The short-term impact of the 2020 pandemic lockdown on employment in Greece

Since the onset of the COVID-19 pandemic, the consequences for economies and labor markets around the world have been immense (e.g., Alstadsæter et al. 2020; Béland et al. 2022; Cowan 2020; Coibion et al. 2020; Cajner et al. 2020; Forsythe et al. 2020; Campello et al. 2020; Bauer and Weber 2021; Kong and Prinz 2020; Goolsbee and Syverson 2020; Aum et al. 2021). As an exogenous, transitory shock, it directly impacted economic activity through a plummeting consumer demand (Kong and Prinz 2020; Kim et al. 2022; Jordà et al. 2022), and it generated heterogeneous effects across several dimensions of economic activity (Barrot et al. 2021; Guerrieri et al. 2022). Also, it forced governments to impose restrictions on both mobility and economic activity, in an attempt to mitigate the virus transmission. However, the demand for goods and services decreased mainly because individuals adjusted their behavior to protect themselves from being infected, especially in times with overwhelmed healthcare systems, and months before the vaccine rollout (Goolsbee and Syverson 2020; Kapetanios et al. 2022). In the short-run, such behavioral adjustments were responsible for the worsening economic conditions (Baek et al. 2021). As a result, the demand for labor decreased. At the same time, labor supply fell because people stayed at home due to concerns about individual and public health. Overall, the immediate responses in terms of employment came through higher separation rates (Adams-Prassl et al. 2020), fewer hires (Guven et al. 2020; Aum et al. 2021; Casarico and Lattanzio 2022; Zou et al. 2022), and fewer vacancy postings (Hensvik et al. 2021; Bamieh and Ziegler 2020).

Despite the timing of the pandemic onset being nearly common for everyone, the short-run labor market effects depended on the structure of economic activity (i.e., region-specific sectoral variation). Therefore, searching for the mechanisms behind the observed short-term employment adjustments will provide insights regarding the process of structural labor reallocation, compatible with a theoretical framework of labor market frictions (Şahin et al. 2014). As Pizzinelli and Shibata (2022) point out, the COVID-19 pandemic did not set in motion a large process of structural reallocation, since the initial exposure of different sectors to the pandemic resulted primarily from the short-run impact of virus transmission concerns, behavioral adjustments, and various mitigation policies. To provide some answers regarding the above, this paper attempts to identify (a) the short-run impact of the pandemic on labor market outcomes, and (b) whether the observed short-term employment losses were due to lower hiring or higher separation rates. Using individual level data from the Greek Labour Force Survey we rely on the region-specific variation in economic activity, and define sectors that were severely affected by mobility restrictions (e.g. transportation, accommodation, restaurants etc.) versus those that were not (e.g. electricity, public administration, education etc.). We should mention three interesting features about the Greek case. First, the economy is traditionally exposed to strong seasonal patterns (OECD 2020).¹ Second, before the pandemic, the country was recovering from a prolonged period of economic recession.² Third, the initial policy response after the pandemic onset was focused on protecting existing jobs and providing financial support to firms mostly affected. A key condition for firms to receive support was that they temporarily prohibited layoffs, a decision aligned with the general consensus among EU policy makers at that time (Giupponi et al. 2022; Ando et al. 2022; Schelkle 2021).³ Constructing sector-region cells on a quarterly basis, and under a difference-in-differences (DiD) framework we compare labor market outcomes (job finding, job separation, employment, and labor force participation rates) before and during the early pandemic phase between affected and not affected cells.

Our estimates show that employment rate was around 9 p.p. lower relative to its pre-pandemic level. Employment losses were due to declines in the demand for labor after the pandemic onset in tourism-exposed cells, which primarily drive the seasonal employment patterns.⁴ This result was solely due to a lower hiring rather than a higher separation rate. Our findings are in line with evidence for the US (Forsythe et al. 2020), Canada (Larue 2021), Korea (Aum et al. 2021), and Sweden (Hensvik et al. 2020), where the employment slowdowns in the early 2020 were due to fewer vacancies in the hospitality sectors. Separation rates were not significantly increased due to the lockdown and this is attributed to the policy measures that protected existing jobs by prohibiting layoffs in suspended sectors. This conclusion contrasts the increased separation rates reported by Adams-Prassl et al. (2020) for the US and Lemieux et al. (2020) for Canada, where job retention was less of a priority. Our findings have important policy implications regarding the severity of labor market disruptions as responses to the nature and the timing of exogenous shocks that are unrelated to the functioning of the economy and the labor market.

