.
(2.) See Gang and Yun (2003) for a convenient explanation of this method.
(3.) Note that this implies that [[[sigma].sup.2].sub.y,B] = [[bar.y].sup.2.sub.Ft] [[bar.y].sup.2.sub.t]
(4.) The Income Survey series is not strictly comparable with the Household Economic Survey series as the latter has a much smaller sample size. In addition, Dixon documented evidence that the Household Economic Survey features a considerably higher fraction of employees working long weekly hours than the Income Survey.
(5.) The 1997 data were avoided due to concerns about their quality, given that this was the first year of the Income Survey.
(6.) In the analysis of the following section, 21 observations are excluded from the sample due to missing values for the control variables.
(7.) Table 1 uses Statistics New Zealand's hierarchical method of classifying ethnic groups; however, a full set of mixed major ethnic group variables is included in the regressions in the next section.
(8.) The appendix provides a full description of the education, ethnic group, industry, occupation and region variables that were used.
(9.) Maani focused on the employed, which includes the self-employed as well as wage and salary earners, and she used annual income instead of hourly earnings.
(10.) The average proportion of males over the two years was used as the value for p in equation (5).
Kerry L. Papps *
Nuffield College, University of Oxford, OX1 1NF, UK and IZA, Bonn, Germany (Received 14 May 2009;final version received 3 May 2010)
* Email: kerry.papps(a)economics.ox.ac.uk
Table 1. Sample characteristics for 1998 and 2008. Men Variable 1998 2008 Age 39.337 40.862 No qualification 0.193 0.171 School qualification 0.184 0.171 Post-school qualification 0.621 0.637 Pakeha only 0.822 0.719 Maori 0.087 0.101 Pacific Islander 0.040 0.054 Asian 0.028 0.101 New Zealand born 0.800 0.719 Full-time 0.950 0.946 Usual hourly wage 18.0 20.8 Actual weekly income 785 887 Number of observations 4,880 5,430 Women Variable 1998 2008 Age 40.325 41.725 No qualification 0.202 0.158 School qualification 0.251 0.205 Post-school qualification 0.543 0.612 Pakeha only 0.832 0.733 Maori 0.081 0.101 Pacific Islander 0.044 0.051 Asian 0.026 0.094 New Zealand born 0.804 0.745 Full-time 0.655 0.706 Usual hourly wage 14.6 17.3 Actual weekly income 497 610 Number of observations 4,850 5,690 Notes: Summary statistics are weighted by the Household Labour Force Survey sampling weights. Wages and incomes are in constant (June 1998) dollar values, adjusted using the Consumer Price Index. Actual weekly income includes income from all sources (at the person's last pay). Table 2. Results of estimating earnings equations for 1998 and 2008. Males Regressor 1998 2008 Age 0.052 * (0.006) 0.046 * (0.006) Age squared (in 100s) -0.057 * (0.007) -0.049 * (0.007) Full-time 0.163 * (0.036) 0.152 * (0.046) Attained school 0.076 * (0.024) 0.075 * (0.027) certificate Attained sixth form 0.122 * (0.031) 0.135 * (0.035) certificate Attained higher school 0.109 (0.060) 0.203 * (0.036) certificate Attained other school 0.017 (0.043) 0.129 * (0.046) qualification Attained vocation or 0.146 * (0.016) 0.165 * (0.017) trade qualification Attained bachelor or 0.314 * (0.029) 0.314 * (0.027) higher degree Other post-school 0.210 * (0.029) 0.094 * (0.034) qualification Unspecified 0.045 (0.101) -0.106 (0.044) qualification New Zealand born 0.001 (0.019) 0.009 (0.020) [R.sup.2] 0.308 0.324 Number of 4,880 5,430 observations Females Regressor 1998 2008 Age 0.007 (0.005) 0.026 * (0.005) Age squared (in 100s) -0.007 (0.007) -0.028 * (0.007) Full-time 0.035 * (0.013) 0.043 * (0.014) Attained school 0.048 (0.019) 0.075 * (0.023) certificate Attained sixth form 0.102 * (0.023) 0.126 * (0.025) certificate Attained higher school 0.121 * (0.045) 0.115 (0.027) certificate Attained other school 0.052 (0.032) -0.014 (0.034) qualification Attained vocation or 0.101 * (0.016) 0.118 * (0.015) trade qualification Attained bachelor or 0.254 * (0.025) 0.283 * (0.020) higher degree Other post-school 0.097 * (0.025) 0.104 * (0.032) qualification Unspecified -0.074 (0.166) 0.157 * (0.040) qualification New Zealand born -0.019 (0.017) -0.023 (0.019) [R.sup.2] 0.341 0.347 Number of 4,850 5,680 observations Notes: A full set of dummy variables for eight ethnic groups, eight occupational groups, 16 industrial groups and 1 l regions was also added to each regression, although the estimated coefficients are omitted. Feasible generalised least