RDP 2002-04: Labour Market Adjustment in Regional Australia 3. Regional Labour Markets in Australia

3.1 What is a Region?

Before investigating regional labour markets in Australia, how should we define our basic unit of analysis? The regional science literature has settled on the following three methods for delimiting the boundaries of a region (Richardson 1973).

Homogeneity: Areas form an economic region if they are homogenous with respect to a key economic element, such as their industry structure. This key element should vary significantly more between regions than it does within regions.

Nodality: Areas form a region if they comprise a single labour market. The boundary of the region is the outer limits over which people can commute to the central location of economic activity.

Programming: Regions comprise administrative and political areas (such as municipalities, electorates or statistical areas) for which data are collected.

Given the practical nature of our investigation into regional labour markets, we must first consider regions for which data are collected. In Australia, economic data for sub-state areas are generally available over three different levels of disaggregation: Statistical Local Areas (SLAs) which are the finest level of disaggregation, Statistical Sub-Divisions (SSDs), and the more aggregated Statistical Divisions (SDs).^[9]

In our analysis, we choose the SLA as the basic unit, because its finer level of disaggregation permits consideration of both homogeneity and nodality when choosing the boundaries of a region. At this level, it is easier to ensure that the region has a common set of economic characteristics and forms a single labour market than is the case for the much larger statistical sub-divisions and statistical divisions.^[10] Furthermore, where SLAs are suburbs of larger cities, they can be aggregated to form a region. We classify all cities with multiple SLAs as single regions, a process that leaves us with a sample of 637 regions.

3.2 Regional Economic Performance between 1986 and 1996

Between 1986 and 1996 the number of people employed in Australia grew by about 17 per cent, or at an average annual rate of about 1.6 per cent. However, as shown in Figure 1, the rate of employment growth was not evenly distributed throughout Australia's regions. Median growth over the ten years was about 5 per cent (or ½ per cent per annum) and, in about 40 per cent of regions, the level of employment actually fell.^[11]

Figure 2 demonstrates that employment growth was also geographically concentrated. The white areas represent regions that experienced employment growth between 1986 and 1996 and the shaded areas represent regions that experienced employment contractions. Regions experiencing employment growth tended to be located along Australia's eastern seaboard, tended to be close to capital cities, or were in remote mining locations. On the other hand, regions experiencing falling employment were mainly rural regions in Australia's interior.

An alternative indicator of regional economic performance is the unemployment rate. Figure 3 shows that, like employment growth, unemployment rates were highly dispersed around the national average in 1996.^[12] However, the geographic distribution of regional unemployment rates looks very different.

The white areas of Figure 4 indicate regions that recorded ‘low’ unemployment in 1996, while the shaded areas recorded ‘high’ unemployment.^[13] In fact, the shading of the eastern states in Figure 4 is almost the negative image of Figure 2, suggesting that regions that recorded employment growth between 1986 and 1996, tended to have relatively high unemployment rates at the end of the period.^[14] Similarly, regions that experienced falling employment tended to have low unemployment rates at the end of the period. This raises questions about how Australia's regional labour markets adjust to changing economic conditions.

3.3 Labour Market Adjustment in Australia's Regions

Typically, we expect nations or states with rapidly expanding employment to experience larger falls in the unemployment rate than those with weak employment growth. This, however, was often not the case in regional Australia between 1986 and 1996. Figure 5 shows the pairwise combination of employment growth and the change in the unemployment rate for each region. The observations form a large cluster rather than a line, confirming that there is only a weak correlation between employment growth and the change in the unemployment rate at the regional level.^[15] This suggests that regions may have adjusted to shocks in different ways.

