The effect of public sector employment on local labour markets

Abstract

This paper considers the impact of public sector employment on local labour markets. Using English data at the Local Authority level for 2003–2007 we find that public sector employment has no identifiable effect on total private sector employment. However, public sector employment does affect the sectoral composition of the private sector. Specifically, each additional public sector job creates 0.5 jobs in the non-tradable sector (construction and services) while crowding out 0.4 jobs in the tradable sector (manufacturing). When using data for a longer time period (1999–2007) we find no multiplier effect for non-tradables, stronger crowding out for tradables and, consistent with this, crowding out for total private sector employment.

Introduction

This paper considers the impact of public sector employment on local labour markets. When a new job is created in an area, additional jobs may be generated as a result of increased demand for locally produced goods and services. This positive effect on employment may be offset by general equilibrium effects induced by changing local wages or prices (Moretti, 2010). In other words, the ‘multiplier effect’ of the additional jobs may be offset by ‘displacement’ or ‘crowding out’ elsewhere in the local economy. For public sector employment these effects may be complicated by the existence of a private–public pay differential and the fact that the increase in employment may be funded through additional local taxes.

In this paper we clarify some of the conceptual issues concerning the impact of public sector employment on local labour markets. However, our main focus is on obtaining empirical estimates of the effects using data on employment for English Local Authorities. We consider the impact of public sector employment growth on private sector employment growth as well as other labour market indicators (unemployment, participation and working age population). Our preferred specification using changes from 2003–2007 implies a short run overall multiplier that is insignificantly different from zero. Public sector employment has little effect on total private sector employment in the short run. In contrast, in line with predictions from a basic conceptual framework, we find evidence of a multiplier effect for non-tradable sectors and a displacement effect for tradable sectors. Our preferred specification implies that each additional public sector job creates 0.5 jobs in the non-tradable sector (construction and services) while crowding out 0.4 jobs in the tradable sector (manufacturing). For a longer time period (1999–2007) we find no multiplier effect for non-tradables, stronger crowding out for tradables and, consistent with this, crowding out for total private sector employment.

These effects are of considerable policy interest. The relocation of public sector employment is sometimes suggested as a tool for helping address employment problems in declining areas. Offsetting this, it is argued that public sector employment (and associated private-public pay differentials) may crowd out the private sector. In the UK, these issues have recently been considered in two government-sponsored reviews (Lyons, 2004, Smith, 2010). They have also provided the background to a series of relocation exercises since World War 2 (Jefferson and Trainor, 1996). Notwithstanding the attention given by successive UK governments to the subject, no robust evidence of the empirical effects are available. In the UK, continued interest in these issues partly reflects concerns over the uneven spatial impact of public sector job cuts (Larkin, 2009, Webber and Swinney, 2010) but also interest in the wider impacts of moving from national to local (or ‘market facing’) pay.¹

Surprisingly, to the best of our knowledge, this issue has been the subject of little (if any) systematic analysis. Since the late 1960s, a series of studies for both the UK and other (mostly European) countries have assessed the case for public sector relocation from capital cities to less-developed areas.² However, these studies focus on the financial costs and benefits of relocation and provide little empirical evidence on impacts on local labour markets. Indeed, descriptive evidence on the impact of past relocations is usually restricted to discussion of secondary data on overall public sector employment rather than actual data on government office relocations. This broadly descriptive literature incorporates academic papers (see, e.g., Marshall et al., 1991); government sponsored reviews (see, Lyons, 2004, Smith, 2010) and a small number of consultancy studies (see, e.g., Experian, 2004, Deloitte, 2004). BIS (2010) does consider the impact of public sector employment on private sector employment using a panel of data for English NUTS 3 regions from 2003 to 2008. We improve on those estimates by focusing on differences (i.e. changes over time), adopting an appropriate functional form and instrumenting to deal with the problems of endogeneity and reverse causality. As we show below dealing appropriately with the latter problem is crucial for understanding the impact of higher public sector employment.

