Since the global financial crisis, a number of central banks have been undershooting their inflation targets. Over this period labour markets have tightened steadily in many countries with unemployment rates reaching record lows. At the same time, wage growth has remained weak by historical standards. Some researchers argued that the Phillips curve, the main paradigm that central banks use in analysing wage pressures, is no longer useful. Some, on the other hand, think it is premature to call for the death of the Phillips curve. It is an important question, however: what is happening to wages? How can we understand them? The recent and influential macro-labour literature with search and matching models emphasises the importance of gross labour flows in analysing labour market developments and their macro implications.
The unemployment rate, the key input in the measurement of the slack in the labour market, is the net change in unemployment. But there is a lot more going on behind the scene. Take the net quarterly change in unemployment for example: the net change in unemployment is made up by flows of workers into unemployment from employment and from outside the labour force; and out of unemployment into employment or to out of the labour force.
The net figure is often a tiny fraction of these gross labour flows. Figure 1 shows these gross flows for the December 2017 quarter for New Zealand for illustrative purposes. By using these flows we can show that the net fall in unemployment in that particular quarter is 2,300 people. While the gross flows from unemployment to employment (EU), and to not in the labour force (EN), are 16,400 and 35,300 respectively, the flows from employment to unemployment (UE), and to not in the labour force (UN), are 30,500 and 23,500. It is obvious that net flows hide the much larger churn and the associated richness in the dynamics.
The net flow in unemployment also misses another important aspect of gross labour flows that is job-to-job flows: employees who switch to another job without having a spell of unemployment.
Search and matching models are a widely used framework for analysing labour market dynamics including wages. In the canonical search and matching model of the labour market, wages are set through decentralised bilateral bargaining between the employer and the employee (Diamond, Mortensen and Pissarides – DMP). The bargaining power of a worker is the key variable, and is determined by the attractiveness of their outside option, namely walking away from the current match and joining the pool of unemployed workers to look for another job. Hence, the pace at which unemployed people find jobs is a crucial factor determining cyclical wage fluctuations. In the case of Nash bargaining specifically, the equilibrium wage turns out to be a weighted average of a worker’s productivity and her reservation wage, where the latter is directly influenced by the job-finding rate (UE, the share of unemployed workers who transition to employment in a given period). According to the DMP model, when the job-finding rate is high, workers have more bargaining power as they can leave the negotiating table and get a good offer elsewhere, so that, all else equal, wages increase.
A key assumption of the DMP model, however, is that unemployed job seekers are the only available source of labour to fill new vacancies posted by firms. This implies that an employed person has to first become unemployed before seeking another job. This simplifying assumption ignores employed people who search for jobs, a phenomenon termed on-the-job search. Strikingly, this group is much larger in terms of numbers than the number of unemployed. In the United States, roughly 40 per cent of all vacancies are filled by people who were employed in another job (Fallick and Fleischman 2004). Moreover, only 16 per cent of people who quit their jobs are next classified as unemployed, while the rest move to other jobs or leave the labour force (Mukoyama et al. 2018). This flow, known as job-to-job transitions, appears to be important.
In an earlier paper, Karagedikli (2018), I showed that in New Zealand around 9 per cent of employed people change their jobs between two quarters. The total number of job-to-job transitions is 5-6 times greater than the number of people who move from unemployment to employment in a quarter. Moscarini and Postel-Vinay (2016, MPV) observe that a corollary of the Burdett and Mortensen (1998, BM) model of wage dispersion is that the job-to-job transition rate is the primary driver of real wages. Competition between firms for workers who are already employed (which is observed through the pace of job-to-job transitions) drives wages higher. This prediction differs strikingly from the DMP model in which the job-finding rate is the main determinant of wage fluctuations.
Motivated by this insight, MPV analyse times series data and find that the evolution of wages over the business cycle is closely linked to the pace of job-to-job transitions. Their empirical finding challenges the theoretical prediction of the DMP model that the job-finding rate should be a key determinant of wage dynamics. More specifically, in their model the primary driver of real wages is the competition among firms for employed workers. Data limitations have partly played a role in these two competing views of wage determination from being tested. The DMP model was easy to test given the availability of long time series for the job-finding probability.
Only recently, however, have reliable job-to-job transitions data started to emerge, mostly coming from administrative data. In a recent paper, Karahan et al. (2017) empirically test the relative relationship of these two views of wage setting. In contrast to the earlier attempt by MPV which used a time series of job-to-job transitions, Karahan et al. (2017) use panel data to exploit state-level variation in job-to-job transitions and job-finding rates, and measure their relative influences on wages. They find strong empirical support in favour of on-the-job search and job-to-job transitions being the prevailing sources of cyclical wage dynamics in the United States.
In a more recent paper (Ball et al. 2020), my co-authors and I use administrative data from the linked employer-employee dataset (LEED) in New Zealand regions to understand the roles of different labour flows in explaining earnings growth. We exploit the pooled cross-regional variation to understand the comparative relationship of the job-finding and job-to-job transition rates for wage fluctuations in New Zealand.
We find that wages react strongly and significantly to the pace of job-to-job transitions and only slightly and less significantly to the job finding rate. We also evaluate the respective explanatory power of the job-to-job transition rate and job-finding rate for wage dynamics at different quantiles of the wage distribution. Our main finding is that job-to-job flows dominate the job finding probability in explaining wage movements. This effect appears particularly noticeable at the lower deciles of the wage distribution. By calling for greater attention to be paid to the lower part of the wage distribution, our result resonates with some earlier findings that the wage Phillips curve only holds at low wage rates. Rather than focusing on the relationship between the unemployment rate and wages (the Phillips curve), or the relationship between the job finding rate and wages (the DMP model), however, our findings emphasise the relationship between job-to-job flows and wages. In addition, our data also allows us to test the effects of job-to-job transitions on the wage rates of employees who do not switch jobs, as well as the effects of separation probability.
What do these results imply for SEACEN economies? Most SEACEN economies have large informal sectors and some member countries have a large fraction of foreign workers. Statistical agencies could include these in their labour force surveys with the aim of having a richer and deeper understanding of labour market dynamics. In addition, these surveys could also include questions about gross labour flows, including job-to-job transitions, similar to the Survey of Consumer Expectations (SCE) by the New York Fed.