To understand why this may have been the case, we can see from Figure 5 that the observations fall into four quadrants:

Quadrant 1: falling employment and rising unemployment
Quadrant 2: falling employment and falling unemployment
Quadrant 3: rising employment and falling unemployment
Quadrant 4: rising employment and rising unemployment

While regions in Quadrants 1 and 3 display the expected inverse relationship between employment growth and unemployment rates, this is not so in the other quadrants. Unexpectedly, we observe a large number of regions where unemployment rates fell, even though employment fell over the full ten-year period (Quadrant 2). The labour market in these regions must have adjusted more through a combination of falling participation rates and out-migration. Similarly, a large number of regions unexpectedly fell into Quadrant 4, where rising employment over the ten-year period was associated with rising unemployment rates. The labour market in these regions must have adjusted more through a combination of rising participation rates and in-migration.

But which adjustment mechanism was more important? Even though regional migration data are not published in the census (on a consistent basis through time), we make inferences about inter-regional migration based on changes in the regional population and labour force.^[16] If we look to Table 1, it seems clear that differences in the strength of the migration response to shocks determined the path of unemployment. In the regions where both the unemployment rate and employment fell (Quadrant 2), there tended to be a proportionately larger fall in the population, implying an important role for out-migration. Similarly, in regions where the unemployment rate increased while employment was growing (Quadrant 4), there tended to be a proportionately larger increase in the population, implying an important role for in-migration.^[17] Differences in the average response of participation rates were, by comparison, much smaller than the changes in population.^[18]

Such divergent regional labour market performance presents some interesting questions. Why, over a ten-year period, has employment grown so markedly in some regions and contracted in others? And how can we explain the regional migration response to such regional variations in conditions? We tackle these questions in turn.

Footnotes

In Australia, there are 1,337 SLAs, 181 SSDs, and 68 SDs. [9]

SDs and SSDs are usually too large geographically for regular commuting and so cannot be characterised as having a single labour market. Furthermore, they are usually sufficiently large to contain multiple nodes of economic activity. [10]

In our sample of 637 regions, just over 200 had populations in 1986 of below 5,000. Because of the low average employment growth rates of these smaller regions, the unweighted median employment growth rate is less than the population-weighted median employment growth rate. That the superior performance of Australia's larger regions is able to mask the poorer performance of smaller regions (which also presumably have different economic characteristics than the larger regions) reinforces the desirability of giving equal weight to each region in our examination of relative regional economic performance. [11]

Note that the dispersion of unemployment rates fell between 1986 and 1996. However, the reduction was small, so that there was only a slight narrowing of unemployment differentials during this period. [12]

A high (low) unemployment rate is defined here as that greater (less) than 8.3 per cent, which was the median unemployment rate for the regions in the sample in 1996. [13]

This supports the work of McGuire (2001), who found, for Queensland Statistical Divisions, that regions with high average unemployment rates between 1991 and 2001 tended to have the highest employment growth rates during this period. [14]

In a simple bivariate regression, employment growth explained only 17 per cent of the variation in changes in unemployment rates between 1986 and 1996. [15]

ABARE (2001) did, however, obtain unpublished regional migration data from the census. Our interest was not, however, in the precise number of domestic migrants, but rather the strength of regional migration relative to regional employment growth. This can be inferred from the published population and labour force data. [16]

Note that ascribing all population growth to inter-regional migration in Quadrant 3 and 4 regions is not strictly correct, since both natural population growth and international migration also made contributions. The role of international migration can be safely ignored because the overwhelming majority of migrants settle in Australia's capital cities, which receive a small weighting in the overall trends discussed here. The role of natural increase may be more important, but due to data constraints cannot be identified. [17]

Our inference that regional migration has influenced unemployment outcomes is not based solely on the median change in labour market variables in each quadrant, but on the prevalence of these changes. For example, in the regions that experienced rising unemployment together with employment growth, 61 per cent had more rapid population growth than employment growth. This contrasts with regions experiencing unemployment falls, where only 12 per cent had more rapid population growth than employment growth. More importantly, after examining the contributions that changes in population and changes in participation made to changes in the size of the labour force, we found that in 80 per cent of cases, changes in the size of the population made a larger contribution. [18]