Ex-ante predictions of the impact of public sector employment can be constructed using methods developed in the extensive literature on regional input–output models. These models use input–output tables to trace through the way in which local supply is likely to respond to an increase in economic activity. Such models usually provide a range of different multipliers but because they assume prices are fixed ignore any general equilibrium constraints that might lead to crowding out. Miller and Blair (2009) provide a classic textbook treatment. The US Bureau of Economic Analysis RIMS II is one of the best known and most widely used applications.³ Regional computable general equilibrium (CGE) models impose more theoretical structure to try to address the problems arising from the fixed-price assumption inherent in input–output models. Such regional CGE models are fitted to data that has been adjusted so as to be consistent with the underlying theoretical model (a process known as calibration). See Partridge and Rickman (2010) for a recent survey. Input–output and CGE models have been widely used to predict the impact of local demand shocks but neither approach provides estimates of the actual impact of such changes.⁴

In the macro-economic literature, a limited number of studies have looked at the potential impact of public sector employment on labour market outcomes (e.g., unemployment and private employment). Using data for 22 OECD countries from the end of the 1960s to 1990, Edin and Holmlund (1997) find that a rise in public sector employment reduces unemployment by about 0.3% in the short-run, whereas there is no significant long-run effect.⁵ When looking at the Swedish experience (with longer time series data), they find that the rapid growth in public sector employment in Sweden over the 1960s and the 1970s contributed to the low Swedish unemployment rate during those years. Their estimates indicate that the effect was at most one percentage point of unemployment during the booming years and much smaller after that.

Boeri et al. (2000) estimate that 10 additional public jobs crowd out 3 private jobs using a sample of 19 industrialised OECD countries over the period 1985–1992. Algan et al. (2002) focus instead on the long-run effects of public sector employment on both unemployment and private sector employment using a panel of 17 OECD countries between 1960 and 2000. They find that a rise in a country’s public sector employment increases that country’s unemployment. Furthermore, they find that public sector employment had, on average, a strong crowding out effect on private employment: creation of 10 public sector jobs tend to destroy 15 private sector jobs.⁶ Although interesting, these estimates do not take into account, or do not solve in a satisfactory manner, possible problems arising from reverse causality or endogeneity.

Moretti (2010) does attempt to isolate the causal impact of local employment changes and is the paper most closely related to our own work. Using US Census of Population data for 1980, 1990 and 2000, he looks at the long-term change in the number of jobs in a city’s tradable and non-tradable sectors caused by a permanent shock in the tradable sector. Results suggest a positive local multiplier of tradables on non-tradables (of about 1.6), but no impact of employment changes in one part of the tradable sector on the rest of the tradable sector. In contrast to our focus here, Moretti’s definition of the non-tradable sector specifically excludes government jobs (along with those in agriculture, mining and the military). Thus, his paper is only concerned with multiplier effects between tradable and non-tradable components of the private sector.

Our work is also closely related to the migration literature, a popular strand of which considers the possible displacement effect of immigrants on natives using cross-regional (city or census track) data.⁷ As we are also interested in the possibility of displacement (or a crowding-out effect) – but of public sector employment on private sector activity – it is possible to draw a parallel between the two approaches: one linking immigrants and natives, the other linking public and private sector employment. Indeed, our methodology uses an adapted version of that commonly found in the migration literature (specifically we use a version of Card’s (2007) model adjusted to take into account improvements as suggested by Peri and Sparber (2011)).

Other related research documents the large share of public sector workers in the economy (ranging from 17% of total employment in the US to about 22% in Western Europe); the sorting and substantial movement of workers between the private and public sector (see Borjas, 2002); the existence of a public–private wage differential and its evolution over time (see, among others, Nickell and Quintini, 2002, Disney and Gosling, 2008 for the UK); and the impact of public investment on local employment and wages (Pissarides and Wasmer, 1999).

Theoretical work considering the interaction between public and private sectors within a local labour market is scarce. To the best of our knowledge, Burdett (2012) is the only study that presents an equilibrium search model of the labour market where a public sector is explicitly modelled.⁸ Under a reasonable set of assumptions, Burdett (2012) shows that public sector employment crowds out private sector employment in regions where public pay is higher than the outside private option. In contrast, in regions where private employers offer a wage at least as large as the public sector, an increase in public sector workers would raise total employment leading to a multiplier effect. Available evidence for the UK (Emmerson et al., 2012) suggests that most regions of the UK pay a substantial public sector premium so we should expect displacement to dominate. Explicitly considering heterogeneity in response as a function of the pay differential would require Local Authority level estimates of the public-private pay differential which are not readily available and so this question is left for further work.⁹

The rest of the paper is structured as follows. The next section sets out a simple conceptual framework. Section 3 describes the data while Section 4 reports results on the impact of public sector employment on total private sector employment and other labour market indicators. Section 5 reports results once we split private sector employment by broad industrial sector, while Section 6 reports results when we consider changes over a longer time period. Section 7 discusses the role of the elasticity of labour supply in explaining our results. Section 8 concludes.