The remainder of this paper is organized as follows. Section 2 presents some stylized facts on mobility and labor market trends before and during the early pandemic period in Greece, and it briefly describes the lockdown and mitigation measures introduced by the government. Section 3 presents the identification strategy, the data sources and provides summary statistics. Section 4 outlines the empirical methodology and presents the estimation results. Section 5 concludes.

The first COVID-19 case was confirmed on February 26, 2020. The outbreak peaked in early April with about 100 new cases per day. This relatively low number of cases was mostly due to the fact that the government reacted quickly, within three weeks after the initial recorded infection, and adopted policy measures that restrained the initial spread, including school closings, closing of all non-essential workplaces, and finally, issuing a general stay-at-home order.⁵ Appendix Fig. 6 plots the timeline of the pandemic in terms of infections and public policy responses. Early in May, the government lifted the stay-at-home order, followed by the opening of schools, and a gradual reopening of commercial and workplace activities by the end of May. Similar to other countries, mobility patterns changed dramatically during the first pandemic wave (Appendix Fig. 7).⁶ Public transit and visits to workplaces and non-essential shops declined by 50%-80% compared to their pre-pandemic levels. At the same time, visits to essential retail stores (grocery stores and pharmacies) were affected to a clearly lower extent. There was a partial return towards the pre-pandemic baseline after the first lockdown was lifted. According to the Hellenic Statistical Authority (ELSTAT) April 2020 press release, suspended sectors covered 14.6% of firms and 25.4% of employees. Accommodation was the most affected sector (87% of firms and 94.2% of workers) and the most exposed regions were South Aegean and the Ionian Islands, where the share of suspended firms was 34.4% and 29.8%, respectively. The severity of the pandemic on tourism-exposed sectors was also reflected in a 94% drop in total revenues in the accommodation sector during the second quarter of 2020, when the overall decline in the economy was 25% (ELSTAT). These suspensions and mobility restrictions concurred with the seasonal peak in job hiring. According to pre-pandemic Ministry of Labor and Social Affairs monthly press releases, new hires during the second quarter of 2019 accounted for 30% of all hires during the year, and half of those were made by firms in the accommodation sector. This sector is very important for the Greek labor market. In 2019, it accounted for 43% of the total number of hires nationally, with the majority of these registered during the second and third quarter of the year.

The demand shock and the subsequent restrictions imposed by the government resulted in a major economic slowdown with significant consequences for firms and workers (Economides et al. 2022). In 2020, real GDP declined by 8.2%, compared to 2019 (ELSTAT).⁷ Seasonally adjusted ELSTAT estimates showed that in May 2020 (relative to February 2020) the labor force participation rate and the employment rate decreased by 0.9 and 1.6 p.p., respectively, while the unemployment rate increased by 1.5 p.p. Figure 1 shows the year over year (2020 vs. 2019) change for each month (in each year February is used as the reference month) of basic aggregate labor market indicators (employment, unemployment, and labor force participation rates).⁸ These changes illustrate how employment and non-employment rates evolved during the early months of the pandemic. Assuming that the trend changes in 2020 reflect the impact of COVID-19, we observe that employment and labor force participation rates were lower, while the unemployment rate was higher in May 2020, compared to 2019.

