1 Introduction

Humans have lived in contact with potential harmful natural phenomena since the beginning of history (Pelling 2003), but events such as earthquakes rarely release their force over human settlements so strongly, that end up reshaping their social order in what we call a disaster. We can firmly assess that “a natural disaster is not a natural event” (Hallegatte 2014, p. 9), because nature forces can be followed and even the strongest event can cause very few casualties. Moreover, in many cases that qualify as a disaster, the hardest time was the medium-term after the tragedy, when institutions were not always able to rearrange a new harmonic situation and many conflicts arose (Alexander 2000).

For these reasons, many scholars consider it more appropriate to consider the human perspective and define such disasters as “socio-natural” (Arteaga and Ugarte 2015). From an economic point of view, these events cause disruption to the functioning of an economic system, with significant negative impacts on well-being, assets, production factors, output, employment, and/or consumption (Hallegatte 2014).

When an earthquake or any other natural disaster occurs, there are direct and indirect costs a region should bear. Direct costs are the ones showing an effect on the physical environment, labor and housing markets in the affected areas, while indirect costs are the ones that slow productive systems and lower purchasing power, with potential negative spillover effects. One of the most relevant markets influenced both by indirect and direct costs is the labor market (Meyer et al. 2013).

We know that many countries are heterogeneous (economically, culturally and institutionally), so in order to have a broader vision on how local labor markets change after these events, we explore short-term changes across three countries in this study: Italy (European high income), Chile (South American high income) and Ecuador (South American upper middle income).

To obtain a broader perspective of the post-disaster labor market adjustment mechanisms, we develop a comparative analysis between these countries in order to understand and systematize the main differences and similarities of their labor market dynamics.

Disasters can generate ambiguous economic effects in the short and long run (Skidmore and Toya 2002; Rampa 2020), but specific economic effects will depend on the context and region in which the event occurs. Earthquakes often destroy local productive infrastructure in the affected areas and cause severe damage to dwellings, private and public infrastructure, e.g.: schools, hospitals, buildings, machinery, equipment, roads, electric transmission lines, among others (Committee for Reconstruction and Productive Reactivation 2016). Therefore, it is reasonable to think that labor markets dynamics within the seismic zones will be affected. Research in this area has documented increases in unemployment (Xiao and Feser 2014) and shifts towards informality (Mendoza and Jara 2020), how job seekers change their behavior (Ohtake et al. 2012) and mismatches between job offers and demand (Higuchi et al. 2012; Ohtake et al. 2012). Earthquakes could also cause structural changes in the medium or long run (Belasen and Polachek 2008; Mehregan et al. 2012). Some economic sectors linked to the reconstruction will experience a boom during the reconstruction period (Chang and Rose 2012); new industries could be born and replace inefficient infrastructure (Ohtake et al. 2012; Xiao and Nilawar 2013).

In order to understand how local labor markets change in the aftermath of an earthquake we use individual labor surveys for each country, focusing in three specific ways local labor markets adapt to new scenarios: (1) employment, (2) wage, and (3) worked hours. Our research objective is to find out whether individuals located within the most seismic regions in each country had a positive or negative likelihood of being employed, increase their wages or change their worked hours. Our empirical strategy consists in a descriptive analysis that compares post-disaster dynamics of labor markets in each country. Firstly, we present descriptive comparisons of labor income differences for affected and unaffected areas between pre and post disaster periods in each country. Second, we perform Mincerian regressions and double difference (DD) estimations, for labor income and worked hours in order to find out if there was an earthquake effect in regional labor markets. Third, we estimate the likelihood of being employed using logistic regressions using a nonlinear DD approach. To compare countries appropriately, we choose specific regions as areas affected by the corresponding earthquakes: (i) the provinces of Manabí and Esmeraldas for Ecuador, (ii) the region of Maule and Biobío for Chile, and (iii) the Abruzzo region for Italy.

We found significant associations between the earthquakes and labor market dynamics in the short term. Distributional differences showed that the earthquakes probably had heterogeneous effects on wages (depending on the labor income quintile). Mincerian wage regressions and DD estimates found higher labor income and worked hours for males and females in Ecuador. In Chile no effect was found, and in Italy we found a positive (significant) effect only for females, who increased their working hours. Finally, we did not find a short run association between the earthquakes and the probability of being employed for individuals located in seismic areas.

This article is divided in six sections, the next section introduces the relevant literature of regional labor markets after a disaster, Sect. 3 describes our empirical cross-country comparison methodology and dataset, Sect. 4 presents information of the three case studies and the dataset that we used in our analysis, Sect. 5 shows the results, and Sect. 6 discusses and concludes.

2 Regional labor markets after a socio-natural disaster

This section will review some of the relevant literature that has explored the relationship between labor markets and disasters in the past. To organize this revision, we first look at studies that explain cross-country differences, how scale and development issues have been considered to study labor market resilience, and finally how specific aspects of the workforce environment is reshaped after a catastrophe.

Although the literature has already empirically determined some effects of disasters on labor markets, some studies have focused on high income countries (Horwich 2000; Belasen and Polachek 2008; Di Pietro and Mora 2015; Karnani 2015; Dresdner and Sehnbruch 2010), while others have looked at developing and middle income countries (Mueller and Quisumbing 2009; Rodríguez-Oreggia et al. 2013; Gagnon 2013; Mendoza and Jara 2020).

