Policy Research Working Paper 10587 Labor Market Impact of the COVID-19 Pandemic in the West Bank and Gaza Jingyuan Deng Nelly Elmallakh Luca Flabbi Roberta Gatti Middle East and North Africa Region Office of the Chief Economist October 2023 Policy Research Working Paper 10587 Abstract This paper studies the impact of the COVID-19 pandemic margin of employment, as working hours declined. The on men’s labor market outcomes in the West Bank and changes in aggregate labor market indicators seem to be Gaza, examining adjustments at the extensive (participa- driven by an increase in job loss and a decline in job gain tion) and intensive (hours of work) margins of the labor in the West Bank and Gaza. Despite the apparent resil- supply. Quarterly panel data from national labor force sur- ience of the labor market, as labor market indicators quickly veys allow observing labor market transitions, job loss and bounced back to their pre-pandemic levels, the results show job gain rates, and labor market stocks. The findings show that the most vulnerable segments of the workforce, such that the COVID-19 pandemic was associated with a decline as informal workers, workers in blue collar occupations, in employment and labor market participation among men the least educated, and residents in refugee camps, bore a in the immediate aftermath of the pandemic. Moreover, the disproportionately heavier burden. analysis finds evidence of large adjustments at the intensive This paper is a product of the Office of the Chief Economist, Middle East and North Africa Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at jingyuan.deng@economics.ox.ac.uk, nelmallakh@worldbank.org, lflabbi@email.unc.edu, rgatti@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Labor Market Impact of the COVID-19 Pandemic in the West Bank and Gaza∗ Jingyuan Deng, Nelly Elmallakh, Luca Flabbi, and Roberta Gatti Keywords: labor market dynamics, COVID-19 pandemic, West Bank and Gaza. JEL Codes: J21, J46, O17, O53. ∗ Deng: Department of Economics, University of Oxford (jingyuan.deng@economics.ox.ac.uk); El- mallakh: World Bank’s Office of the Chief Economist for the Middle East and North Africa (nelmal- lakh@worldbank.org); Flabbi: University of North Carolina at Chapel Hill (lflabbi@email.unc.edu) and IZA; Gatti: World Bank’s Office of the Chief Economist for the Middle East and North Africa (rgatti@worldbank.org). The authors wish to thank Samira Hillis and Gianluca Mele for useful dis- cussions. 1 Introduction The COVID-19 pandemic was a global economic crisis as much as a public health one. The pandemic itself and the measures necessary to combat it wrought severe disruptions to labor markets everywhere in the world. However, even as the pandemic has already ended, its toll is still not yet well-understood in some parts of the world. Evidence from the Middle East and North Africa (MENA) region is conspicuously lacking. And while there is some evidence on the impact of the pandemic on labor market stocks in MENA, the pandemic’s effect on labor market transitions is not well- documented. This paper aims to shed light on the extent to which the pandemic affected the labor market in the West Bank and Gaza, analyzing both labor market stocks and flows, while also exploring the specific segments of the workforce that were more adversely affected. Within the under-studied MENA region, the West Bank and Gaza deserves special attention. The country’s labor market is effectively divided into two segments, with the West Bank interacting closely with Israel, while Gaza remains under continuous eco- nomic isolation and constriction. In the former, almost a quarter of workers commute to Israeli-controlled territories for work, creating a highly idiosyncratic labor market phenomenon. In both, a sizable minority resides under the dire conditions of refugee camps. Gaza, in particular, has 14% of its population in refugee camps. The effec- tive division between the West Bank and Gaza itself led to significant within-country differences in pandemic progression, and created scenarios where national policy may significantly mismatch local conditions. These country-specific conditions allow for a rich and nuanced examination of the pandemic’s effects on labor markets. We first document large effects of the pandemic on both labor market stocks and flows, in both the West Bank and Gaza. We estimate that the pandemic induced an immediate 5 percentage point loss in employment rate, compared with 2018 and 2019 among men aged 20-59. The loss is accounted for by a corresponding rise in non- participation in the two territories, while total unemployment as a share of population changed little. Our results also showcase large intensive margin adjustments. In the immediate aftermath of the COVID-19 shock, weekly hours worked among men aged 20-59 declined by 15 hours in 2020Q2 in the West Bank before quickly bouncing back, compared with a smaller but more persistent decline of 5 hours in Gaza. The rate of job loss rose by over 5 percentage points in the West Bank compared to pre-pandemic levels, in 2018 and 2019. A smaller increase is observed in Gaza, though again persisting for three quarters. The rate of job gain in the West Bank fell by just under 5 percentage points in 2020Q2 before recovering and then overshooting. Meanwhile, in Gaza, job gain dropped slightly below normal in 2020Q2 and 2020Q3, and later plunged by 10 percentage points below normal in 2020Q4. Second, we find both labor markets to be highly resilient despite repeated waves 2 of the pandemic, especially in the West Bank. By 2020Q3, all aggregate labor market indicators in the West Bank had returned to their pre-pandemic levels (2018 and 2019). By 2020Q4, the job gain rate had risen to 5 percentage points above normal in the West Bank. Gaza faced a more prolonged struggle with the pandemic, as its major wave only hit in 2020Q4. Yet just one quarter later in 2021Q1, all aggregate indicators in Gaza had also returned to normal, with the job loss rate falling below normal. However, important heterogeneities underlay the aggregate narrative. Both in the West Bank and in Gaza, job loss is almost entirely accounted for by private sector workers, informal workers, and those employed in blue collar occupations. Next to no effect is found among public sector, formal sector, and white collar workers. Cross- border commuters experienced higher job losses