Economics Letters 186 (2020) 108787
Contents lists available at ScienceDirect
Economics Letters journal homepage: www.elsevier.com/locate/ecolet
Redistributive public employment? A test for the South of Italy Marta Auricchio a ,1 , Emanuele Ciani a ,1 , Alberto Dalmazzo b ,1 , Guido de Blasio c , a b c
Bank of Italy, Italy Department of Economics, University of Siena, Italy Bank of Italy, via Nazionale 91, 00184 Rome, Italy
Article history: Received 5 August 2019 Received in revised form 16 October 2019 Accepted 20 October 2019 Available online 23 October 2019
a b s t r a c t This paper proposes an econometric test to verify whether public employment discourages the development of private economic activity in the South of Italy. Our results suggest that this is indeed the case. © 2019 Elsevier B.V. All rights reserved.
JEL classification: H53 J21 Keywords: Public employment Private employment Italy
1. Introduction In their deeply provocative 2001 paper, Alesina et al. (ADR hereafter) suggest that public employment in the South of Italy is, for about a half of the wage bill, a redistributive device. In short, it is a subsidy from the North to the lagging South that eventually discourages the development of the local private sector.2 Since the public sector pays higher wages and provides secure jobs, the competitiveness of local private firms is hampered. This paper provides a rigorous empirical verification of the ADR claim. To this purpose, we study whether the impact of exogenous contractions in public employment – which free up resources for the local private sector – is stronger in Southern Italy. We elaborate on Auricchio et al. (2019), where we study the impact on Italian municipalities of the downsizing of public sector employment that took place between 2001 and 2011. Exogeneity comes from the fact that, over that period, the Italian central government cut down local public employment by halting turnover: this measure was essentially taken for nationwide budgetary reasons, regardless of the economic conditions of the municipalities. ∗ Corresponding author. E-mail addresses: [email protected]
(M. Auricchio), [email protected]
(E. Ciani), [email protected]
(A. Dalmazzo), [email protected]
(G.d. Blasio). 1 The views expressed in this paper are those of the authors and do not necessarily correspond to those of the institutions they are affiliated. 2 The South of Italy refers to the regions of Abruzzo, Molise, Campania, Puglia, Basilicata, Calabria, Sicilia, and Sardegna. https://doi.org/10.1016/j.econlet.2019.108787 0165-1765/© 2019 Elsevier B.V. All rights reserved.
We focus on the consequences of the retrenchment on private employment both in the tradable and nontradable sectors and, also, on additional outcomes, such as house prices and population. Spatial general equilibrium theory (Roback, 1982) suggests that, if the ADR claim is correct, an exogenous decrease of public employment in an area will exert a downward pressure on local salaries, and will also decrease the demand for locally produced goods, such as housing. Diminishing wages and house prices will spur the growth of the local private sector. To the extent that private employment fully counterbalances the loss of public jobs, no net population movements are expected. The rest of the paper is structured as follows. Section 2 describes the data. Section 3 highlights the identification strategy. Section 4 provides the results on the ADR conjecture. Section 5 concludes. 2. Data We exploit municipality-level data on private and public employment from the 2001 and 2011 Italian Industry and Service Censuses. Private sector is defined by economic activities that contribute to gross domestic product at market prices, in the industry, commerce and services sectors. The public sector refers to the provision of non-market goods and assets for the benefit of the community. In both sectors, employment includes workers with fixed-term or permanent employment contracts, and the self-employed (but not contractors and temporary workers, apart from those on standard fixed-term contracts, because they are available only for the public sector). Data are disaggregated at the
M. Auricchio, E. Ciani, A. Dalmazzo et al. / Economics Letters 186 (2020) 108787
industry level, with the ATECO 3-digit classification, corresponding to the NACE classification adopted by Eurostat. The issues about ATECO reclassification across the two Census waves have been solved by an algorithm, which is available upon request. Public employment is spread across 8 sectors. The main ones are education (one third of total public employment in 2011), health (one fourth), and administration of the state and the economic and social policy of the community (one fifth). Census data do not provide information on the skill level of the workforce. In order to include a set of additional variables and controls, we exploited the information on population and housing Censuses available on 8milacensus (http://ottomilacensus.istat.it/). Data on house prices per square meter are, instead, obtained by using the Bank of Italy Index built on the OMI house prices database. However, since the OMI’s dataset is available from 2003, we use prices in 2003 as a proxy for the prices in 2001. Census data do not provide any information on local wages. There is no available data source on wages at the municipality level between 2001 and 2011. Finally, we select only the municipalities that exist in both the census waves considered. 3. Identification
∑ Njpubl ,c ,2001 j∈pub
Nc ,2011 − Nc ,2001 Nc ,2001
= β0 + βpub
Nc ,2011 − Nc ,2001 Nc ,2001
+ xc ,2001 βx + ϵNpriv ,c (2)