A good starting point for comparisons between countries using cross country data is the article by Skidmore and Toya (2002). They show how ambiguous the economic effect of disasters could be, finding that geological disasters are negatively correlated with growth, but climatic disasters are positively correlated with increases in total factor productivity, economic growth and human capital accumulation. Creative destruction (Schumpeter 1942) might be a possible explanation, as pointed out by some authors (Ohtake et al. 2012; Xiao and Nilawar 2013), since it causes the birth of new industries by freeing up resources, replacing less efficient infrastructure, and increasing productivity due to technological improvements. Crespo Cuaresma et al. (2008) point out that this relationship might not hold in developing countries, as it is difficult for them to introduce and disseminate new technologies, weakening the role of creative destruction. This ambiguity in the effects motivates us to contribute to close this gap in current research.

Another important insight from this literature is the idea of economic resilience of scale (Xiao and Nilawar 2013), which states that—in aggregate—an economy can be resilient since the destructive effect of a disaster disappears as you move away from its core. When we analyze smaller spatial units, the negative effects of a disaster are perceived less, and the opposite happens when the affected area is larger. This is another important motivation for this article, as our case studies include large, medium, and small affected geographic areas. In the article by Ohtake et al. (2012) it is mentioned that some studies have determined that a large-scale disaster will produce a negative economic impact in the short run (Loayza et al. 2009; Raddatz 2007; Noy 2009); a smaller scale natural disaster would have a positive economic impact given that the losses are less than the effects of reconstruction (Loayza et al. 2009).

Many relevant findings in the broad disaster literature are focused in high income countries. Ewing et al. (2005) evaluate changes in the labor market as a result of Hurricane Bret (1999) and the following recovery, finding that the unemployment rate in the long run is reduced. In developing countries other perspectives must be considered, such as social vulnerability, segregation and informality (Mendoza and Jara 2020).

Some disasters have been studied using multiple approaches. One example is the 1993 Midwest flood in the US. Xiao and Drucker (2013), Xiao (2011), and Xiao and Feser (2014) studied its effects through time series econometrics and impact evaluation techniques, finding that, on average, counties with greater economic diversity experienced rapid growth in employment, showing evidence of resilienceFootnote 1 in their labor markets. They also find negative effects on personal income and increased unemployment rates in the short run and for the most affected counties. Apparently, local economies in aggregate seem to be resilient to natural disasters, being able to absorb the shock caused by the flood, although the agricultural sector seems to have experienced long lasting effects.

Xiao and Drucker (2013) specifically linked economic diversity with resilience, and explain this relationship because the effects of shocks in a specific place are less extensive if few industries are the main source of employment, similar to a portfolio effectFootnote 2 (Frenken et al. 2007; Malizia and Ke 1993).

Ohtake et al. (2012) study the effect on the Japanese labor market of the Great Hanshin-Awaji earthquake (1995) over time. They found that the number of job placements was reduced for part-time workers in the short run, possibly due to a shortage of job offers. For full-time workers, they found that growth in the number of job placements decreased abruptly, possibly due to a mismatch between labor supply and demand.

Other studies have focused on understanding what happens in the post-disaster labor market from a broad perspective (see Kirchberger 2017). Belasen and Polachek (2008) find a positive sectoral impact for hurricane-hit counties in construction and services, while a negative one in manufacturing, commerce, transportation, utilities, finance, investment and real estate. Mehregan et al. (2012) studied the composition of the labor market in Bam (Iran) after the 2003 earthquake through a shift-share analysis. They found that for affected zones, manufacturing and mining gained employment between 1996 and 2006 (increased their shares), which probably occurred due to the reconstruction process. Similarly, Mueller and Quisumbing (2009) looked into the effects on the labor market caused by the 1998 flood in Bangladesh, and found that non-agricultural labor markets are more affected in the long run (reduction of wages) possibly because they depend on the recovery of other markets, while on the other hand, agricultural labor markets had a negative effect in the short term. Sectoral changes seem to happen frequently after a disaster, at least in the short run, some economic sectors might be winners while others might be losers (Xiao and Nilawar 2013).

There is also a strand of literature that studies post-disaster labor markets tied to reconstruction. Chang and Rose (2012) review the literature that includes both empirical studies and formal models. Among the main findings they point to significant increases in employment in the short run as a result of the reconstruction (Dacy and Kunreuther 1969; Chang 2010), with sectors related to reconstruction showing significant gains (Chang and Rose 2012). It has been shown that pre-disaster economic trends and performance are usually accentuated during the recovery period (Chang 2000, 2010; Dahlhamer and Tierney 1998; Alesch et al. 2009). The authors conclude that the more severe the disaster is, the more it can cause structural changes in the affected local economies in the long run, such as sectoral composition, competitiveness and business types (Alesch et al. 2009; Lam et al. 2009; Chang 2010). Ohtake et al. (2012) indicate that after the reconstruction period the number of vacancies, job searches and jobs placements would be lower than pre-earthquakeFootnote 3 levels if there were initial spillover effects. In the first years after a disaster, households should find their own way to restart their lives, some authors believe that the best way to get out from the “infantilization trap” of earthquake victims is through labor (Castorina and Pitzalis 2019).