during the disease-ridden quarters of Q2 and Q4 in 2020. Both in the West Bank and in Gaza, job loss among refugee camp residents increased significantly compared to others, especially in the West Bank. They also recovered much more slowly, suffering larger declines in job gain due to the pandemic on top of having lower job gain rates pre-pandemic. In the West Bank, the better educated, specifically those with a secondary degree, almost always fared better. However, in Gaza, the pandemic’s effect on job loss was as large for them as for their less-educated peers during the major shock of 2020Q4. Additionally, in Gaza, the effect on job gain was almost identical for the two groups throughout the periods of analysis under consideration. The importance of heterogeneous effects in the West Bank and Gaza is illustrative of common effects found in labor markets with large informal sectors and significant cross-border commuting. First, job separation can be rapid and sudden, but so can be finding a job post-pandemic. A similar result is found, for example, by Viollaz, Salazar- Saenz, Flabbi, Bustelo, and Bosch (2023) on women working in the informal sector of some major Latin American countries. Second, disadvantaged groups suffered more during the pandemic, in particular those lacking good Internet access to sustain remote work. A large number of contributions came to a similar conclusion, including Peluffo and Viollaz (2021) in their study of Mexican households and Wahby and Assaad (2023) in their study of Syrian refugees in Jordan. Finally, labor markets where cross-border commuting and migration are significant are disproportionally affected. While not all studies in this area come to this conclusion, many do, including Borjas and Cassidy (2020) on data from the US and Barker et al. (2020) on data from Bangladesh. This paper contributes to the broader literature on the impact of the COVID-19 pandemic on labor markets. While there is a considerable body of literature on the impact of COVID-19 on labor markets in developed economies, the evidence on devel- oping economies in general, and in MENA economies in particular, remains sparse. In the developed economies’ context, Baek, McCrory, Messer, and Mui (2021) find that, in the United States, there was a 1.9% increase in a state’s weekly initial unem- ployment insurance claims for every week of stay-at-home exposure relative to other 3 states. Using a highly granular dataset of economic activities in the United States, Chetty, Friedman, Stepner, and Team (2020) find that the closure of small businesses involving in-person contact was a major driver of job loss. Alon, Coskun, Doepke, Koll, and Tertilt (2022) provide evidence from 26 European countries, the United States, and Canada and find disproportionate effects on women, a result supported also by Albanesi and Kim (2021). Adams-Prassl, Boneva, Golin, and Rauh (2020), using data from the UK, the US and Germany, report disproportionate effects also on workers in alternative work arrangements and on the less educated. In the developing economies’ context, Bottan, Hoffmann, and Vera-Cossio (2020) found large effects of the pandemic on job loss and business closures in 17 countries in Latin America and the Caribbean (LAC). With more systematic evidence on labor market stocks and flows in LAC, Viollaz et al. (2023) found a larger impact on women in the region, driven by childcare responsibilities. Providing evidence from Africa, Alon, Doepke, Manysheva, and Tertilt (2022) also highlight that, similar to high- income nations, mothers with school-age children also faced the largest decreases in employment due to the pandemic. These findings were also echoed in the context of other developing countries such as India (Allard et al., 2022) and Zimbabwe (Mabugu, Maisonnave, Henseler, Chitiga-Mabugu, & Makochekanwa, 2023). In the MENA economies’ context, the evidence is considerably sparcer. Early work on the impact of the COVID-19 pandemic on labor markets in the MENA region re- lied on high-frequency phone surveys and highlight important job losses among wage workers and an uneven impact across industries (Krafft, Assaad, & Marouani, 2021, 2022). Providing evidence from Labor Force Surveys in the Islamic Republic of Iran, Dang and Salehi-Isfahani (2023) find that the pandemic exacerbated the pre-existing low participation of females in the labor force. Wahby and Assaad (2023), on the other hand, focus on the impact of the pandemic on Syrian refugees in Jordan and find a divergence in job finding and separation rates of Syrian refugees relative to their Jorda- nian hosts after the onset of the pandemic. Focusing on cross-border commuters in the West Bank and Gaza, Adnan and Etkes (2022) find that undocumented commuters benefited relative to their documented peers after the pandemic, as Israeli policies inadvertently created incentives for employers to favor the former. This sharply con- trasts the results by Borjas and Cassidy (2020) on the impact of the pandemic on immigrants in the United States. The rest of this paper is organized as follows. Section 2 provides background information on labor markets in the West Bank and Gaza, as well as background in- formation on the COVID-19 pandemic and government responses. Section 3 describes the data. Section 4 discusses our methodology. Section 5 presents the main regression results and investigates heterogeneous effects. Section 6 provides robustness checks. Finally, we provide concluding remarks in Section 7. 4 2 Background Information 2.1 Characteristics of West Bank and Gaza’s labor markets The labor markets of the West Bank and Gaza exhibit features typical of the broader Middle East and North Africa (MENA) region, but also have attributes that are highly unique. Additionally, important differences exist between the West Bank and Gaza. This section provides an overview of these characteristics. We use data from the Labor Force Surveys (LFS) of the West Bank and Gaza and we focus on 20-59 years old men. In Section 3.1, we provide more information about the data sources and sample selection. We divide each labor market into five mutually exclusive and jointly exhaustive states: public sector employment, private formal sector employment, private informal sector employment, unemployment, and out of labor force. 