The local labour market effects of earthquakes

ABSTRACT

Using a balanced panel of local labour market areas (LLMAs) and adopting a new difference-in-differences approach with multiple time periods and multiple groups, this paper assesses the causal effects on local labour market outcomes of the earthquakes that occurred in Italy in 2009, 2012 and 2016. The results show a strong heterogeneity: the 2009 event had significant negative and persistent impacts on the employment rate of the LLMAs involved, while there were no adverse effects for the LLMAs affected by the earthquakes in 2012 and 2016. We also extend the analysis at the industry level in order to highlight the sectoral shifts that typically characterise the aftermath of these natural disasters.

KEYWORDS:

• natural disasters
• labour market
• counterfactual analysis
• difference-in-differences approach

JEL:

• C32
• J61
• Q54

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. To compute this indicator, we consider only the municipalities with significant damage and, therefore, with a Mercalli value above grade VI.

2. A thorough description of the socio-demographic and economic characteristics of the LLMAs in Italy is reported in Appendix B in the supplemental data online.

3. In some cases, such as floods or drought, natural disasters are not completely exogenous events. However, the exogeneity of the treatment can be easily assumed in the case of earthquakes, as also discussed by Porcelli and Trezzi (2019) and Mendoza et al. (2020).

4. All estimates are performed using the doubly robust approach developed by Sant’Anna and Zhao (2020) and available in the R package DRDID.

5. Five categories are considered based on the level of labour cost (€, thousands): (1) up to 24.2 ($LC1$); (2) between 24.2 and 31.2 ($LC2$); (3) between 31.2 and 34.6 ($LC3$); (4) between 34.6 and 38.1 ($LC4$); and (5) more than 38.1 ($LC5$).

6. Four categories are identified based on the quartile distribution of the ratio between imports plus exports (€, thousands) and the number of employees: (1) up to 2.3 ($Q1$), (2) between 2.4 and 10.3 ($Q2$), (3) between 10.4 and 31.6 ($Q3$), and between 31.7 and 40.1 ($Q4$).

7. Spatial lags are computed using a row-standardised symmetric binary weights matrix based on the cut-off distance of 41 km, which is the minimum distance ensuring that each location has at least one neighbour. In other words, each element of the matrix takes the value of 1 if the great circle distance between LLMA $i$ and j is less than the minimum cut-off distance of 41 km, 0 otherwise.

8. Callaway and Sant’Anna (2021) suggest choosing the ‘never treated’ option with respect to the ‘not yet treated’ one when there is a sizeable group of units that do not participate in the treatment in any period. We have also checked the robustness of the results compared with the alternative choice. The results do not change at all.

9. We used a restricted official definition of the LLMAs involved in these events. In the case of the 2009 earthquake, this definition excludes coastal areas, which were included in another enlarged official definition of the area hit by the L’Aquila quake. Our decision was based on the observation that municipalities and LLMAs in coastal Abruzzo were actually much less directly involved in the event and worked as reception areas for people displaced in the immediate aftermath of the quake.

10. The 2016 treatment group does not include the three LLMAs located in Abruzzo (L’Aquila, Penne and Teramo) which were hit by the 2016 earthquake, as they had been previously hit by the 2009 earthquake and have been included in the 2009 group.

11. As discussed above, our results are based on the conditional parallel trends assumption. In other words, we assume that, in the absence of treatment, only LLMAs with the same characteristics would follow the same trend in the outcome variable. This conditioning scheme is extremely important to us because the economic structure of LLMAs hit by the earthquake is different from those that are not, and because the paths of employment rates (absent the earthquake) are affected by the economic structure of each LLMA.

Additional information

Funding

This study was supported by Territori Aperti, a project of the University of L'Aquila financed by the Fondo Territori Lavoro e Conoscenza, established by the Confederazione Generale Italiana del Lavoro (CGIL), the Confederazione Italiana Sindacati Lavoratori (CISL) and the Unione Italiana del Lavoro (UIL).