Our article also aims to separate labor market effects by gender. Many reports have stated the importance of gender equality (Bradshaw 2004) in post-disaster scenarios, as women might be more likely than men to lose their job or decrease their wages. This undesirable effect can occur in any kind of economic crisis (Hallward-Driemeier et al. 2017; Weisser 2019), but in the case of disasters reconstruction policies have often ignored gender issues. Women can lose their jobs because of general labor market responses (Möller and Aldashev 2006; Kirchberger 2017), but also because households dynamics can lead women to be less encouraged to participate in labor markets in times of crisis (Dagsvik et al. 2013). Men might ask their wives to stay at home and to not look for a job for many reasons specific to each culture. These phenomena are not only driven by pure market reactions, Lewis (1992) gives some insights about how welfare regimes could help or discourage female work participation.

Even after all this evidence, it is difficult to grasp a generalized insight on what are the possible outcomes for labor markets that experienced these types of shocks. Our article tries to shed light on this literature by performing a simple descriptive analysis in three countries that have comparable enough data.

3 Data set and methodology

As the number of cross-country academic papers increases, it is of utmost importance to make cases comparable with similar data and methods of evaluation (Maddison 2005; Jilke et al. 2015). Because of its relative novelty as a research field, disaster studies have strongly relied on cross-country comparisons. It is very important to compare coherent cases with similar data and process them with the same methodology (Smith 2001; Maddison 2005; Beghelli et al. 2020). For this article, the three selected case studies are Ecuador, Chile and Italy, in order from the latest to the oldest disaster.

3.1 Data set

For all three countries, we will use short-term micro data. For the Ecuadorian case we used a two-period data panel from December 2015 (pre-earthquake) to December 2016 (post-earthquake) of the National Survey of Employment, Unemployment and Underemployment (ENEMDU), developed by the Ecuadorian Institute of Statistic and CensusesFootnote 4 (INEC). For Chile we use the post-earthquake National Socioeconomic Characterization Survey (CASEN) which was an extension of a cross-sectional survey taken in 2009, repeated in 2010 for six regions in central Chile. For Italy we use the “Rilevazione sulle Forze di Lavoro—Dati trasversali trimestrali” compiled by the National Statistics Institute (ISTAT), containing information for 2009 (post-earthquake) and 2008 (pre-earthquake). On this case the data was collected for Abruzzo and its neighboring regions.

In order to have a clear and meaningful comparison, we only use variables that are similarly collected across countries. Since the surveyed populations differ, the comparable groups of analysis are not whole nations, but only the affected regions for each country. So, there are between 1.3 and 3 million people in the most affected areas (with the presence of both seaside and mountainous areas) and about 15 million people in the overall population for all cases. The country capital cities (Roma, Quito and Santiago) are included in the analysis. The most affected regions (provincias) for Ecuador are Esmeraldas and ManabíFootnote 5. In Chile the most affected regions were Biobío and Maule, while the others are O’Higgins, Valparaiso, Metropolitana and Araucanía. In Italy, the Abruzzo region was the most affected while Umbria, Marche, Lazio and Molise comprise the rest of the sample. In the final datasets (cleaned databases), the number of observations per year for Ecuador are 51,126, 58,293 for Chile and 19,695 for ItalyFootnote 6.

Because of several mismatches in survey design and the inter-temporal construction, the number of independent variables are restricted to the comparable ones, such as occupational status, sex, age, education, household size, a dummy for self-employment, dummy of work stability, and the economic sectorFootnote 7.

3.2 Methodology

In our research we use individuals distributed across different regions as the unit of analysis (provinces for Ecuador, regions for Chile and Italy). The individuals located in the most affected regions (see Fig. 1) are compared in pre and post-earthquake scenarios in the short term with the individuals in their respective surrounding regions (with exception of Ecuador in which we can use the remaining provinces as control). A more granular or nation-level analysis was not applicable because we lacked the proper data for Italy and Chile, so in order to find a proper control cluster of regions, the analysis had to be implemented just for the closer ones, so results are fully comparable.

Fig. 1
figure 1

Most affected regions by country. (Own elaboration)

Full size image

Although we do not attempt to find causal relationships in this article, it is very important to note that our identification strategy assumes total exogeneity of the shocks. We believe that this is not a problematic assumption, as the two South American countries are effectively located inside a continental fault line, making self-sorting within both countries unfeasible. In Italy this is different at a national scale, but for the treatment and control regions used in this study (central Italian regions) the assumption still holds.

Short term labor market changes are assessed through different econometric models: Mincerian wage regressions, difference in difference (DD) models to estimate shifts in wages and working hours, and logistic regressions implemented as a nonlinear DD to look for changes in the probability of being employed.

In order to understand gender dynamics in a post-earthquake scenario, the regressions for labor income and worked hours will be estimated for men and women altogether, and then as separate estimationsFootnote 8. We estimate them separately because men and women could have different reasons to decide to participate in labor markets, so their reserve salary could be determined by different variables.

Brown et al. (2011) analyzed the differential in reserve salaries between men and women, finding that the presence of children at home increases the reserve salary more for women than for men, this would be explained because women would experience higher opportunity cost for taking a job outside the home. Caliendo et al. (2017) found that the wage gap between men and women is reduced once the reserve salary was incorporated as a control variable, this would indicate that the gender differences in reserve salaries would determine the observed differences in the wage.