1 We focus our discussion exclusively on men, as women’s labor force participation in both the West Bank and Gaza is very low, never reaching values above 25%. This low participation rate is common in MENA countries and makes the role of the pandemic on women’s labor market outcomes relatively less important than other, more relevant structural factors. Figure 1: Labor market stock in West Bank and Gaza over time, by quarter Notes : The figure shows distribution of labor market states among men aged 20 to 59, over the periods of analysis. The five labor market states are mutually exclusive and jointly exhaustive. Informality is defined as either having no work contract, or having only a verbal contract and no health insurance. We include the self-employed in the private informal sector, and employers in the private formal sector. The latter category is numerically immaterial. 1 Data sources will be discussed in greater detail in Section 3.1. 5 Clear differences between the West Bank and Gaza can already be seen in Figure (1), and more shall be discussed in the succeeding paragraphs. Among men, non- participation runs twice as high in Gaza than in the West Bank. Persistently high unemployment characterizes Gaza, with just under a third of these prime working-age men in this state. Non-employment is also more persistent in Gaza than in the West Bank, as Figure (3b) will show. The public sector is much larger in Gaza than in the West Bank. In 2019Q4, 37% of employment in Gaza, among the demographic in question, was in the public sector, compared with 14% in the West Bank. Figure (1) shows that the two labor markets are characterized by very large informal sectors, bigger than the respective formal sectors in both the West Bank and Gaza. This large informal sector is more characteristic of MENA than comparable economies in other regions. For example, in Jordan, there’s a slightly higher number of men aged 20 to 59 working in the informal sector compared to those in the formal sector.2 Likewise, in the Arab Republic of Egypt, among the same demographic group, the informal sector is significantly larger than the formal (Deng, Elmallakh, Flabbi, & Gatti, 2022). However, of the four Latin American and Caribbean countries surveyed in Viollaz et al. (2023), only the Dominican Republic exhibits this pattern. Among men aged 25 to 55, Chile’s formal sector employs more than thrice as many workers as does the informal sector, Mexico’s more than twice as many, while Brazil’s half as many. The large share of informal employment makes the West Bank and Gaza particularly vulnerable. As highlighted in the literature, informal workers bear a disproportionate brunt of the pandemic’s impact (Viollaz et al., 2023). A number of features uniquely set the West Bank and Gaza apart from almost all other economies. Chief among them is the presence of a large minority of workers who commute to Israel and the occupied territories for work, but reside in the West Bank.3 This group usually accounts for between a fifth to a quarter of all employed workers from the West Bank. None however can be found in Gaza, which had been under a de facto state of siege throughout and long before the period under consideration. De jure, the commuters must enter and work in Israel and occupied territories either with valid work permits, or with Israeli or Jerusalem IDs. However, permit-holding status is neither sufficient for, nor necessary to, working in Israel. As of 2019Q4, just under a fifth of West Bank residents had the right to work in Israel and the occupied territories, but almost a quarter among them were not commuting across the border for work. The vast majority of such individuals are holders of Israeli or Jerusalem IDs. Conversely, among those who do commute to Israel and occupied territories, 17% do not hold valid permits or IDs. Likewise, permit-holding status does not logically affect the formality status of the commuter. The frequent border crossings between the West 2 This is derived from authors’ own calculations using the 2016 Jordan Labor Market Panel Survey. 3 It is worth noting that the LFS is representative of the residents of the West Bank and Gaza, whose work may not lie in the country. 6 Bank and Israel and occupied territories strongly suggest that commuters would be differently affected during the pandemic when border closures were enacted. Another unique feature is the presence of a large refugee population in the West Bank and Gaza. However, it is important to note that refugees in this context are de- fined quite differently from other contexts. Here, not only the individuals immediately displaced are considered refugees, but also their patrilineal descendants, even if born many decades later. In particular, the LFS dataset follows the United Nations Relief and Works Agency (UNRWA) definition of refugees, which is “persons whose normal place of residence was Palestine during the period 1 June 1946 to 15 May 1948, and who lost both home and means of livelihood as a result of the 1948 conflict,” as well as “the descendants of Palestine refugee males, including adopted children” (UNRWA, 2023). Consequently, most refugees are indistinguishable in socio-economic outcomes and labor market behavior from non-refugees. However, residence in refugee camps does make a significant difference. As of 2019Q4, 5% of the West Bank’s residents live in refugee camps, as do 14% in Gaza. These camps are run by UNRWA, which effectively acts as a government in the pro- vision of public goods and services. 4 The agency defines a refugee camp as “a plot of land placed at the disposal of UNRWA by the host government to accommodate Palestine refugees and set up facilities to cater to their needs” (UNRWA, 2023). The conditions in the refugee camps are poor. UNRWA (2023) describes them as follows, “socioeconomic conditions in the camps are generally poor, with high population den- sity, cramped living conditions, and inadequate basic infrastructure such as roads and sewers.” Camp residents lack many basic public services, including primary healthcare or education beyond primary school. Further, refugee camps lack internet access, fore- stalling any possibility of remote work during the pandemic. In Wahby and Assaad (2023)’s study of Syrian refugees in Jordan during the pandemic, they found refugee camp residency to be primarily responsible for the refugees’ worse outcomes during the pandemic, as camp residents faced restrictions on movement and were less able to compete with other refugees. 