How local are labour markets? A look at the London Olympics

How local are labour markets? A number of important questions in labour economics hinge on the answer.

In recent years there has been a resurgence of interest in the consequences of localisation of economic activity for workers' welfare (see Moretti 2011, for a recent overview) and in policies aimed to improve labour market outcomes in disadvantaged areas (Glaeser and Gottlieb 2008).

In the US, federal, state and local governments combined spend nearly $50 billion per year on local development policies. Notable examples are the Moving to Opportunities programme and the Empowerment Zones programme. These policies need to know about the size of local labour markets to decide about the appropriate nature of the intervention.

• If labour markets are very local then an effective intervention will have to be targeted to the disadvantaged areas themselves even if those areas are not conducive to generating employment.
• If labour markets are not as local, then there is less need for the targeted intervention and a targeted intervention may simply attract workers from other more advantaged areas.

An important, related question concerns the incidence of local shocks to labour demand and their impact on labour mobility, local unemployment, and earnings. To answer such questions one needs a clear definition of a 'place'.

Existing evidence

Most existing research on the topic and government statistical agencies divides geographical space into non-overlapping areas, which are then assumed to be single labour markets. Examples would be the 367 Metropolitan Statistical Areas for the US, or the 320 Travel to Work Areas (TTWAs) for the UK.

While these efforts are understandable and useful, they do have their problems. For example, UK TTWAs are constructed so that at least 75% of the population resident in a TTWA actually work in the area, and at least 75% of those who work in the area also live in the area. Because people commute from large distances to central London to work, the whole of the Greater London area is classified as a single labour market. But those who live in the northern suburbs of London do not really think of the far southern suburbs as part of their local labour market.

The non-overlapping nature of local labour markets constructed in this way also causes inevitable discontinuities around the boundaries. Someone living just inside the London TTWA will be classified as living in an enormous labour market while someone living just across the border in the Luton TTWA will be classified as being in a modestly-sized local labour market. But these people are in essentially the same labour market.

The fundamental cause of these problems is a failure to recognise the continuous nature of geographic space; the labour market for one individual at one location will overlap with that for another individual in a different but not too distant location. Consequently, there is no way to segment an economy into non-overlapping areas without mismeasuring local labour markets. But, how can one model geographical space in a more realistic way while preserving tractability?

When geography is treated as continuous

If geographic space is treated as continuous, as opposed to a collection of non-overlapping administrative units, this avoid problems of mismeasurement of local labour markets described above, and turns out to have crucial implications for the impact of local policies. In recent work (Manning and Petrongolo 2011) we model the size of local labour markets by the cost of distance, whether measured as geographic distance, commuting time, or commuting cost. If the cost of distance is high then workers will be more reluctant to consider jobs at more distant locations than if the cost of distance is small. This approach means that the boundary of a local labour market is fuzzy, but that is the right way to think about a jobseeker's decision problem.

Let us briefly consider how one might approach the question of estimating how local are labour markets from individual behaviour. Commuting data might be expected to contain useful information about the size of local labour markets as they tell us about how far workers seem prepared to travel to jobs. But commuting patterns represent the outcome of a bunch of decisions (e.g. residential location) that muddy the waters. To give a specific example, consider the academic job market. From commuting patterns one would observe that most faculty live reasonably close to their place of work and perhaps conclude that the labour market for academics was relatively local. But, of course, it is better described, albeit with some hyperbole, as global. What information would allow us to detect that? We argue that one could detect that by looking at the address of the job market candidate when they applied for a job and looking at the distances they are considering. In the academic market a job candidate in a specific current residential location is prepared to take a job over a very large geographical area but will then change residential location to be close to whatever job they obtain. In this situation it makes sense to think of the individual being in a very large local labour market as they will consider a very wide range of jobs. But they will end up with a low commute, so commuting data would not reveal the true extent of the local labour market.