To keep the economy afloat, the government mobilized a support package amounting to €6.8 billion (3.5% of GDP) in March and April, and approved an additional package of €24 billion in May in order to stimulate the restart of the economy. The first legislative act to support employers (March 11) included various firm-level liquidity measures.⁹ On March 18, the Ministry of Finance and the Ministry of Labor and Social Affairs introduced new measures,¹⁰ including an €800 stipend (covering the period from March 15 to April 30, eventually extended through May) to workers whose contracts had been suspended because of the suspension of operations of their workplaces. By early May, approximately 1.2 million employees and 550,000 self-employed and freelancers benefited from this support scheme.¹¹ Unemployment and long-term unemployment benefits, and unemployment allowance payments for the self-employed were extended by two months for those whose entitlement ended on March 31, and it was further extended to cover those whose entitlements ended at the end of April and May.¹² In addition, a lump sum stipend of €400 was introduced for 155,000 long-term unemployed individuals, registered with the Public Employment Agency (OAED) from April 1, 2019 who were maintaining their status until April 16, 2020 and were not receiving any other state benefit.¹³ In terms of numbers affected and financial commitment, the government’s mitigation measures emphasized the preservation of employment in firms operating in suspended sectors. To receive the benefits, a key condition was that affected firms were obliged not to reduce their pre-pandemic headcount. In fact, layoffs in suspended sectors of economic activity were prohibited from March 18 until June 16 when the restriction was lifted. Estimates from the National Institute of Labor and Human Resources in Greece (EIEAD 2021) show that during March–April 2020, almost half of the total salaried employment in Greece was affected by those schemes.

Aggregate evidence on whether the employment changes were due to changes in job hiring and/or job separation rates can be derived by daily data on job flows. Results presented in the Appendix sub-section A.1 show that the employment decline, by 7.8 p.p., (column 10 of Table 5) was entirely due to the collapse in job hiring rather than increases in job separations. For instance, the post-COVID hiring rate was 23.1 p.p. lower compared to what was expected, and almost half of this decline was due to the decrease in full-time hires. Similarly, the separation rate during the first pandemic wave was 14 p.p. lower and the majority of this decline was due to the 6.9 p.p. reduction in the contract termination rate. Moreover, to complement the evidence on how the labor market responded after the pandemic onset, we used daily vacancy data. The results (Appendix sub-section A.2) indicate a sharp decline in job vacancies due to the implementation of workplace restrictions in mid-March (Fig. 9), particularly in the accommodation and food services sectors (Table 6).

During normal times, firms base their employment decisions on their experiences and expectations, mostly shaped from seasonal patterns, regarding the demand for their goods and services. The pandemic outbreak interrupted this normality. The Greek government imposed horizontal mobility restrictions in all regions (international travel restrictions were also imposed), eliminating the ability of people to move to places outside their region of residence. However, despite the horizontal implementation of those restrictions, their impact on economic activity should not be expected to be geographically homogenous. Instead, it should depend on the exposure of each region to activities related to mobility of people across regions. Therefore, productive units whose economic activity depends strongly on interregional mobility should be more adversely affected, compared to productive units less economically dependent on such mobility. For example, people who planned to visit specific locations and spend some nights in local hotels could not do so in the presence of mobility restrictions. Hence, revenues in the associated sectors could not be realized. Therefore, employment in firms with tourism-related activities should be more affected compared to employment in firms not dependent on such activities.

To empirically identify the short-run labor market impact of the pandemic, we exploit the variation of the economic activity (2-digit NACE Rev. 2 classification, 88 sectors) within each region (NUTS-2 classification, 13 regions). We classify sectors into two groups, a treated (Transportation, Accommodation, Food & beverages, Real estate, Travel agencies, Creative activities, Gambling & Betting, and Sports activities) and an untreated one, based on the International Recommendations for Tourism Statistics 2008, UN/UNWTO taxonomy (Appendix, Table 7). To explore whether the above grouping identifies differences across regions regarding their exposure on those activities that are mostly affected by mobility restrictions, we use pre-pandemic (2019) information from ELSTAT on sales (turnover), and calculate for each region the share of sales generated by the treated sectors (as a percentage of that region’s total sales). In addition, we calculate the number of total arrivals (natives plus foreigners) in tourist accommodation establishments per 1,000 inhabitants (using the regional population of 2019).¹⁴ After sorting regions based on their tourism-related share of total sales, there are four regions that are relatively more exposed to tourism-related activities (South Aegean, Ionian Islands, Crete and North Aegean) compared to other regions (Fig. 2). Moreover, those tourism-exposed regions have the highest number of arrivals in tourist accommodation establishments per 1,000 inhabitants. Thus, grouping sectors into treated and untreated ones will help us to identify the short-run labor market impact of the pandemic, given that the sectoral composition is not homogenous across regions.