2.2 COVID-19 pandemic and government responses in the West Bank, Gaza, and Israel This section provides background information on the pandemic and government re- sponses during our period of analysis. Figure (2) plots the daily Stringency Index, compiled by the Blatvatnik School of Government, with signposts highlighting events of potential interest. 4 In addition to the official refugee camps, all of which are UNRWA-run, there are four unofficial camps in the West Bank not run by the agency. The LFS defines refugee camps as only those run by UNRWA. See: https://www.pcbs.gov.ps/Downloads/book2622.pdf. 7 Figure 2: Timeline of COVID response measures in West Bank, Gaza, and Israel Notes : The two lines plot the Stringency Index compiled by the Blatavnik School of Government. The index records as behavior. See BSG’s site for a more detailed the strictness of “lockdown style” policies that primarily restrict peopleˆ description. The index is measured at daily intervals and at the national level, hence not distinguishing between West Bank and Gaza. Yellow text boxes describe relevant measures undertaken by Israel, medium green ones by West Bank and Gaza jointly, light green ones by West Bank, and dark green ones by Gaza. The COVID-19 pandemic affected all three territories in question, but over sub- stantively different time horizons. Each territory’s response also differed significantly in scope and intensity. Israel recorded its first identified case of the coronavirus dis- ease COVID-19 in late February 2020, and the West Bank shortly afterwards in early March. Gaza did not record any cases of COVID-19 outside quarantine centers until late August. Nevertheless, beginning in early March, governments in all three terri- tories swiftly imposed measures that gradually increased in stringency. Schools were closed in both the West Bank and Gaza on March 6. In mid-March, Egypt and Israel closed their respective borders with Gaza. By March 22, the West Bank imposed a comprehensive lockdown of 14 days, exempting only the purchase of daily necessities (OCHA, 2020). In the meantime, Israel progressively limited the movement of workers commuting from the West Bank. This brought the status of the commuters to the forefront, resulting in an agreement between Israel and the West Bank’s Palestinian Authority whereby commuters who wish to continue their current work must stay in Israel and occupied territories, but Israeli employers must provide accommodations for them (OCHA, 2020). This arrangement however increased the cost of employing commuters for the Israeli employers, with the associated labor demand implications. It also led to an adverse incentive to favor undocumented commuters over documented ones (Adnan & Etkes, 2022). The West Bank began to gradually ease its first lockdown on May 25, but consis- 8 tently kept social-distancing measures in place. However on July 3, rising caseloads in the West Bank led to a reinstatement of lockdown measures for two weeks. Weekend lockdowns only ended in mid-August (ILO, 2021). On August 24, COVID-19 finally began communal transmission in the isolated Gaza. This led to a lockdown starting August 25 and lasting throughout September. In November, after Gaza had eased its restrictions, the West Bank tightened its own, imposing lockdowns over weekends and curfews over weekdays. Such restrictions became normalized over the winter. In February 2021, the West Bank and Gaza began receiving vaccines from Israel, the Russian Federation, and the WHO COVAX initiative. Nevertheless, towards the very end of 2021Q1, the West Bank experienced a major surge in cases and fatalities, which lasted until April. In Israel and the occupied territories, the course of the pandemic is punctuated by three major lockdowns in our period of analysis. Israel imposed its first almost contemporaneously with the West Bank and Gaza, but began easing the restrictions a month earlier in late April. The period of reopening lasted well into late September, when a resurgence of case loads led to Israel’s second lockdown between September 25 and October 17. After a short respite in November, Israel imposed a third lockdown from December to February the next year. Gradual easing continued throughout March. During this lockdown, vaccination was rolled out, including to commuters with valid permits. Given this context, we expect the main shock to occur in 2020Q2 in the West Bank, with negative effects of smaller magnitudes to persist over the next quarters. In Gaza, we expect another major shock in 2020Q3 that carries over to Q4. We expect the commuters’ outcomes to instead track events in Israel and occupied territories more closely, with a major shock in 2020Q2, recovery in Q3, and another dip in Q4. 3 Data 3.1 Data sources and definitions We employ data from the Labor Force Surveys (LFS) of the West Bank and Gaza, collected by the Palestinian Central Bureau of Statistics and prepared by the Economic Research Forum. The surveys are conducted on a quarterly basis, covering periods from 2000 onwards. The LFS is meant to represent all households whose ordinary residence is in the West Bank and Gaza, though their place of work need not be. The LFS is representative at the region level (respectively of the West Bank and Gaza), as well as at the level of locality types (urban, rural, and refugee camps) (Palestine - Labor Force Survey, LFS , 2021). Importantly for the purpose of our analysis, the data have a panel dimension, en- abling the study of labor market transitions. The sample rotation scheme is described 9 as follows (Palestine - Labor Force Survey, LFS , 2021), Each round of the Labor Force Survey covers all the 536 master sample areas. The areas remain fixed over time, but households in 50% of the EAs are replaced each round. The same household remains in the sample over 2 consecutive rounds (Q1, Q2), skips the next two rounds and is represented again in the sample for the next two consecutive rounds (Q5, Q6) before it is dropped from the sample. A 50% overlap is then achieved between both consecutive rounds and between consecutive years to make the sample efficient for monitoring purposes. The sample rotation therefore allows us to observe each individual four times in four quarters, over the same seasons in two consecutive years. This allows us to construct two types of panels for each individual. First, we can construct