Australian labour market flows over the business cycle.Abstract
This paper analyses the behaviour of Australian labour market transition rates. Since the early 1980s, the job-finding rate has been significantly more volatile than the job-loss rate and it is strongly pro-cyclical. The economic downturns of the early 1980s and early 1990s were associated with up to a 10 percentage points decline in the average job-finding rate. In comparison, the recent economic downturn was associated with a less significant decline in the job-finding rate. During these periods, the job-loss rate has shown less significant volatility. The findings of this paper suggest that the job search activities of workers are potentially more relevant in explaining the volatility of labour market variables such as the unemployment rate, and whether emerging skill shortages can be addressed. Policies that assist job search and the skills development of workers are important, as is the intensity of workers' search activity.
Increasing attention is given to the recent performance of the Australian labour market and to how much further the national unemployment rate will decline, as the economy recovers from an economic downturn associated with the global financial crisis (GFC). The re-emergence of skills shortages across a number of occupations, industries and regions is also relevant in the current environment.
The direction of the national unemployment rate and the degree to which skills shortages can be tackled partly depend on the efficiency of the labour market in 'matching' people looking for work with job vacancies (see Pissarides 2000; Mortensen and Pissarides 1994). The movement (or flows) of people between the labour market states of employed, unemployed or not in the labour force varies over the business cycle and can help to explain aggregate labour market outcomes (see, for the United States, Shimer 2005).
This paper analyses the behaviour of Australian labour market transition rates over the last 30 years using data for the flows of people between the three labour market states. A unique contribution of this paper is its examination of job-finding and job-loss rates over six identified economic downturns for Australia. Other related studies relevant to Australia include Panomareva and Sheen (2010), which examines transitions using a four-state model and finds that difficulties in finding employment--rather than job losses--are the key feature of recessions. Chindamo and Uren (2010) examine the ability of a search and matching model of labour market flows to explain the behaviour of Australian labour market variables over the business cycle. Similarly to Shimer (2005), they find that the model fails to identify substantial volatility among unemployment or vacancies. Dixon, Freebairn and Lim (2004) provide a framework for understanding the impact of labour market flows on the dynamics of the unemployment rate, including adjustment of labour market flow data to reflect stock estimates.
The present paper finds that since the early 1980s the job-finding rate is significantly pro-cyclical and that it has been much more volatile than the job-loss rate. The economic downturns in the early 1980s and early 1990s were associated with up to a 10 percentage points decline in the average job-finding rate. In comparison, the economic downturn period in 2008-09 was associated with a less significant decline in the job-finding rate. During these same episodes the job-loss rate has shown less significant movement; declines in the job-finding rate appear to be more significant in explaining aggregate labour market behaviour.
2. Labour Market Gross Flows
2.1 Gross Flows from the Matched Sample Data
The Australian Bureau of Statistics (ABS) provides estimates of movements of people between the three labour force states from one month to the next based on the Labour Force Survey, and these data are referred to as labour market gross flows. (1)
The ABS is able to match Labour Force Survey respondents who report in consecutive months se that the transition of individuals (or gross flows) between the different labour force states can be estimated. The gross flows figures represent approximately 80 percent of the Labour Force Survey sample. Table 1 provides an example of monthly labour market gross flows for Australia.
Table 1 shows that of the matched sample, there was a flow of 130 thousand persons from the not in the labour force state to unemployment from March 2010 to April 2010. This was smaller than the number of persons moving from not in the labour force directly to employment (195 thousand). The number of unemployed persons in March 2010 who moved to not in the labour force, or to employment were equal (118 thousand), while the number of employed persons who moved to unemployment was 82 thousand.
2.2 Calculating Transition Rates: Job-Finding and Job-Loss Rates
From the gross flows data we can calculate monthly transition rates between the labour market states, which provide us with a better understanding of labour market dynamics. Two key transition rates are the job-finding rate and the job-loss rate.