To validate the ability of our treatment assignment strategy to capture the responsiveness of economic activity on mobility restrictions, we show the monthly evolution of annual sales growth between January 2018 and June 2020 for treated and untreated sectors, at the national level (Fig. 3).¹⁵ Sales dropped considerably more in treated relative to the untreated sectors, e.g. 75% versus 25% in April 2020, respectively. This finding was also confirmed using firm-level information from the ICAP Data.Prisma dataset, that provides balance sheet data and information about the sector of economic activity (2-digit NACE Rev. 2) and NUTS-2 region for each firm. Using those data, we calculated the annual sales growth for each firm between 2020 and 2019. Then, for each region, we calculated the median sales growth for firms operating in tourism-related (treated) and untreated sectors (Fig. 4). We observe that the annual sales growth in the treated sectors was lower in regions that are more exposed to tourism-related activities.

Based on the above strategy to identify treated and control units, we calculated a series of labor market outcomes for each sector-region cell using individual-level data from the Labor Force Survey (LFS) covering the period 2015Q1-2020Q2. The LFS contains information on demographic, human capital, employment status and economic characteristics of surveyed households and their members. In particular, to construct the labor market outcomes of interest, we use information on the region of residence, and on current and past sector of economic activity at the individual level. More specifically, for individuals who are (a) employed in the surveyed quarter we know whether they started working during that quarter and (b) non-employed in the surveyed quarter we know whether they stopped working during that quarter. Although information for the non-employed who left or lost their job in the last two years is available, we only considered those who left or lost their job in the previous year, in order to be more accurate about their sectoral affiliation. Therefore, we have a pool of individuals who became employed during a surveyed quarter in a specific sector and region or continue to be non-employed over the last 4 quarters (but for which the sector of economic activity of the last employment and current region of residence is known). The constructed labour market outcomes of interest are quarterly series on job finding, job separation, employment, and labor force participation rates for each sector-region cell (1,144 cells in total). For the job finding rate, the numerator is the number of individuals who started working during the current quarter and the denominator is the number of individuals who were not employed over the last 4 quarters. For the job separation rate, we first defined a pool of individuals who either became non-employed during the current quarter in a specific sector and region or continued to be employed; this is the denominator. The numerator is the number of individuals who stopped working in the current quarter and the denominator is the number of employed individuals in that quarter. For the employment rate, the number of working individuals in the current quarter was divided by the sum of the currently working individuals plus the number of non-employed who stopped working the last 4 quarters. For the labor force participation rate, the sum of those working in the current quarter plus those currently unemployed (i.e. those who lost their job during the last 4 quarters) was divided by the sum of those currently working individuals plus the unemployed and inactive individuals who lost their job during the last 4 quarters. All series are weighted by the LFS population weights.