a “short panel” from the two consecutive quarters, allowing us to observe quarterly labor market transitions of the same individual. Second, since we observe two sets of consecutive quarters, we could also construct a “panel of short panels”, or a panel of labor market transitions of the same individual over the same seasons in two different years. However, there are two limitations that prevent us from using this full set of observations. First, each “panel of short panels” necessarily contains only one cohort of sampled individuals, who all enter the survey in the same quarter and are then subject to attrition when re-sampled in the subsequent quarters. This introduces attrition bias, and also concerns over representativeness, as the LFS is designed to be representative for each cross-section but not necessarily for each cohort. Second, in the dataset, we observe that the aforementioned ideal rotation scheme has been significantly altered since the onset of COVID-19, sometimes enabling panel observations over longer periods, but usually disrupting panels that we would otherwise expect to observe. The irregularities prevent a uniform application of the “panel of short panels”, even though it would be more conceptually interesting (see Appendix A.1 for issues in the rotation). The paper will instead focus on the quarter-on-quarter “short panels”, and only use the longer panels as robustness checks when they are available. We restrict the sample to the quarters between 2018Q1 and 2021Q1. The upper limit of the timeline is again limited by the availability of the “short panels”, as the Q3/Q4 transition is the only transition we can observe in 2021. The choice of the lower limit hinges on our research design. As our counterfactual for the COVID-19 treated labor market is the period before COVID-19, we would ideally prefer a set of observations with the same characteristics as the treatment period but without the treatment itself. However, selecting a window that is too close to the onset of COVID- 19 could result in emphasizing quarter-level idiosyncrasies and yielding a small sample size. Consequently, we commence our sample from 2018Q1, slightly over two years 10 prior to the pandemic, a period characterized by less distinct time trends in the labor market. Since the focus of the paper is on the impact of COVID on labor market stock and dynamics, we also restrict the sample of interest to include only prime-aged working adults (aged 20 to 59). The dataset contains standard variables expected of a labor force survey, including those denoting employment, unemployment, and inactivity. It also contains information on the intensive margin of the labor supply, including hours worked and full-time and part-time status. Information on employment sector, industry, contract status, health insurance coverage, and mode of work (distinguishing between employees and self-employed, for example) is also available, allowing us to construct indicators of formality and to differentiate different modes of employment. Information on occupation is also available, but only at the level of 2-digit ISCO-08 classification. This information is enough to distinguish between white- and blue-collar occupations but it is not enough to observe additional relevant pandemic-related job characteristics such as the degree of contact with the public. 3.2 Descriptive statistics Figure (1) tracks the evolution of labor market stocks in the West Bank and Gaza respectively over time. A significant increase in non-participation among men can be clearly seen in the West Bank in 2020Q2, mostly at the expense of the private infor- mal sector. The labor market then quickly bounces back. By 2020Q3, labor market stocks in the West Bank appear indistinguishable from pre-pandemic periods. Gaza, on the other hand, experienced three consecutive quarters of depressed employment from 2020Q2 to 2020Q4. Non-participation spiked twice, first in 2020Q2 and then in 2020Q4, corresponding respectively to the initial lockdown orders and the subsequent outbreak in Gaza. Recovery also appears to be slow and uneven. Figure (3) shows the labor market flows. We exploit the specific panel structure of the LFS dataset, described in Section 3.1 by focusing on one cohort of the same respon- dents who were surveyed in 2019Q1, 2019Q2, 2020Q1, 2020Q2, and finally 2020Q4. This cohort of individuals allows us to observe labor market transitions into the pan- demic; to compare with a period over the same quarters in 2019; and, finally, to observe their recovery outcomes in 2020Q4. Overall, Figure (3) shows two labor markets with high levels of churning. On average, 29% of individuals in the sample would change their labor market states after just one quarter. These churns are especially prominent between informal employment and unemployment, and in Gaza between unemployment and non-participation. The figure also illustrates the significant differences between the West Bank and Gaza in labor market dynamics, differences already observed in the labor market stocks presented in Figure (1). In addition to the significant flows between unemployment 11 and non-participation, Gaza also exhibits few flows occurring between the private formal and informal sectors. The unemployment state is far more persistent. Figure 3: Labor market transitions in West Bank and Gaza over time, by quarter (a) West Bank (b) Gaza Notes : The figure shows labor market transitions of the same individuals whose labor market states are observed consistently in all five quarters in question. The sample is restricted to men aged 20 to 59. The figure shows 704 unique individuals in West Bank and 537 in Gaza. The five labor market states are mutually exclusive and jointly exhaustive. Informality is defined as either having no work contract, or having only a verbal contract and no health insurance. In the West Bank, the informal employment is by far the most numerous labor market state. Before the pandemic, flows between the informal employment on the one hand and private formal employment and unemployment on the other are fre- quent, but the sizes of the three states are generally invariant. Public employment and non-participation rarely interact with other states, and are comparatively small by themselves. This pattern very clearly changed in 2020Q2, when COVID-19 hit 12 the West Bank. The public sector is virtually unscathed, while exchanges