First, we use wage Mincerian equations (Mincer 1974) to identify changes in labor income and working hours (data panel regressions), our coefficient of interest is the one associated with the natural disaster (d.earthquake1, see Eqs. 1.1 and 1.2):

Equation 1.1

Mincerian regression, labor income

$$\ln \left(wage\right)_{it}=\alpha +{\beta _{1}}d.\text{earthquake}1_{i}+\beta _{2}\text{schooling}_{it}+\beta _{3}\text{potencial}\_ \text{experience}_{it}+\beta _{4}{\text{potencial}\_ \text{experience}_{it}}^{2}+\beta _{5}d.\mathrm{sex}_{i}+\beta _{6.kt}d.\text{ecomic}\_ \text{sector}_{ik,t}+\beta _{7}d.\text{selfemployment}_{it}+\beta _{8.\mathrm{skt}}\left(d.\text{ecomic}\_ \text{sector}_{ik,t} x d.\mathrm{sex}_{i}\right)+\beta _{9.\mathrm{sft}}\left( d.\text{selfemployment}_{it} x d.\mathrm{sex}_{i}\right)+\beta _{10}d.\mathrm{time}_{t}+\beta _{11}\ln \left(\textit{worked}\_ \textit{hours}\right)_{it}+\beta _{12}d.\text{stability}_{it}+\varepsilon _{it}$$

Equation 1.2

Mincerian regression, worked hours

$$\ln \left(\textit{worked}\_ \textit{hours}\right)_{it}=\alpha +{\beta _{1}}d.\text{earthquake}1_{i}+\beta _{2}\text{schooling}_{it}+\beta _{3}\text{potencial}\_ \text{experience}_{it}+\beta _{4}{\text{potencial}\_ \text{experience}_{it}}^{2}+\beta _{5}d.\mathrm{sex}_{i}+\beta _{6.kt}d.\text{ecomic}\_ \text{sector}_{ik,t}+\beta _{7}d.\text{selfemployment}_{it}+\beta _{8.\mathrm{skt}}\left(d.\text{ecomic}\_ \text{sector}_{ik,t} x d.\mathrm{sex}_{i}\right)+\beta _{9.\mathrm{sft}}\left( d.\text{selfemployment}_{it} x d.\mathrm{sex}_{i}\right)+\beta _{10}d.\mathrm{time}_{t}+\beta _{11}d.\text{stability}_{it}+\varepsilon _{it}$$

Second, we apply a difference in difference (DD) approach to find whether the earthquake had a robust effect on labor income and worked hours. In this case, our coefficient of interest was the one associated with the interaction between the treatment (d.earthquake2) and time:

Equation 2.1

DD approach, labor income

$$\ln \left(wage\right)_{i}=\alpha +{\beta _{1}}d.\text{earthquake}2_{i}+{\beta _{2}}d.\mathrm{time}+{\beta _{3}}{d.\text{earthquake}2_{i}} x d.\mathrm{time}+\beta _{4}\text{schooling}_{i}+\beta _{5}\text{potencial}\_ \text{experience}_{i}+\beta _{6}{\text{potencial}\_ \text{experience}_{i}}^{2}+\beta _{7}d.\mathrm{sex}_{i}+\beta _{8.k}d.\text{ecomic}\_ \text{sector}_{ik}+\beta _{9}d.\text{selfemployment}_{i}+\beta _{10.sk}\left(d.\text{ecomi}{c_{\text{sector}}}_{ik} x d.\mathrm{sex}_{i}\right)+\beta _{11.sf}\left(d.\text{selfemployment}_{i} x d.\mathrm{sex}_{i}\right)+\beta _{12}\ln \left(\textit{worked}\_ \textit{hours}\right)_{i}+\beta _{13}d.\text{stability}_{i}+\varepsilon _{it}$$

Equation 2.2

DD approach, worked hours

$$\ln \left(\textit{worked}\_ \textit{hours}\right)_{i}=\alpha +{\beta _{1}}d.\text{earthquake}2_{i}+{\beta _{2}}d.\mathrm{time}+{\beta _{3}}{d.\text{earthquake}2_{i}} x d.\mathrm{time}+\beta _{4}\text{schooling}_{i}+\beta _{5}\text{potencial}\_ \text{experience}_{i}+\beta _{6}{\text{potencial}\_ \text{experience}_{i}}^{2}+\beta _{7}d.\mathrm{sex}_{i}+\beta _{8.k}d.\text{ecomic}\_ \text{sector}_{ik}+\beta _{9}d.\text{selfemployment}_{i}+\beta _{10.sk}\left(d.\text{ecomi}{c_{\text{sector}}}_{ik} x d.\mathrm{sex}_{i}\right)+\beta _{11.sf}\left(d.\text{selfemployment}_{i} x d.\mathrm{sex}_{i}\right)+\beta _{12}d.\text{stability}_{i}+\varepsilon _{it}$$

Third, we estimate logistic regressions models (nonlinear DD) for each country in order to explore the probability of being employed after the earthquake (see Eqs. 3.1 and 3.2):

$$Prob\left(x_{i}=1|z_{i}\right)=\frac{e^{\beta _{i}^{\prime}{z_{i}}}}{1+e^{\beta _{i}^{\prime}{z_{i}}}}$$