The average job-finding rate is defined as the flow of persons from unemployment to employment plus the flow of persons from not in the labour force to employment, divided by a measure of the non-employed at the end of the previous month (that is, persons not employed but who wanted a job at some point during the next month, including those officially classified as not in the labour force but who want a job). The denominator in the job-finding rate is calculated using the approach of Fujita and Ramey (2006) which assumes that the average job-finding rate for unemployed workers is equal to the rate for persons not in the labour force who want a job in the current month. The average job-loss rate is defined as the flow of persons from employment to unemployment plus the flow of persons from employment to not in the labour force over the current month, divided by the stock of employed persons at the end of the previous month. (2)
The matched sample is approximately 80 per cent of the corresponding Labour Force Survey sample. (3) The transition rates are adjusted by a multiplicative factor of 1.25 to account for the inability to match workers across time. A more sophisticated method involves using the residual allocation method (see Dixon, Freebairn and Lim 2004), and the X12 seasonally adjusted procedure is also applied to convert the transition rates into seasonally adjusted rates. Table 2 compares the descriptive statistics of the resulting job-finding and loss rates for Australia.
Table 2 shows that the monthly job-finding rate had averaged 25.1 per cent since the early 1980s or an average non-employment period for those looking for work of about a third of a year. (4) The job-loss rate averaged 4.9 per cent over the same period, suggesting that a job lasts for an average of 1.7 years. The next section examines the behaviour of the transition rates over time.
3. The Behaviour of the Labour Market Transition Rates
Figure 1 presents the average job-finding rate for Australia from December 1981 until December 2008. The figure also identifies periods of economic downturn, approximated as periods of two or more consecutive quarters of negative growth in domestic final demand. The periods of economic downturn identified are: December 1981 to March 1982; March 1983 to June 1983; September 1989 to December 1989; December 1990 to June 1991; September 2000 to December 2000; and December 2008 to March 2009.
The economic downturns in the early 1980s and early 1990s were associated with up to a 10 percentage points decline in the average job-finding rate. In comparison, the economic downturn in 2008-09 was associated with a less significant decline in the job-finding rate, and there was minimal movement in the job-finding rate for the downturn period in the early 2000s.
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
The average job-loss rate for Australia from December 1981 to December 2008 is plotted in Figure 2; it shows that there was a slight rise in the job-loss rate in the economic downturn during the early 1980s and again during the early 1990s. The last two downturn periods were hot associated with any significant rise in the job-loss rate.
Figure 3 shows the job-finding and loss rates together. This more clearly demonstrates that the periods of economic downturn have been associated with a significant drop in the job-finding rate, rather than a significant upward movement in the job-loss rate. Over the whole sample there is a correlation of -0.92 between the job-finding rate and the unemployment rate. In contrast, the correlation between the job-loss rate and the unemployment rate is 0.82.
[FIGURE 3 OMITTED]
The decline in the job-finding rate during the periods of economic downturn can be explained in the context of the job vacancy to unemployment ratio. During periods of economic downturn, businesses may lower their hiring intentions and hence advertise fewer vacancies. This makes the job search success of workers more difficult. Conversely, during periods of economic upturn, businesses may raise their hiring intentions and increase the number of vacancies they advertise (Pissarides 2000). In this situation, for a given number of unemployed persons, there is a greater number of job vacancies and a higher probability that a searching worker will be able to find a job. This relationship is illustrated in Figure 4 where the job vacancy to unemployment ratio (a measure of labour market tightness) is plotted against the job-finding rate. Clearly the job-finding rate is a positive function of the job vacancy to unemployment ratio.
The comparative performance of the job-finding and job-loss rates over time also suggests that factors impacting on workers' job search are particularly important in explaining the extent of unemployment. Factors include the discouraged worker effect, skills atrophy while (long-term) unemployed and transaction costs associated with job searching.
[FIGURE 4 OMITTED]
To examine the nature of the transition rates further, Figure 5 presents the cyclical components of these time series using a Hodrick Prescott filter with a smoothing parameter set at 1600. (5) Figure 5 shows that the job-finding rate is clearly procyclical. The cyclical component of the job-finding rate shows extreme volatility in the economic downturn of the early 1980s and early 1990s. The volatility is also more significant in the economic downturn during 2008-09. In contrast, the cyclical component of the job-loss rate has been much less volatile.
When comparing the two cyclical time series over the entire sample period, the cyclical component of the job-finding rate was approximately 18 times more volatile than the cyclical component of the job-loss rate. This suggests that the job search and job-finding activities of workers in the labour market are potentially more relevant in explaining the volatility of labour market variables such as the unemployment rate.