between the private formal and informal sectors continue. However, large numbers of informal workers are driven into unemployment and non-participation. Far fewer unemployed gain (informal) jobs; many drop out of labor force. Hence the stock effect of the pandemic is most visible in the rise of non-participation. The effect of the pandemic therefore appears to be dominated by job loss. The recovery in 2020Q4 is a straight- forward reversal of the above, with individuals moving back to informal employment from non-employment. Figure (3b) paints a very different picture for Gaza. Pre-pandemic labor market dynamics are dominated by the vast churns in and out of informal employment and unemployment. Entrance into the pandemic in 2020 Q1 to Q2 appears very similar to previous transitions, with only slight upticks in job loss, which is mostly accounted for by informal workers like in the West Bank. However, much fewer are gaining jobs than before. The flows from unemployment to informal employment are cut by half. Hence we should expect the pandemic to affect Gaza primarily through job gain. Gaza was in the midst of its own major outbreak in Q4, and hence we do not observe a recovery. Instead, respective exchanges between informal employment, unemployment, and non-participation stabilized to keep the stocks broadly similar to Q2. 4 Methodology In order to identify the impact of the COVID-19 pandemic on labor market dynam- ics, we compare the proportions of workers in different labor market states and the transitions of those workers between those states before and after the beginning of the pandemic. Since the advent of the pandemic was such a significant and sudden shock to the economy, it is unlikely that the usual confounding factors affecting the identification – such as anticipation effects or other concurring shocks to the economy – play an important role. Moreover, the descriptive statistics show a period of relative stability before the pandemic, providing a control group free from major trends or contamination from other events. A final identification concern is that the impact of the pandemic shock is persistent over time and that perhaps our period of observation is not long enough to assess it. This is a valid concern since we can observe individuals only up to 2021Q1, less than a year from the beginning of the pandemic. We then propose our estimates as short-term effects of the pandemic, leaving to future work the study of long-term effects. 13 4.1 Stock regressions The first specification we propose is the standard regression estimated by now in many countries across the world.5 An indicator variable for a given labor market state, for example employment, is regressed on a series of dummies denoting the pandemic periods. Possible heterogeneous effects are captured by interaction terms. To our knowledge, this is the first paper replicating this specification on data from the West Bank and Gaza. Formally, we estimate the following model: Yiyq = α + βyq Dyq + γq + x′iyq τ + ϵiyq (1) yq ∈T where Yiyq is the outcome of interest for individual i, in quarter q of year y . For example Yiyq = 1 if i is employed in yq . Dyq define the treatment since it denotes a set of indicators equal 1 for each different quarter potentially affected by the pandemic. The set of these quarters, as discussed in Sections 2 and 3, runs from 2020Q2 to 2021Q1 and therefore T = {2020Q2, 2020Q3, 2020Q4, 2021Q1}. For example, D2020Q2 = 1 for all individuals observed in the second quarter of 2020. Since we compare different quarters of different years and we want to account for seasonality, we add quarters fixed effects γq . We also add individual-level controls, which we denote with xiyq . The coefficient of interest is βyq , estimating the difference in the dependent variable between a quarter-year combination affected by the pandemic and all the quarter-year combinations not affected by the pandemic, conditional on seasonal variations and individual-level controls. 4.2 Flow regressions The second specification we propose identifies the impact of the pandemic on labor market dynamics. We define as dependent variable the transition between labor market states for a given worker. We focus on the indicator job loss, which records when an individual leaves the state of employment; and job gains, which records when an individual gains the state of employment. To our knowledge, no previous contribution has estimated such models on MENA countries, in part because these models richer data set to be implemented. For the same reason, only a handful of contributions have done so for other areas of the world.6 Formally, we estimate the following model: Fiyq = α + βyq Dyq + γq + x′iyq−1 τ + µiyq (2) yq ∈T 5 For high-income countries, see for example Alon, Coskun, et al. (2022) and Fairlie, Couch, and Xu (2021); for low- and middle-income countries, see for example Viollaz et al. (2023); in the MENA regional context, see Dang and Salehi-Isfahani (2023). 6 See Albanesi and Kim (2021) for high income countries and Viollaz et al. (2023) for a sample of middle-income countries. 14 where Fiyq measures the quarterly job loss status of individual i between periods yq − 1 and yq . The variable is defined as follows. If the individual was employed in period yq − 1 but no longer in yq , then his job loss status Fiyq is 1. If his employment persisted through the quarterly transition, then Fiyq = 0. If he was not employed in the initial period yq − 1, then his job loss status in yq is undefined. Job gain is defined conversely. If an individual was not employed in period yq − 1, but is employed in period yq , then his job gain status in the current period is 1. If he was not employed in period yq − 1 and still is not in yq , then his job gain status is 0. If he is already employed in period yq − 1, then his job gain status is undefined. 