Equation 3.1

Logistic model

Where \(x_{i}\) represents the condition of being employed (variable employment equal to 1) and \(z_{i}\) is the vector of explanatory variables for each individual i. The vector of the explanatory variables that we use in our nonlinear DD estimations is:

$$z_i=d.\text{earthquake}2_{i},d.\mathrm{time},{ d.\text{earthquake}2_{i}} x d.\mathrm{time}_{,}\,\text{schooling}_i,\text{potential}\_ \text{experience}_i,\text{potential}\_ \text{experience}_{i}^2,d.\mathrm{sex}_{i},d.\text{selfemployment}_{i}$$

Equation 3.2

vector of explanatory variables, nonlinear DD for employment

The label descriptions for the variables in our models are:

ln(wage):

is the natural logarithm of individual monthly labor income.

ln(worked_hours):

is the natural logarithm of the individual monthly worked hours.

employment:

is a dummy variable taking the value of 1 if the individual is employed, 0 otherwiseFootnote 9.

d.time:

is a dummy time variable (1 post disaster, 0 pre disaster). Pre disaster periods are December 2015 for Ecuador, 2008 for Italy and the year 2009 for Chile. Post disaster periods are December 2016 for Ecuador, the year 2009 for Italy and 2010 for Chile.

d.earthquake1:

is a dummy variable taking the value of 1 if the individual is located in the seismic areas in the post-earthquake year, otherwise 0.

d.earthquake2:

is a dummy variable taking the value of 1 if the individual is located in the seismic areas, otherwise 0.

schooling:

is the individual’s years of education.

potential_experience:

was calculated as age—schooling—six.

d.sex:

is a dummy variable for sexFootnote 10 (1 female).

d.economic_sector:

are dummy variables for “k” economic activities (see footnote 7).

d.selfemployment:

is a dummy variable that takes the value of 1 if the individual was an employer or self-employment, otherwise 0.

d.stability:

is a dummy variable which takes 1 if the individual has work stability (permanent contract or job appointment), 0 otherwise.

4 Case studies: Abruzzo (Italy), Biobío and Maule (Chile), and Manabí and Esmeraldas (Ecuador)

Jowell (1998) and Maddison (2005) explained the issues about the selection of case studies in cross country comparison for social sciences: The cases should present a right mix of similarities and differences, their number should be not too large but possibly more than two, in order to be able to describe them in depth and to prevent possible syncrasies. Finally, the authors should not be completely alien from their context.

The similarities identified in advance are: all three countries are earthquake-vulnerable across the whole country (no feasible self-selection in earthquake-free areas); they have comparable size and population (the biggest one is no more than three times the smallest) and morphology. They also had a similar independent State history (independence in 1800s, liberal and authoritarian regimes, democratic transitions and a slow ongoing process from a centralized to a decentralized State in the late decades), same majority religion, similar languages, legal systems (civil law), and relative currency stability.

Their main differences are their geopolitical role, their gross and per capita GDP (Italy 2009: 2.391 trillion USD—40,640 per capita. Chile 2010 172.4 billion—10,243 per capita. Ecuador 2015 99.29 billion—6150 per capita, according to the World Bank). They also experienced very different macroeconomic trends before and after the disaster in study: Italy was at its peak, and since the earthquake struck it has experienced a severe crisis, mainly due to the European financial turmoil in 2008–2011. In Chile the reverse was happening, since 2009 was the middle of a rapid, yet volatile growth cycle. Ecuador was at the end of a sustained growth period that was becoming flatter after the earthquake hit. Their economic models of growth are also different, the Chilean one had more merits to be called neoliberal at the time, while the Italian and Ecuadorian ones were more balanced. Indexes of economic freedom show this: Chile (18th), Italy (80th) and Ecuador (170th). Most of these differences are given by government spendingFootnote 11 (UN 2018). Ecuador and Italy present similar characteristics with respect to Chile when workers’ rights are the center of analysis. Starting from a fundamental rights point of view, the Chilean Constitution recognizes to workers the fundamental rights given by human nature and the “freedom of work and its protection” (art 5 and 19). The Ecuadorian Constitution explicitly considers work as a “right and a social duty” (art 33), deeply articulated in its right and subject to State intervention (art. 34, 37, 38, 39, 47, 66, 276, 284, 319–333). Italian Constitution is even more explicit, starting with a clear “Italy is a democratic Republic founded on Labor” (Art. 1), with a clear stand on workers’ rights, the right to work and the role of the State actively promoting these (art. 4, 35–40, 46). All three States have their own Worker Statute Law, comprehending contract norms, workers securities and union’s rights.

Regarding the role of the State in case of disaster, the Constitutional framework is different: the Italian text ignores the issue, mentioning only in a note the possibility of exceeding the national budget (art. 81), forwarding to lower rank laws any protection from disaster hazards. The Chilean Constitution explicitly introduces the state of catastrophe (art. 32, 40, 41) where the Government can exceed its normal powers, also restricting workers freedom, but there is no mention of any risk management. In the Ecuadorian Constitution the concept of risk reduction is widely introduced, with rights in case of disaster and the possibility of a state of exception (art. 35, 38, 46, 164, 261, 389, 397).

