4  Alternative specifications

In this section we experiment with a number of alternative specifications, formed by systematically varying the underlying assumptions. Table 8 summarises the specifications and results.

In the benchmark specification, the increase in employees comes from labour force participation rates rising while unemployment rates remain constant. It represents a situation where more people join the labour market, and the employment prospects of those in the labour market remain constant. In Alternative 1, the increase in employees comes from unemployment rates falling while labour force participation rates remain constant. This specification represents a situation where the labour force stays the same size, but the employment prospects of those already in the labour market improve.

Figure 3 shows unemployment rates in New Zealand and summary measures for the sample of OECD countries. To generate Alternative 1, we set the unemployment rate, both for women aged 25-34 and for the population aged 15 and over, equal to the average of the 5 countries with the lowest unemployment rates.

Figure 3 – Unemployment rates, females aged 25-34, and population aged 15-64

Women aged 25-34

Population aged 15-64

 

Source: Calculated from data in the OECD Labour Market Data online databases.

Notes – The sample of OECD countries excludes the following: Hungary, Korea, Luxembourg, Mexico, the Slovak Republic, and Turkey. Employment rates are defined as the percent of people in a given age group who work at least one hour a week.

As is apparent in Table 8, the associated increase in GDP is much smaller than in the benchmark specification. The main reason is that unemployment is already low, so reducing it further generates only a small number of extra employees (2,800 in the “young women” scenario, and 46,400 in the “overall” scenario.)

Alternative 2 shows the effect of following the conventional practice of assuming that the average new employee is identical to the average existing employee, or, equivalently, that GDP increases one-for-one with employment. The rise in GDP under this assumption is considerably higher than in the benchmark specification. This suggests, as expected, that the characteristics of new workers drawn into employment matters a great deal for economic growth. It also suggests that our benchmark result can be expressed as follows: GDP does not increase one-for-one with employment.

Alternatives 3 and 4 investigate why Alternative 2 gives higher results than the benchmark scenario. Alternative 3 retains the benchmark assumption that extra employees have the same productivity as people of the same age and sex who are not in employment, but follows Alternative 2 in assuming that the extra employees work the same number of hours as the average existing employee. This choice of assumptions gives a slightly higher GDP increase than in the benchmark specification. Differences between the hours worked by new employees and existing employees is evidently one reason why GDP does not increase one-for-one with employment, but is not the main reason.

If the main reason is not hours, it must be productivity. Alternative 4 bears this out. It assumes that the extra employees have the same productivity as the average existing employees, while working different hours. It yields almost as high an increase in GDP as Alternative 2. It is clear that the relative productivity of extra and existing workers matters—there is a 40% difference in the increase GDP between the benchmark and Alternative 4 for the overall scenario, accounted for by the lower productivity of the extra workers. The difference for the “young women” scenario is 38%.

Alternative 5 applies only to the “overall” scenario. The increase in the participation rate for the population aged 15 and over is identical to the benchmark scenario. However, the breakdown of participation by age and sex is different. Whereas the age-specific rates in the benchmark scenario were obtained by adding 5.1 points to the actual New Zealand rates, the age-specific rates in Alternative 5 were obtained by making an adjustment to Swedish rates (Appendix 3 describes the calculations in detail). Sweden was chosen for two reasons. First, its overall rate was very close to the target participation rate of 71.1%. Second, and more important, the age and gender profile for labour force participation in 2001 was markedly different from New Zealand, with much lower rates for men and women at younger ages, and much higher rates for women aged 25 and over (see Figure 4). Sweden also has early retirement from the labour market and low youth employment.

Figure 4 – Age-specific labour force participation rates, New Zealand 2001

Males

Females

 

Source: Calculated from data in the OECD Labour Market Data online database.

Alternative 5 therefore tests whether the results on GDP increases are sensitive to assumptions about the age and gender of the new employees. As Table 8 shows, the GDP increase is larger under Alternative 5 than it is in the benchmark case. These results evidently are sensitive to assumptions about the age and gender of new employees. The reason why GDP is higher under Alternative 5 is that it implies a decrease in proportional share of low-productivity younger employees.