5 Estimation Results 5.1 Stock results Turning to the empirical results, first, we examine the effect of the COVID-19 pandemic on labor market stocks such as inactivity, employment, unemployment, and hours of work. This allows us to quantify the impact of the pandemic on both the extensive and intensive margins of labor market adjustments. Figure (4) reports the results from Equation (1). Our results show that labor markets in the West Bank and Gaza adjusted to the COVID-19 shock through both margins. In the first quarter following the onset of the pandemic (2020Q2), we observe a large increase in the probability of exiting the labor force. In 2020Q2, the pandemic was associated with 8 and 7 percentage points increase in the probability of dropping out of the labor force for men in the West Bank and Gaza, respectively. In the West Bank, in 2020Q2, the decline in male labor force participation seems to coincide with both a decline in employment (6 percentage points) and a decline in unemployment (2 percentage points). Male unemployment, on the other hand, increased by 3 percentage points in 2020Q3. Large intensive margin adjustments in the aftermath of the COVID-19 pandemic are also taking place through a reduction in weekly hours worked. Indeed, men worked on average 15 hours less per week in the immediate aftermath of the crisis (2020Q2) in the West Bank versus 5 hours less per week in Gaza. This reduction in weekly working hours seems to persist up to 2020Q4, although the largest declines are documented in the immediate aftermath of the shock. 15 Figure 4: Effect of the pandemic on labor market stocks Notes : The figure reports coefficients on period dummies and associated 95% confidence intervals from Table (A1) in Appendix A2, where further information can be found. The labor market state outcomes - OLF, Unemployed, Employed - are defined as 1 if the individual is in said state and 0 if in any other state, and thus these coefficients are interpreted as changes in percentage points of each labor market state as a share of the total sample. Hours worked refer to the total hours worked in the past week. The sample for this outcome is restricted to the employed. The estimation sample is restricted to men aged 20-59. The control group is all observations between 2018Q1 to 2020Q1. The regression controls for seasonality fixed effects and the basic demographic characteristics of age and marital status. The employment declines documented in Figure (4) may mask differential effects among segments of the labor force. To have a better understanding of the underlying 16 dynamics, in Figure (5), we investigate the effect of the pandemic on labor market stocks distinguishing between the various types of employment. This exercise reveals some interesting patterns. First, we find large declines in informality, unconditional on employment. In the first quarter after the onset of the pandemic (2020Q2), we document 8 and 5 percentage point declines in informality in the West Bank and Gaza, respectively. Relative to pre-pandemic mean of 0.470 and 0.228, these declines correspond to a 17% and a 22% decline in informality in the West Bank and Gaza, respectively. This decline in informality persists up until 2021Q1 in the West Bank and 2020Q4 in Gaza. As we will show in the following sections, the decline in informality reflects larger job separation rates among informal workers. On the other hand, formal workers are found to be more sheltered from the COVID-19 labor market disruptions. These dynamics extend to other labor market segments such as private sector work- ers, workers in blue-collar occupations, and commuters. Indeed, our results suggest that these groups were disproportionately more affected by the pandemic relative to public sector workers, those employed in white-collar occupations, and non-commuters. For instance, we find a 6 percentage point decline in private sector employment, uncon- ditional on employment, in each of the West Bank and Gaza in 2020Q2. Employment in blue-collar occupations equally decreased by approximately 7 percentage points in 2020Q2 in the West Bank and in Gaza, while the probability of commuting declined by 4 percentage points in the West Bank in the first quarter after the pandemic. It is also worth noting that the labor market in the West Bank seems to bounce back faster than in Gaza. 17 Figure 5: Effect of the pandemic on labor market stocks, by employment type Notes : The figure reports coefficients on period dummies and associated 95% confidence intervals from Tables (A2) and (A3) in Appendix A2, where further information can be found. Each labor market state outcome is defined as 1 if the individual is in said state and 0 if in any other state, and thus these coefficients are interpreted as changes in percentage points of each labor market state as a share of the total sample. The estimation sample is restricted to men aged 20-59. The control group is all observations between 2018Q1 to 2020Q1. The regression controls for seasonality fixed effects and the basic demographic characteristics of age and marital status. 18 5.2 Job loss The labor market stocks presented in Section 5.1 tend to fluctuate based on changes in the frequency of individuals transitioning into unemployment or into inactivity (job loss rates) and out of unemployment or out of inactivity (job finding rates). During the pandemic, these transitions may be influenced by several factors, including a decrease in job creation resulting in fewer available positions, an increase in job destruction resulting in the elimination of existing positions, and a decline in job churning or reallocation as fewer workers decide to voluntarily leave their jobs in search of better opportunities. Figure (6) reports the results from Equation (2) with a focus on job separation rates in the West Bank and in Gaza, separately. In the immediate aftermath of the pandemic (2020Q2), job separation rates among men increased by 5 and 4 percentage points in the West Bank and Gaza, respectively. The recovery in the West Bank was once again faster than in Gaza as job separation rates remained relatively higher in Gaza up until 2020Q4. The increase in job loss in the aftermath of the pandemic is very large when compared with the pre-pandemic means. Relative to a job loss rate of 0.090 and 0.181 in the West Bank and Gaza, respectively, the increase in job loss is estimated to be 60% in the West Bank and 20% in Gaza in 2020Q2. In 2020Q3, we already observe a decline in job loss in the West Bank, while Gaza still suffered from an increase in job loss rate up until 2020Q4. Figure 6: Benchmark impact of the pandemic on job loss Notes : The figure reports coefficients on period dummies and associated 95% confidence intervals from Table (A4) in Appendix A2, where further information can be found. If an individual is employed in the previous period, but no longer in the current, then his job loss status in the current period is 1. If