Figure 5 displays how labor market outcomes trended for treated and untreated cells before and during the early pandemic period. Regarding the job finding rate, our treatment assignment fully captures the seasonal character of the Greek labor market according to which new hires in treated cells peak during the second quarter each year before the pandemic (2015–2019). For untreated cells, however, within year fluctuations were limited. In 2020Q2, the job finding rate for treated cells did not increase as in previous years, and it remained at the levels of precedent pre-pandemic quarters. Seasonality is also evident in the case of job separation. Treated and untreated cells trended similarly before the pandemic onset. However, separation rates in treated and untreated cells were not different after the pandemic onset. Regarding the employment rate, we observe strong seasonal patterns in the pre-pandemic year for the treated cells while this is not the case for the untreated ones. In particular, during the pre-pandemic period, the employment rate, on average, was around 70% and it increased notably during the second and third quarters of the year, while it decreased in the first and fourth quarters. Combing the evidence for the job finding and job separation rates, it seems that in tourism-related cells, within-year fluctuations in hires drove the seasonal employment patterns more relative to fluctuations in job separations. In the untreated cells, the employment rate did not fluctuate within the year during the pre-pandemic period. It should be noted that employment rates for both treated and untreated cells were trending slightly upwards since 2015, reflecting the mild recovery in the Greek economy in the years after the 2008–2009 financial crisis. However, in 2020Q2, the employment rate for tourism-related cells reversed its pre-pandemic trend and dropped to 60%, which is equal in magnitude to that of untreated cells. Moreover, in 2020Q2, employment rate in untreated cells did not considerably change from its pre-pandemic trend. Regarding labor force participation rate in treated cells (2020Q1-2020Q2), it decreased both relative to its pre-pandemic trend, and relative to how it evolved in untreated cells.

Mean differences in the labor market outcomes by treatment group and period are shown in Table 1. Given the pandemic onset in February and the quarterly frequency of the LFS data, the first two quarters of 2019 are the pre-pandemic period, and the first two quarters of 2020 are the pandemic period. Regarding the second quarter each year, the job finding rate for the treated (untreated) cells was 22.4% (6.5%) in 2019 and it dropped to 3.9% (3.2%) in 2020. These within-group differences correspond to a 3.3 p.p. reduction for the untreated cells (column 3) and 18.5 p.p. reduction for the treated ones (column 6). The between-groups difference in the above two within-group differences (column 7) is a 15.3 p.p. reduction for the treated cells over and above the developments that took place over time in the untreated cells. The job separation rates did not change significantly for either group although job separation for treated cells was higher in both periods. Employment rate for the untreated did not change significantly before and during the first pandemic wave, fluctuating around 65%-67%. However, for the treated cells, this figure dropped from 75.9% in 2019 to 64.5% in 2020. The between-groups difference in the above two within-group differences (column 7) results in a reduction of 8.9 p.p. in the treated units over and above the developments that took place across time in the untreated units. The labor force participation rate decreased for both groups during the first pandemic wave but mostly for the tourism-related cells; the net change in column 7 represents a reduction of 5 p.p. Repeating the same exercise for the first quarters of 2019 and 2020 does not reveal any significant differences between treated and untreated cells in the two periods. However, the results regarding the first quarters should be viewed with caution because 2020Q1 includes a few weeks after the pandemic onset.

We also calculated averages for the composition of each sector-region cell in 2019Q2 (pre-pandemic period) regarding age, gender, country of birth, marital status, education, and occupation. Table 2 presents the pre-pandemic composition of the LFS sample, in terms of demographic, human capital, occupation, sector, and location characteristics. In the treated group, 70% of respondents were affiliated to the accommodation and food services sectors. Moreover, respondents in treated cells were slightly younger, not married, less educated, employed in sales and clerical jobs, and located in tourism-dependent regions.