In all three nations, norms about risk reduction and disaster management are numerous and prone to change after each occurring disaster (especially in Italy) so it is not easy to predetermine precisely how each state could have reacted to such disastrous events presented hereafter.

4.1 Brief description of the earthquakes

4.1.1 2016 April 16th: Ecuadorian Earthquake

The Ecuadorian coast was shaken by a 7.8 Mw earthquake on April 16, 2016 (20 km deep), its hypocenter was located off the coast of Manabí, a province which also suffered the highest casualty rate. The provinces of Manabí and Esmeraldas were the most affected regions in terms of infrastructure damages. The disaster left a total of 671 human lives lost, 113 individuals were rescued alive, and in the three days after the earthquake almost 4859 people received health care. The areas most affected by the disaster were vulnerable areas, both physically and socio-economically. In general, the estimated reconstruction costs reached the amount of 3344 million USD: 1369 million in the social sector, 862 million in the infrastructure sector, 1032 million in the productive sector, and 81 million in other costs (IG-EPN 2018; Committee for Reconstruction and Productive Reactivation 2016).

The national government quickly developed actions and regulations for the recovery of the affected areas. On May 20Footnote 12, 2016, the Ecuadorian government passed a law stipulating solidarity contributions to be collected on remunerations, wealth, utilities, property and representative rights of capital existing in Ecuador. The value added tax was temporarily increased by 2 percentage pointsFootnote 13 throughout the country, except for the provinces of Manabí and Esmeraldas. Incentives were established for new productive investments in these two provinces.

By May 30, 2018, 2876 million USD were allocated, which were distributed as follows: 2100.8 million for reconstruction, 412.6 for productive reactivation and 362.6 for emergency. As of April 14th, 2017, the Productive Reactivation Committee indicates that more than 44,000 direct and indirect jobs have been generated (80% identified as local unskilled labor). Finally, the Technical Secretariat of the Committee of Reconstruction and Productive Reactivation (2019), reported that by October 2018 Public Banking had placed 689 million dollars in the provinces of Manabí and Esmeraldas.

4.1.2 2010 February 27th: Chile earthquake

On the night of February 27th 2010, an earthquake of 8.8 moment magnitude (Mw) shook the central part of Chile for four minutes long. The epicenter was located 17 km from the coastal town of Cobquecura, the magnitude and the distance induced a tsunami wave that hit the Chilean coast 35 min after the seismic shake. The event left 525 official losses, of whom 100 perished because of the tsunami, about 500,000 people were immediately displaced, and more than 200,000 buildings reported damages. The earthquake mainly affected six Chilean regions (Biobío, Maule, O’Higgins, Valparaiso, Metropolitana and Araucanía). But, while in the regions away from the epicenter most of the damages were due to a scattered single building vulnerability, the catastrophic outcome emerged mainly in the two regions close to the epicenter: Biobío and Maule.

The two regions most affected by the 2010 earthquake were both poorer in terms of GDP than the national average, but Biobío was slightly richer than Maule. Their economy is mainly based on forestry, fishery, other agricultural activities and energy production. The 2010 earthquake affected highly dense areas and the two coastal cities of Concepción (Biobío) and Constitución (Maule), where the signs of the earthquake are still persistent almost ten years later.

In April 16th, 2010, the Presidency released a reconstruction plan where almost 30 billion USD were estimated as the overall cost of the catastrophe. To boost reconstruction, the government engaged 8.4 billion USD for the 2010–2014 period, mostly earmarked for public infrastructure projects (Government of Chile 2010). The remaining part was supposed to be covered by 6.2 billion USD from private insurance companies, but most of these funds were never disbursed because of insolvency (Brain and Mora 2012). The last half of the remaining costs have not been explicitly covered. No explicit post-disaster policies have been enacted for the creation jobs or local development initiatives. In 2010, only 16,634 workers got unemployment subsidies.

4.1.3 2009 April 6th: Italy earthquake

At 3:32 am on April 6th, 2009, the city of L’Aquila (Italy) was hit by a 6.3 Mw earthquake, the highest of a long seismic swarm. 309 inhabitants died, 1600 were injured and about 100,000 were evacuated from L’Aquila and the surrounding municipalities during the following weeks. Differently from the South American earthquakes, European and especially Italian seismic events are characterized by lower, yet highly concentrated energy releases in smaller areas prone to devastation, either because of shorter fault lines, complicated geology or older construction techniques (Lam et al. 1996; Viti 2019). The Abruzzo region, where L’Aquila is the capital, has always been one of the most vulnerable regions of Italy, both in terms of socio-economic and seismic vulnerability. Besides the cost in human lives and livelihoods, the capital losses were also enormous and remarkable since more than 22,000 private houses were damaged, 2000 firms suspended their business and the total cost from the earthquake is estimated at 11.2 billion USD.

On April 28th, the government approved the emergency law Decreto Abruzzo (L 77-2009) as well as the first reconstruction norms, which described no-tax areas and social securities for the affected population, including unemployment subsidies for self-employed workers and employers. In 2012, this effort was followed by a larger law for overall reconstruction (L 83-2012), bringing the total public reconstruction expenditure at almost 13 billion USD, more than the original cost. This huge amount of public funds should have been used mainly for private housing reconstruction subsidies, but a relevant quota (530 million USD) ended up going instead for local development projects. Table 1 summarizes the most important information for the three earthquakes.