In our approach, by a 'local labour market' we mean the set of jobs that an unemployed worker, currently in a particular location, will apply for. It may be that, if the application is successful, the individual chooses to change residential location but that would not concern us. We use data on unemployment and vacancies in small neighbourhoods in England and Wales (8,850 in total), providing a much closer approximation to the continuous nature of geographic space than existing studies.

We argue that data on the filling of vacancies can be used to infer the distance over which people look for work. The intuition is the following. Consider a vacancy in area A. It is plausible to think that the ease of filling this vacancy depends on the number of unemployed workers for whom the vacancy is in their local labour market. If the ease of filling a vacancy in A is influenced by the number of unemployed people in area B but not in area C, then a reasonable conclusion is that area A is in the local labour market for people who live in area B but not for those who live in area C.

In our framework, unemployed workers decide whether to apply to jobs at different locations based on the cost of distance to target jobs and on the probability that an application is successful, in turn depending on how many other jobseekers across the economy would find these jobs attractive. Interdependencies across areas arise because the number of applicants to jobs in a given area is likely to be influenced (even if only very slightly) by unemployment and vacancies in all other areas, insofar as they are ultimately linked through a series of overlapping markets. The resulting number of job applications in each area is the central ingredient of the process by which local vacancies are filled.

Our estimates suggest that the cost of distance is relatively high. Specifically, the attractiveness of a job falls by about 4.5 times if one pulls the job five kilometres further afield from a jobseeker's location. Also, other things being equal, workers tend to apply to jobs in areas where they expect relatively less job competition. These results imply that, despite the fact that labour markets are relatively 'local', location-based policies in the form of local stimulus to labour demand or improved transport links are rather ineffective in raising the local job finding rate, because labour markets overlap and the associated ripple effects in job applications largely dilute the effect of local policies across space.

As an example, we consider an increase in the number of job openings in Stratford and New Town ward in East London, which is the main venue of the 2012 Olympic Games, and has an unemployment to population ratio nearly three times higher than the average ratio for England and Wales. This example thus combines very large increases in numbers of vacancies as a result of Olympic-related projects with a relatively depressed local labour market. Our framework predicts that, after a doubling of job vacancies in Stratford, there is only a tiny (0.4%) increase in the local job finding rate. This happens because unemployed workers living relatively close to Stratford divert some of their job search effort from their local wards towards Stratford. This raises job competition in Stratford, reduces job competition in their local wards, attracts applications from elsewhere, and so on. This mechanism explains the spatial propagation of local shocks in the presence of relatively high costs of distance.

How does this prediction about the impact of a local labour demand shock compare with what has actually happened in Stratford in the run-up to the 2012 Olympic Games? Much of the increase in labour demand is taking place right now running the Olympics itself, while some has built up steadily over time (such as in construction). But there is one instance of a sharp increase in labour demand that has taken place: on 13 September 2011, a new shopping centre, Westfield Stratford City, opened next to the London Olympic Park. This is one of the largest urban shopping centres in Europe, and has been expected to contribute significantly to local development. We can then look in the recent data for any evidence of the corresponding labour demand shock on vacancy creation and the job finding rate around Stratford. While vacancy data record a sharp rise in Stratford in the summer of 2011 that is associated with the opening of the shopping centre, we can find little or no evidence of an increase in job finding in Stratford or surrounding areas as a result of the spike in vacancies. These early indications are exactly in line with the predictions of our model and are certainly consistent with negligible local effects of targeted labour demand stimulus. The bottom line is that even strong local stimulus has a limited bite on the local outflow rate from unemployment, because a series of spatial spillovers would greatly dilute any local shock across space.

References

Glaeser, Edward L and Joshua D Gottlieb (2008), “The Economics of Place-Making Policies”, Brookings Papers on Economic Activity, Spring, 155-253.

Manning, Alan and Barbara Petrongolo (2011), “How Local are Labor Markets? Evidence from a Spatial Job Search Model”, Centre for Economic Performance DP No.1101.

Moretti, Enrico (2011), “Local Labor Markets”, in O Ashenfelter and D Card (eds.), Handbook of Labor Economics, 4B Amsterdam: Elsevier-NorthHolland, 1237-1314.