Our primary units of analysis are sectors in each region observed in quarters before and after the pandemic onset. As described in Sect. 3, sectors were classified into treated and untreated ones based on the grouping reported in Table 7 (Appendix), and sectoral composition varies by region. Hence, we rely on a DiD framework to compare labor market outcomes between sector-region cells that should be more affected, relative to those less affected, by the economic slowdown due to the lockdown. We expect that employment-related outcomes in treated cells should have exhibited more substantial adjustments compared to untreated cells. Furthermore, we also constructed an alternative treatment definition based on the economic activity that was suspended by governmental orders (also shown in Table A.3). This should provide some evidence on whether labor market responses differed between suspended and not-suspended cells. For our DiD estimates to be valid, two assumptions should be satisfied, i.e. treatment exogeneity and parallel trends. Given that the mobility restrictions across regions that imposed by the government were homogeneously implemented throughout the country, i.e. without taking into account the structure of the local economy, our treatment indicator is considered to be exogenous (discussed also in Sect. 3). Regarding the parallel trends assumption, Fig. 5 reassures that it is satisfied. Labor market outcomes for treated and untreated cells trended similarly over time before the pandemic onset. Based on the above, our main hypothesis is that following the initial virus spread, any mitigation measures to protect public health should have had more adverse short-run labour market impacts in treated sector-region cells. Due the strongly seasonal character of the Greek labor market, our estimation sample uses data from the second quarters of 2019 and 2020. Our linear DiD regression model with cell-specific fixed effects is the following:where, \({\Upsilon}_{srt}\) is the labor market outcome of interest for each sector \(s\) – region \(r\) pair at time \(t\), \({P}_{t}\) takes the value of 1 for 2020 and 0 for the pre- pandemic year (2019), \({T}_{sr}\) is a dummy indicator denoting whether a specific sector \(s\) in each region \(r\), is considered as treated, i.e. it is a tourism-related sector, as defined in Appendix Table 7, \(\left({{P}_{t}\times T}_{sr}\right)\) is an interaction term equal to 1 for the treated group in the second quarter of 2020, and \({u}_{srt}\) is the unobserved error term. The parameter of interest \(\delta \) indicates the short-term labor market effects of the pandemic on the treated cells. To test whether our DiD results are driven by the suspensions imposed by the government in specific cells after the pandemic onset, we augment Eq. (1) by an interaction term between a dummy indicator that denotes whether economic activity within a cell was suspended (as defined in Table 7. in the Appendix) and the post-pandemic dummy indicator \({P}_{t}\). Furthermore, we estimate Eq. (1) for various sub-samples based on characteristics such as age, gender, country of birth, marital status, and education to examine how the short-term labour market impact of the pandemic varied across groups.

Table 3 presents our baseline results based on Eq. (1). Job finding rate was 15.3 p.p. lower in treated sector-region cells during the second quarter of 2020, compared to the untreated ones (column 1). Column 2 adds an interaction term between the pandemic period and a suspended sector binary indicator. This was done to examine whether the observed decline was due to the pandemic or the subsequent governmental order to suspend economic activity in certain sectors in order to prevent the virus spread. The results remain unchanged implying that the negative impact was solely driven by the decline in activity in treated sector-region cells rather than the lockdown per se. Job separation rate (columns 3–4) was not statistically different between the treatment and control group during the pandemic. We attribute this finding to the early government intervention that prohibited layoffs in suspended sectors during the early pandemic phase.

However, despite the fact that separations did not increase, the observed decline in finding rates translated into lower employment. The employment rate was 8.9 p.p. lower (column 5) for the treatment group during the early pandemic. Including the interaction term between the pandemic period and the suspended sectors, the decline in the employment rate of the treatment group in the second quarter of 2020 was 7.8 p.p. lower compared to the employment rate of the untreated (column 6). Regarding the labor force participation rate (columns 7–8), the estimated DiD parameter is negative, but not statistically different from zero. It should be noted that a negative estimated coefficient of the post dummy indicator is also reported in all cases but is statistically significant only in the job finding and employment equations, around 3.3 p.p. and 2 p.p., respectively. These baseline results indicate that the COVID-19 pandemic and lockdown did have a negative impact on labor market outcomes in sectors and regions that largely depend on tourism-related activities.¹⁶ Although the government suspended layoffs in certain sectors that were critical for controlling the spread of within the community, the overall impact on employment rates was driven by a drop in employment entry in treated sector-region cells. Changes in the job separation rate during the pandemic did not seem to determine the observed drop in the employment rate. Lastly, job seeking activity, as captured by the job finding, employment, and labor force participation rates, slowed down in both treated and untreated cells.