We condition on a large set of covariates (see the Notes to Table 1) to make sure that: (a) spillovers due to commuting do not invalidate the SUTVA and, (b) pre-trends are adequately controlled for. This empirical strategy is robust to an extensive sensitivity analysis (results available upon request). A reasonable degree of failure of the exclusion restriction does not invalidate our estimates (Oster, 2019). Moreover, our findings nicely survive Machine Learning selection of covariates (Belloni et al., 2014). 4. Results
Following Faggio and Overman (2014), we adopt an IV strategy built on a specification in contributions to growth. Our instrument follows the logic of Bartik (1991), applied to public employment. For each municipality c, we first compute national growth of public employment in each sector j between 2001 and 2011, omitting the municipality itself from the calculation (as indicated by the subscript -c).3 We then multiply it by the ratio between public employment in sector j and overall employment (private plus public in all sectors) in the municipality c at the beginning of the decade. Finally, we aggregate across sectors: instc ,2011 =
allows us to deal with both omitted variable and endogeneity issues (measurement is not a concern as we work with Census data). publ Using instc ,2011 as an instrument for the actual public sector contribution to employment growth between 2001 and 2011 we estimate the following equation:
Nj,−c ,2011 − Nj,−c ,2001 publ
The instrument is the predicted contribution of public employment to overall local employment growth, calculated by using national trends, which are strongly influenced by the central government’s decision to downsize public employment. The exclusion restriction is likely to be satisfied for the Italian case, where local public employment is largely financed by transfers from the central government, rather than local taxation. Consequently, the allocation of public employees over the national territory is mostly decided at the central level. This circumstance weakens the link between the wealth of the local private sector and the size of the public one. No need to say, lagging areas might still get a higher share of public workers, when the government attempts to relieve local unemployment. In the decade we consider, however, limitations imposed both by the EU and national laws in 2002, 2004, 2006, set a total or partial stop to new hirings.4 Such stops in turnover implied proportional cuts in employment, since the fraction of public employees entering retirement was not replaced by new hires.5 Our IV approach 3 As standard in the literature, this leave-one-out procedure is aimed at reducing the concerns of simultaneity. 4 Between 2001 and 2011 the total number of employees in the public sector decreased by 11 percent (in the previous census period, 1991–2001, the number increased by 3 percent). 5 Clearly, this fraction is not necessarily the same in all municipalities, as it depends on the age structure of public employment. However, the purpose of the instrument is to exclude variations in public employment that may be systematically related to private employment growth. From this perspective, the age structure of the public employment observed in the decade 2001–2011 depends on the hiring decisions made from the 1960s to the 1980s when public sector employment boomed.
Table 1 reports the second-stage estimates for the 2557 municipalities of the South of Italy (the first-stage F statistics always display values that reassure us of weak-instrument concerns). We find that the ADR’s claim receives empirical support. Column (1) suggests that a reduction of one public employee will crowd in 1.06 private employees. Columns (2) and (3) suggest that crowding in involves both tradables and nontradables. The impact on the latter is somehow unexpected, as previous literature often finds that variations in public employment go hand in hand with variations in local private employment in nontradables. In our case, the more favorable supply-side conditions caused by the reduction in public employment seem to outweigh the adverse impact on the local demand for services and constructions. We also find that public sector downsizing implies a substantial reduction in local rents (Column 4). Finally, the fact that we fail to find significant population movements (Column 5) suggests that, not only public employment losses are fully offset by private employment gains but, also, that real wages remain constant. In other words, the downward pressure on local nominal wages has been compensated by lower rents. Table 2 reports similar estimates for the 5528 Northern municipalities, which are used for comparisons with the findings about the Southern ones. A crowding-in effect still seems to be there – suggesting that a positive effect from public sector contraction also materializes in Northern areas – but the magnitude (Column 1) is less than a half of the estimate for the South, and entirely concentrated on tradables (Column 2). We also find a minor impact on house prices (not statistically significant, Column 4) and a pronounced effect on population outflows (Column 5). Given that the crowding in of private jobs does not fully compensate for the loss of public employment, this finding suggests that those who lost a public job have moved elsewhere. Some of the crowding out effects could be mechanical.6 On the one hand, when job opportunities in the public sector shrink, residents may decide to start their own business, and work as self-employed. In this case, the average local firm size should decrease. On the contrary, if private business competes with the public sector in the local labor market, a reduction in public employment will make it simpler for private firms to expand their workforce. In this case we should find that the average local firm 6 We thank a Referee for this suggestion.