[FIGURE 5 OMITTED]
The movement of people between the three labour market states varies over the business cycle and the job-finding and job-loss rates can explain the magnitude of movements between the three states. Since the early 1980s the job-finding rate has been much more volatile than the job-loss rate, and has also been much more pro-cyclical. The economic downturns in the early 1980s and early 1990s were associated with up to a 10 percentage points decline in the average job-finding rate. In comparison, the economic downturn period in 2008-09 was associated with a less significant decline in the job-finding rate. During these same episodes, the job-loss rate has shown less significant movement. This suggests that downward movements from unemployment or not in the labour force to employment are far more important in explaining aggregate labour market behaviour--such as the unemployment rate--than flows resulting in job losses are.
The implications of the findings of this paper are that frictions in labour market search and matching are relevant to Australia's labour market performance and whether the unemployment rate can decline further after the GFC. Job search and job-finding activities of workers in the labour market are potentially more relevant in explaining the volatility of labour market variables--such as the unemployment rate--and whether emerging skills shortages can be addressed in the short term. Policies that assist in the job search and skills development of workers are important, as is the search intensity of workers.
Australian Bureau of Statistics (ABS), Labour Force, Australia, June 2010, cat. 6206.0.
Australian Bureau of Statistics (ABS), Labour Statistics: Concepts, Sources and Methods, April 2007, cat. 6102.0.55.001.
Chindamo, P. and Uren, L. (2010), 'Vacancies and Unemployment in Australia', Australian Economic Review, vol. 43, pp. 136-152.
Dixon, R., Freebairn, J. and Lim, G. (2004), 'A Framework for Understanding Changes in the Unemployment Rate in a Flows Context: An Examination Net Flows in the Australian Labour Market', accessed at <http://www.economics. unimelb.edu.au/downloads/wpapers-04/910.pdf> on 5 February 2010.
Fujita, S. and Ramey, G. (2006), 'The Cyclicality of Job Loss and Hiring', Federal Reserve Bank of Philadelphia Working Paper, no. 06-17.
Mortensen, D. and Pissarides, C. (1994), 'Job Creation and Job Destruction in the Theory of Unemployment', Review of Economic Studies, vol. 61, pp. 397-415.
Panomareva, N. and Sheen, J. (2010), 'Cyclical Flows in Australian Labour Markets', Economic Record, vol. 86, pp. 35-48.
Pissarides, C. (2000), Equilibrium Unemployment, 2nd edn., MIT Press, Cambridge.
Shimer, R. (2005), 'The Cyclical Behavior of Equilibrium Unemployment and Vacancies', American Economic Review, vol. 95, pp. 25-49.
Shimer, R. (2007), 'Reassessing the Ins and Outs of Unemployment', accessed at <http://sites.google.com/site/robertshimer/research/workingpapers> on 5 February 2010.
(1) Labour market gross flows data from 1997 were obtained from ABS (2010) and data from 1981 to 1997 were obtained from Professor Robert Dixon of The University of Melbourne.
(2) Shimer (2007) outlines a method to account for intra-period flows in the transition rates (such as flows from one employment state to another employment state within a month).
(3) Given the 80 per cent matched sample, it means that gross flows data cannot be reconciled with labour force survey stock estimates. The ABS is working on expanding the matching sample to allow such reconciliation.
(4) The average monthly job-finding rate is 0.251, implying an annual period of non-employment of (1/0.251)*12 = 0.332 years.
(5) The cyclical series were constructed using a Hodrick Prescott filter. Other filters were tested, including the Baxter-King band pass filter. The results do not change appreciably.
Phillip Chindamo, Chief Economist, Australian Industry Group, 20 Queens Road, Melbourne 3004, Australia. Email: phillip, firstname.lastname@example.org. The views expressed in this article are entirely these of the author, who thanks three anonymous referees for comments en an earlier draft.
Table 1: Gross flows March 2010 to April 2010(a) Labour force status April 2010 Persons, thousands, To. Not in To: To: original labour force Unemployed Employed Labour force From: Not in 4413 130 195 status in labour force March 2010 From: 118 293 118 Unemployed From Employed 258 82 8734 (a) Data source: Australian Bureau of Statistics (ABS), Labour Force, Australia, June 2010, cat. 6206.0. Table 2: Descriptive statistics for estimated job-finding and job loss rates for Australia (a) Sample: 1981Q4 Job-finding Job-loss to 2010Q1 rate rate Mean 0.251 0.049 Median 0.243 0.051 Maximum 0.358 0.058 Minimum 0.189 0.042 (a) Quarterly averages of monthly transition rates data.