Table 4 presents the DiD estimates regarding the heterogeneous impact of the COVID-19 pandemic on labor market outcomes across groups of individuals. For all demographic sub-groups considered here, job finding rates were significantly lower for the treated group during the early pandemic. However, those who were younger, males, native-born, singles, and better educated were the most severely affected groups. It is interesting to notice that the estimated negative impact on job finding rates steadily declines with age. This likely reflects the fact that new hires in tourism-related sectors in the pre-pandemic period were mostly younger individuals, consistent with evidence shown in Table 2. The impact of the pandemic and lockdown on job separation rate was not statistically different from zero for any demographic sub-group. Regarding reductions in employment rates, these were more severe for males, native born, and those with completed secondary and tertiary education. With respect to age, employment rates were lower for individuals aged 30–44 years old. Labor force participation rates were particularly reduced for males in the treated group, a finding that implies a decline in the job seeking behavior of males during the early months of the pandemic. Moreover, all the reported DiD parameters were robust to the inclusion of an additional interaction term between the early pandemic period and the sectors that were suspended by the government.

The COVID-19 pandemic induced severe labor market disruptions worldwide. Following its initial outbreak, governments implemented a series of policies to support firms and their employees, especially those in precarious labor market environments. However, policies and their outcomes varied strongly, depending on the characteristics of each labor market and country-specific idiosyncrasies, e.g. institutions, fiscal and healthcare system capacity. For instance, job retention was not prioritized in Canada where unemployment rose from 6 to 14% and weekly work hours declined by 32% in the early pandemic (Béland et al. 2022; Lemieux et al. 2020). To buffer labor market impacts, furloughing schemes were introduced in the US and the UK, however, 20% and 17% of workers lost their jobs by early April 2020, respectively (Adams-Prassl et al. 2020). In Germany, the eligibility criteria for short-time work became less stringent and the percentage of workers who lost their jobs was considerably lower, i.e. around 5% (Adams-Prassl et al 2020; Mayhew and Anand 2020).

Our paper adds to the literature considering Greece, a country that focused on protecting existing jobs, following a gradual recovery from a prolonged recession and given that a large part of its economy is exposed to seasonal demand for services that involve tasks that cannot be done remotely. Several sectors of economic activity were suspended to restrain the spread of the virus. However, layoffs were prohibited and firms in those sectors could receive financial support conditional on preserving their pre-pandemic headcount. Another distinctive feature of our case is that the national lockdown was imposed during a period when the seasonal, heavily reliant on tourism-related activities (especially in certain regions) economy would normally be gearing up with increased hiring.

To investigate the short-run labor market impacts of the pandemic, we used Labor Force Survey data (2015–2020). We constructed a sector-region panel with information on several labor market outcomes, and we defined cells that were mostly exposed to tourism-related activities and cells that were not affected by tourism-related activities. Under a difference-in-differences framework, we compared the evolution of labor market outcomes for treated and untreated cells before and after the pandemic onset. Our results show that the employment rate in the tourist-affected group fell by approximately 9 p.p. during the first pandemic wave relative to the unaffected group. This reduction was solely due to reduced job finding rates; separation rates did not increase in treated cells due to the layoff prohibitions implemented by the government. Our findings highlight the relevance of policy-making in determining how labor markets adjust to external shocks. For example, Greece (along with some other European countries) emphasized job-retention measures to mitigate the consequences of the pandemic. This resulted in low job separation rates in the early months. Although unemployment modestly increased, it was not because of layoffs. This stands in contrast to countries like the US, the UK and Canada, where unemployment rose quickly as policies emphasized income support more than job protection.

Our findings have important policy implications regarding the nature of labor market disruptions as responses to purely exogenous shocks. The drop in consumer demand due to generalized concerns about the pandemic affected the labor market, however, in the short-run it did not fuel the pool of unemployed with layoffs as in previous crises, e.g. the 2008–2009 one, due to the adopted job retention schemes. Hence, any adverse employment outcomes should be viewed as temporary only, given that the pandemic onset coincided with the time when hires in specific sectors normally peak. Indeed, the gradual re-opening of the economy showed that it was only after the first quarter of 2021 when vacancies increased back to their pre-pandemic level, and unemployment returned to a declining trend. Therefore, the seasonal pattern was not simply postponed, in the sense that hiring did not catch up later on in 2020.

See Figs. 6, 7, 8, 9 and 10 and Tables 5, 6 and 7.