M. Auricchio, E. Ciani, A. Dalmazzo et al. / Economics Letters 186 (2020) 108787
Table 1 The impact of public sector contractions in the South. Private employment (1)
House prices (4)
Average plant size (6)
Notes: The table reports the second-stage estimates (Eq. (2)) for the coefficient of public employment in the South. The sample includes 2557 municipalities. The instrument is defined in Eq. (1) and the first stage in Eq. (2). * p-val < 0.1, ** p-val < 0.05, *** p-val < 0.01. The standard errors, in brackets, are clustered at the LLM level (2001 definition). First-stage F is equal to 271. We censor the contribution to growth at the 5th and 95th percentiles, while the instrument is censored at the 1st and 99th. The unit of observation is the municipality across 2001–2011. We keep only municipalities that exist in both years. Public employment, Private employment, Tradables, Nontradables and Population are expressed as contributions to overall (public+private) employment growth. House Prices and average plant size are expressed as growth rates. Tradables include the manufacture sector and the extraction of natural resources. Nontradables include the service sector and construction. Population refers only to working age population (aged 15–64). We condition for a large set of covariates (see Auricchio et al., 2019) to make sure that (a) spillovers due to commuting do not invalidate the SUTVA and (b) pre-trends are adequately controlled for (since our specification is basically a diff-in-diffs with continuous treatment). Estimates are produced using the command ivreg2 by Baum et al. (2010). Table 2 The impact of public sector contractions in the North. Private employment (1)
House prices (4)
Average plant size (6)
Notes: See notes of Table 1. The sample for the North includes 5528 municipalities and the first-stage F is equal to 251.
size increases when the public sector stops hiring. We estimate Eq. (2) using growth in average plant size as dependent variable. The results (Column 6) suggest that the second argument – focusing on local labor market tightness – is more likely to prevail, as public sector retrenchments bring about an increase in private plant size, in particular in the South. 5. Conclusions ADR claim that public employment in the South of Italy has largely been a redistributive device which hindered private sector development. Our estimates suggest that this claim appears to be justified. In the South, public employment contractions have produced strong beneficial effects on market activities: each public job that disappeared has been replaced by a private one. Differently, the crowding-in effect in Northern municipalities has been less than a half.
References Alesina, Alberto, Danninger, Stephan, Rostagno, Massimo, 2001. Redistribution through public employment: The case of Italy. IMF Staff Pap. 48 (3), 447–473. Auricchio, Marta, Ciani Emanuele, Dalmazzo Alberto, de Blasio, Guido, 2019. Life after public employment retrenchment: evidence from Italian municipalities. J. Econ. Geogr. http://dx.doi.org/10.1093/jeg/lbz013. Bartik, Timothy J., 1991. In: Upjohn, W.E. (Ed.), Who Benefits from State and Local Economic Development Policies. Institute for Employment Research, Kalamazoo, Mich. Baum, Christopher, Mark, Schaffer, Stillman, Steven, 2010. Ivreg2: Stata module for extended instrumental variables/2SLS, GMM and AC/HAC, LIML and k-class regression. http://ideas.repec.org/c/boc/bocode/s425401.html. Belloni, Alexandre, Victor, Chernozhukov, Hansen, Christian, 2014. Highdimensional methods and inference on structural and treatment effects. J. Econ. Perspect. 28 (2), 29–50. Faggio, Giulia, Overman, Henry, 2014. The effect of public sector employment on local labour markets. J. Urban Econ. 79 (C), 91–107. Oster, Emily, 2019. Unobservable selection and coefficient stability: theory and evidence. J. Bus. Econom. Statist. 37, 187–204. Roback, Jennifer, 1982. Wages, rents and the quality of life. J. Polit. Econ. 90 (6), 1257–1258.