Table 1 Summary information for the case studies. (Own elaboration)
Full size table

An important point about these three earthquakes is that both emigration and immigration represented a very limited phenomenon, at least in the short term time period we are examining here. This is important to note, as attrition, representativeness and biases could have increased if these disasters had also triggered flows of people across geographic areas.

5 Results

5.1 Labor income distributional differences

To describe how labor income shifts developed in each country we first performed a quantile comparison of labor incomeFootnote 14 between two years, first for affected areas and then for unaffected areas of each country using the Harrell and Davis (1982) quantile estimator for robust tests for two independent groupsFootnote 15.

In the Ecuadorian case (see Table 2), there was a positive and significant difference in labor income only for the first wage quintile in the affected areas in 2016, but for unaffected areas there was a negative difference in 2016 for the first and second quintile.

Table 2 Ecuadorian labor income quantile difference (natural logarithm). (Source: National Survey on Employment, Unemployment and Underemployment (ENEMDU). Own elaboration)
Full size table

In Chile no wage differences were found for the affected areas, only for unaffected areas there was a negative and significant wage difference for the first, third and fourth quintiles (Table 3).

Table 3 Chilean labor income quantile difference (natural logarithm). (Source: National Socioeconomic Characterization Survey (CASEN, post-earthquake). Own elaboration)
Full size table

In Italy there were positive and significant wage differences in 2016 for the affected areas in the first to fourth quintiles, both in affected and unaffected zones (see Table 4).

Table 4 Italian labor income quantile difference (natural logarithm). (Source: Rilevazione sulle Forze di Lavoro. Own elaboration)
Full size table

In summary, we found heterogeneous wage differences between years for some quantiles. In Ecuador only the first income quintile in the affected zones showed statistically significant wage increases, while unaffected areas in the first two quantiles showed income decreases. For Italy, both seismic and not seismic areas had wage increases (quintile 1–4), only for the first quintile is higher for affected locations. Finally, in Chile no significant differences were found for affected territories. In the final section we discuss how future research could explore deeper questions about the distribution of wages in post-disaster labor markets.

5.2 Wage regressions

For the panel regression and the difference in difference estimates, we performed separate regressions for men and women, we also made joint regressions with interactions for robustness. Differences in the reserve salaries could due to gaps in productivity and expectations of both groups, since men and women have very different reasons for deciding whether or not to participate in the labor market (Brown et al. 2011; Caliendo et al. 2017).

For data panel models we performed a Hausman test, which suggests that the more appropriate models were fixed effect estimations, so we report only these models for all three countries. In Ecuador we found a positive association between the earthquake, labor income and working hours (11.6% and 6.5% respectively for males and females jointly, Table 5). The earthquake labor income coefficient for males was greater in magnitude than for females, but the coefficient of worked hours was higher for females. In Italy, we found an increase in worked hours (3.8% for men and women jointly, 9.8% for females), but not for male workers. Surprisingly, in Chile we did not observe statistically significant associations between the earthquake, wage changes and working hours.

Table 5 Mincerian Regressionsa (data panel). (Source: ENEMDU, CASEN, ISTAT. Own elaboration)
Full size table

5.3 Difference in difference (DD) approach

The double difference estimations were coherent with the Mincerian wage regressions. We found a positive earthquake effect in wages and worked hours in Ecuador (11% and 6% respectively, joint regressions, Table 6). Also, we found that the wage coefficient was higher for men and the working hours coefficient was greater for women. It seems that, in the short term, the Ecuadorian labor markets in the regions affected by the earthquake improved their performance in these two dimensions. For Italy, we found an effect only in working hours but not in labor income, and we also found a gender effect, since only women experienced an increase in their working hours. It is possible that reserve salaries for women might have decreased due to the earthquake in some affected regions. In Chile there was no significant effect, neither in labor income nor in worked hours.

Table 6 Double Difference Estimationsa. (Source: ENEMDU, CASEN, ISTAT. Own elaboration.)
Full size table

5.4 Employment status logistic regressions

We performed a regression analysis to compute how the three earthquakes are associated with the probability of being employed in the immediate post-disaster months using a nonlinear DD. As in previous regressions, comparable samples (most affected regions and the rest of the territories) were included in the regression analysis for all three countries.

The parameter of interest is the interaction term between time (post-earthquake) and being in one of the most affected regions (Manabí and Esmeraldas for Ecuador, Maule and Biobío for Chile and Abruzzo for Italy). Controlling for other variables, we did not observe statistically significant parameters for interaction terms in any of the three countries (see Table 7).

Table 7 Logistic regressions for employment dummy (average marginal effects, nonlinear DD)a. (Source: ENEMDU, CASEN, ISTAT. Own elaboration)
Full size table

6 Discussion and conclusions

6.1 Discussion

Understanding the effects of an earthquake on the workforce during the first months after the event is not a linear process. After having identified changes in working hours and wage parameters, the first distinction to adopt in this discussion is between the most affected regions and the surroundingFootnote 16 population, starting from a simple comparison of employed, unemployed or inactive people before and after the earthquake.

