4  Estimating policy impacts

In order to estimate the impact of the NZS eligibility age on retirement behaviour, we need consistent data on the labour force participation patterns of older males and females covering the whole period between 1976 and 2001. Census data is available at five yearly intervals for the whole period, but two particular limitations need to be noted.

First, tabular data for the early years is available only in 5-year age groups rather than by single year of age, so the model uses these groupings, starting with those aged 45-49 years and with an upper category of age 65 and over.

Second, the definition of labour force participation has been revised over the years, reflecting first the exclusion of people in part-time employment and then a change in the number of hours per week that constituted part-time employment. For this estimation exercise the current Statistics New Zealand definition of a labour force participant has been adopted, namely a person who either works regularly for one or more hours per week or is unemployed and was seeking work during the week prior to the census. Earlier census data have been adjusted to make them as consistent as possible with this definition.[22]

Figure 7 and Figure 8 illustrate how the labour force participation rates of males and females in different age groups have changed over the period 1976 to 2001. It is these data that the following model seeks to explain. Key features to note are: the trend for participation rates for males to generally fall over time, while rates for females generally rise; and the changing trend in participation rates within the 60-64 year age group compared with younger age groups.

Figure 7 - Labour force participation rates of males, by age group, 1976-2001

Source: Census documents and author’s calculations.

Figure 8 - Labour force participation rates of females, by age group, 1976-2001

Source: Census documents and author’s calculations.

4.1  The model

The determinants of labour force participation rates among older people have been modelled as follows.

(1)     P_gjt = a + bU_t + cE_gjt + dD_gj + ε_gjt

where P_gjt is the labour force participation rate of a particular gender (g) and age group (j) in a census year (t); U_t is the general unemployment rate in that year; and E_gjt indicates the proportion of each gender and age group that is eligible to receive NZS in census year t[23]. D_gj is a vector of dummy variables that capture age group and gender influences and ε_gjt is a normally distributed random error term.[24]

The eligibility variable, E, is interacted with the gender variable to allow for differential responses by males and females to being eligible for NZS. In addition, the model allows for the possibility that females who are within five years of being eligible for NZS might retire at the same time as their (older on average) husbands, since there is some financial assistance for those who do so.[25]

Separate age-group dummies for males and females have been used to pick up possible gender differences in age/retirement profiles and to avoid imposing any particular functional form on the profile. This specification constrains the profiles to be the same across different cohorts. However, the long-term trend of rising female labour force participation has been captured using a logarithmic time trend in age groups that are not eligible for NZS.[26]

The estimated model uses data covering 6 census years (1976, 1981, 1986, 1991, 1996, 2001), 2 genders and 5 age groups (45-49, 50-54, 55-59, 60-64, 65 and over), giving a total of 60 observations. The estimated coefficients are shown in Table 3 and all are statistically significant at the 99% level, except for one age-group dummy.

Dependent variable: age group/gender specific %age participating in the labour force in census year t.

N=60; Adjusted R squared = 0.9928.

The underlying participation profiles for men and women as they age can be estimated by combining the constant term and the average unemployment effect with the values for the age/gender dummy variables – these are shown in Figure 9.

Figure 9 - Estimated shape of the age/ participation curves for males and females

Notes: (1) Calculations include the estimated effect of unemployment, based on the average unemployment rate of 4.6% for the six census observations. (2) Female participation trends include a log time trend interacted with being ineligible for NZS.

For men, abstracting from NZS eligibility, the pattern is for a decelerating rate of participation, i.e. an accelerating rate of retirement, from age 45 onwards, with no particular ‘kink’. This is what one would expect as a result of factors such as maturing private retirement savings or rising rates of health problems that trigger retirement decisions. However, when eligibility for NZS is triggered, the estimated coefficient on the eligibility variable in Table 3 indicates that the participation rate drops by a further 21 percentage points (this is not shown in Figure 9).

The picture is more complicated in the case of women, since part of the impact of NZS eligibility shows up in Figure 9 through the female participation time trend term, which stops applying once a female becomes eligible for NZS. In addition there is a direct depressive effect of a further 7 percentage points (-21.44 +14.54 from Table 3) on women’s labour force participation rates after they become eligible for NZS. This second effect is not shown in Figure 9.

The total percentage point effect on participation rates of NZS eligibility turns out to be similar for women and men. For example, in 1991 NZS eligibility was set at age 60. This model suggests that if the eligibility age had in fact been 65 at that time, the labour force participation rates of 60-64 year old males and females in 1991 would each have been about 21 percentage points higher.[27] Of course these percentage point changes would have translated into different total percentage changes for men and women (53% and 170% respectively) because of their different base levels of participation.

Finally, the results in Table 3 also indicate that there is a drop in the average participation rates of women if they are within five years of becoming eligible for NZS in their own right. This effect is estimated to lower participation in the near-eligible age band by 11 percentage points. This is most likely to reflect ‘joint retirement’ choices by couples when the (typically older) husband starts to receive NZS.