squares estimation was used with the Household Labour Force Survey sampling weights. Standard errors are shown in parentheses; * denotes significance at the 1 % level. Table 3. Decomposition of changes in earnings inequality between 1998 and 2008. Males [DELTA]x [DELTA][??] Source of inequality effect effect Age and age squared -0.0002 0.0000 Full-time 0.0004 -0.0002 New Zealand born 0.0001 0.0000 Education dummy 0.0018 0.0004 vector Ethnicity dummy 0.0033 * 0.0023 * vector Occupation dummy 0.0010 0.0056 * vector Industry dummy 0.0018 0.0005 vector Regional dummy -0.0009 0.0004 vector All variables 0.0073 0.0083 Residuals 0.0244 * Total 0.0400 * Females [DELTA]x [DELTA][??] Source of inequality effect effect Age and age squared 0.0002 0.0007 * Full-time -0.0002 0.0003 New Zealand born -0.0001 0.0000 Education dummy 0.0033 0.0017 vector Ethnicity dummy 0.0016 * 0.0009 * vector Occupation dummy 0.0022 -0.0012 vector Industry dummy 0.0013 0.0005 vector Regional dummy -0.0008 0.0003 vector All variables 0.0073 0.0032 Residuals 0.0167 * Total 0.0272 * Between genders [DELTA]x [DELTA][??] Source of inequality effect effect Age and age squared 0.0002 * 0.0007 * Full-time -0.0001 -0.0001 New Zealand born 0.0000 0.0001 Education dummy -0.0006 -0.0004 vector Ethnicity dummy -0.0020 0.0032 * vector Occupation dummy 0.0017 -0.0048 vector Industry dummy 0.0004 -0.0022 vector Regional dummy 0.0010 0.0006 vector All variables 0.0004 -0.0026 Residuals -0.0010 Total -0.0031 Note: [DELTA]x effect denotes the characteristics effect; [DELTA][??] effect denotes the coefficients effect. * denotes that the effect was significant at the 10% level using bootstrapped standard errors, with 50 replications. Table 4. Decomposition of changes in earnings inequality between 1998 and 2003 and 2003 and 2008. Males [DELTA]x [DELTA][??] Source of inequality effect effect 1998-2003 Age and age squared -0.0004 -0.0001 Full-time 0.0003 0.0018 New Zealand born -0.0001 0.0003 Education dummy vector 0.0001 0.0031 Ethnicity dummy vector 0.0008 0.0013 Occupation dummy vector 0.0000 0.0065 Industry dummy vector 0.0005 0.0019 Regional dummy vector 0.0000 0.0012 All variables 0.0012 0.0160 Residuals 0.0575 Total 0.0747 2003-2008 Age and age squared 0.0001 0.0001 Full-time 0.0003 -0.0021 New Zealand born 0.0001 -0.0002 Education dummy vector 0.0017 -0.0028 Ethnicity dummy vector 0.0025 0.0010 Occupation dummy vector 0.0008 -0.0007 Industry dummy vector 0.0008 -0.0010 Regional dummy vector -0.0009 -0.0008 All variables 0.0052 -0.0063 Residuals -0.0331 Total -0.0343 Females [DELTA]x [DELTA][??] Source of inequality effect effect 1998-2003 Age and age squared 0.0000 0.0011 Full-time -0.0001 0.0014 New Zealand born 0.0000 -0.0001 Education dummy vector 0.0015 0.0002 Ethnicity dummy vector 0.0003 0.0009 Occupation dummy vector 0.0015 -0.0022 Industry dummy vector 0.0009 0.0007 Regional dummy vector -0.0002 0.0008 All variables 0.0040 0.0029 Residuals 0.0782 Total 0.0851 2003-2008 Age and age squared 0.0001 -0.0004 Full-time -0.0002 -0.0011 New Zealand born -0.0001 0.0000 Education dummy vector 0.0014 0.0018 Ethnicity dummy vector 0.0013 0.0000 Occupation dummy vector 0.0008 0.0008 Industry dummy vector 0.0002 0.0000 Regional dummy vector -0.0007 -0.0004 All variables 0.0028 0.0008 Residuals -0.0616 Total -0.0579 Between genders [DELTA]x [DELTA][??] Source of inequality effect effect 1998-2003 Age and age squared 0.0001 0.0009 Full-time -0.0002 -0.0010 New Zealand born 0.0000 0.0000 Education dummy vector 0.0000 -0.0010 Ethnicity dummy vector -0.0001 -0.0002 Occupation dummy vector -0.0001 0.0019 Industry dummy vector 0.0001 -0.0005 Regional dummy vector 0.0000 -0.0001 All variables -0.0002 -0.0001 Residuals -0.0037 Total -0.0038 2003-2008 Age and age squared 0.0000 -0.0002 Full-time 0.0000 0.0015 New Zealand born 0.0000 0.0000 Education dummy vector -0.0002 0.0003 Ethnicity dummy vector -0.0011 0.0018 Occupation dummy vector 0.0009 -0.0021 Industry dummy vector -0.0001 0.0001 Regional dummy vector 0.0008 0.0004 All variables 0.0003 0.0019 Residuals -0.0015 Total 0.0007 Note: [DELTA]x effect denotes the characteristics effect; [DELTA][??] effect denotes the coefficients effect.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||RESEARCH ARTICLE|
|Author:||Papps, Kerry L.|
|Publication:||New Zealand Economic Papers|
|Date:||Dec 1, 2010|
|Previous Article:||Reflections of a journal editor.|
|Next Article:||The economic impact of the New Zealand fiscal stimulus package.|