In Ecuador there was a reduction in the share of unemployed and inactive citizens in the most affected areas (defined in our research). This trend of rising labor indicators could have been caused by a young demographic structure and general State interventions to boost the local economy in the short run. In Ecuador, the most vulnerable workers (lower quintile) faced a significant increase in wagesFootnote 17, while in unaffected areas we observed the opposite (negative difference in 2016 for the first and second quintile).

Despite the magnitude of the earthquake and the relatively low levels of social protection, Chile faced the earthquake with good resistance, both in the most affected regions, as well as the rest of the country, showing only a slight transition from unemployment to inactivity. We observed that wages did not face significant changes in the different quintiles. The Italian case is very peculiar, an old population in the year of an economic crisis made the Abruzzo region more likely to increase unemployment, even considering that there were no subsidies for workers passing from inactivity to unemployment. Note that the most affected regions in the three nations are very similar under a profile of peripheralityFootnote 18 with respect to the national economic flows. Our regression analysis shows that the earthquake seemed to improve wages and working hours in Ecuador, while in Chile the disaster did not seem to influence labor income, nor working hours. In Italy, only an increase of working hours was found.

The gender gap is multidimensional and strongly persistent in the three cases, where women have lower probability of being employed (see sex dummy in Table 9 in the Appendix). The heterogeneity of each case is relevant, in Ecuador both men and women earn and work more due to the earthquake, in Italy only women increased their worked hours, but labor income did not increase overall. In Chile no significant effects were found.

It is important to mention that all three labor markets are different from each other. In Ecuador, a large proportion of the workforce (nearly 50%) is employed in the informal sector, the percentage is even higher if we use informal employment definition (Mendoza and Jara 2020). This has implications for the provision of public services, as the Ecuadorian Social Security Institute only guarantees formal workers access to health and retirement pensions. The Chilean market has a legal workday of 45 h, while in Ecuador and Italy the legal workday is 40 h. In many ways, Chile is also a country that has focused its economy on market liberalization for the last decades, yet it has not guaranteed some social protection benefits for workers (i.e. low retirement pensions). The Italian labor market at the time (2009) was mainly made up of formal workers, it is these workers who have access to social security.

6.2 Conclusions

This brief study of short run effects of earthquakes on local labor markets shows different scenarios where responses and impacts had differences and similarities across all three countries. The speed of recovery could depend both on intrinsic characteristics of the regions and the capabilities of policy makers to help these processes, as the Ecuadorian dynamism, Chilean resistance, and Italian support has shown. The amount of public money spent for reconstruction and local development could have played an important role on short term labor market reactions, although some of these funds might have also exacerbated expectations from local economic agents. We saw how a smaller and poorer, but younger and more dynamic economy such as the Ecuadorian was able to react in a decisive way to the disaster shock. The case of Chile, a mature economy with a more liberalized system, did not seem to suffer considerable short run effects, as the Chilean labor market reacted well due to increased economic activity in the years following the earthquake. At the same time, an older economic system such as the Italian one, seemed to experiment short run effects in working hours, but not in workers’ salaries in affected areas. However, an increased State role during catastrophic events appears to be necessary in order to maintain social cohesion and securities.

In conclusion, this paper remarks the importance of institutions, the economy and public support have after a disaster in the short run. In that moment, when the attention is mainly paid to basic needs such as food or shelter, affected households function with relevant changes in their working conditions: some families can be forced to find other sources of income and women are often at a disadvantage when gender equality policies are not considered in these scenarios. Investigating previous short-term responses is important for policymakers to adjust disaster responses to the needs of societies and avoid some pitfalls that other countries experienced (such as the compounding effect the financial crisis had in Italian labor markets).

Work is not only important in the mid-run from a disaster, in order to assess the economic development of an affected area, it is also crucial to give proper attention to post-disaster employment conditions in the first months. An integrated system of workforce protection may help affected households to avoid a second personal disaster after the collective one.

There are many challenges for future research in this topic, we show a first distributional approach with the salary comparisons per quintile in our manuscript. According to our results, Ecuador is a special case in which there was an average salary increase for the first quintile in seismic areas, but there was an average wage decrease for unaffected areas in the first and second quintiles between the pre and post-earthquake periods. In this article there is not enough evidence to conclude that an existing distributional effect was found. For this reason, causal distributional approaches are among the future challenges in order to reveal whether earthquakes affect local labor markets homogeneously (or not) across the wage distribution. This could be a crucial part in designing good mechanisms in public policy to minimize potential damages that earthquakes could cause to workers, and the labor markets they comprise.

Further research directions in this area can be both plentiful and offer a lot of insight, especially when comparing multiple countries that have suffered similar disasters. Ohtake et al. (2012) suggest to study changes in industrial structure, which could help policy makers to identify economic sectors in need, and guide the post-disaster economy with better tools. Ewing et al. (2005) suggest that new research questions might focus in determining which characteristics make resilient regions, a very broad concept that has been explored in depth in this literature, but rarely using cross-country comparisons with microdata. Xiao and Feser (2014) also point out that future research could focus on determining the influence of pre-disaster economic conditions (sectoral structure, urban and rural composition, employment and unemployment), on the labor market reaction and adjustment over long periods.