he is employed in previous period and still is in the current, then his job loss status is 0. If he is not employed in the previous period, then his job loss status is undefined. The estimation sample is restricted to men aged 20-59. The control group is all job loss transitions between 2018Q2 and 2020Q1. The regression controls for seasonality fixed effects and the basic demographic characteristics of age and marital status. Figure (7) further corroborates the findings on the unequal toll of the pandemic on different segments of the workforce. Indeed, job losses among men in the West Bank and Gaza were much more pronounced among the most vulnerable groups such 19 as informal workers, private sector workers, those employed in blue-collar occupations, and West Bank commuters. For instance, informal male workers were much more likely to lose their jobs relative to formal workers. In 2020Q2, informal workers were 7 percentage points and 9 percentage points more likely to lose their jobs in the West Bank and Gaza, respectively, compared to the pre-pandemic period. Formal workers in the West Bank were only 2 percentage points more likely to lose their jobs in 2020Q2, while formal workers in Gaza were 2 percentage points less likely to lose their jobs compared to the pre-pandemic period. Likewise, private sector workers were 6 percentage points and 7 percentage points more likely to lose their jobs in 2020Q2 in the West Bank and Gaza, respectively, while their public sector counterparts were 3 percentage points less likely to lose their jobs in the West Bank and only 2 percentage points more likely to lose their jobs in Gaza. Workers in blue-collar occupations bore a disproportionately heavier burden in both the West Bank and Gaza, as reflected by higher job loss rates. While both commuters and non-commuters from the West Bank witnessed an increase in job loss rates in 2020Q2 after the onset of the pandemic, commuters witnessed an even larger penalty. 20 Figure 7: Heterogeneous impact of the pandemic on job loss, by employment type Notes : The figure reports coefficients on period dummies and associated 95% confidence intervals from Tables (A5) and (A6) in Appendix A2, where further information can be found. We run separate regressions for subsamples of men whose initial states are the respective heterogeneous employment types. If an individual is employed in the previous period, but no longer in the current, then his job loss status in the current period is 1. If he is employed in previous period and still is in the current, then his job loss status is 0. If he is not employed in the previous period, then his job loss status is undefined. The estimation sample is restricted to men aged 20-59. The control group is all job loss transitions between 2018Q2 and 2020Q1. The regression controls for seasonality fixed 21 of age and marital status. effects and the basic demographic characteristics Figure (8) further documents the higher job loss rates among the less educated and those residing in refugee camps. In the West Bank, the less educated men (those with less than secondary education) witnessed an increase in the likelihood of losing their job by 9 percentage points in 2020Q2, while the most educated ones (those with above secondary education) only witnessed a 2 percentage point increase in the probability of losing their job in 2020Q2. Similarly, in Gaza, the least educated men witnessed an increase in the job loss rate by 7 percentage points in 2020Q2 versus only a 2 percentage point increase among the most educated individuals. Residents in refugee camps in the West Bank and Gaza experienced higher rates of job loss relative to those residing outside of refugee camps. Nonetheless, it is worth noting that both groups witnessed an increase in job loss rate in the immediate aftermath of the pandemic. Figure 8: Heterogeneous impact of the pandemic on job loss, by demographic characteristics Notes : The figure reports coefficients on period dummies and associated 95% confidence intervals from Tables (A5) and (A6) in Appendix A2, where further information can be found. We run separate regressions for subsamples of men who have the respective demographic characteristics. If an individual is employed in the previous period, but no longer in the current, then his job loss status in the current period is 1. If he is employed in previous period and still is in the current, then his job loss status is 0. If he is not employed in the previous period, then his job loss status is undefined. The estimation sample is restricted to men aged 20-59. The control group is all job loss transitions between 2018Q2 and 2020Q1. The regression controls for seasonality fixed effects and the basic demographic characteristics of age and marital status. 22 5.3 Job gain In this section, we further explore the impact of the pandemic on job finding rates. We find a decline in job gain among men in 2020Q2 by 5 and 4 percentage points in the West Bank and Gaza, respectively. The decline in job gain rates in 2020Q2 is similar in magnitude to the increase in job separation rates, as outlined in Section 52. As shown in the Figure (9), the recovery in the West Bank was faster than in Gaza. Job gain in the West Bank quickly recovered and actually increased by 2020Q4, while job gain in Gaza was still significantly lower (8 percentage points) in 2020Q4 compared to the pre-pandemic period. Figure 9: Benchmark impact of the pandemic on job gain Notes : The figure reports coefficients on period dummies and associated 95% confidence intervals from Table (A7) in Appendix A2, where further information can be found. If an individual is not employed in the previous period, but is employed in the current, then his job gain status in the current period is 1. If he is not employed in previous period and still is not in the current, then his job gain status is 0. If he is employed in the previous period, then his job gain status is undefined. The estimation sample is restricted to men aged 20-59. The control group is all job gain transitions between 2018Q2 and 2020Q1. The regression controls for seasonality fixed effects and the basic demographic characteristics of age and marital status. While it is possible to investigate the heterogeneous effects of the pandemic on job loss rates by demographic characteristics and employment types (informal versus formal employment, public versus private sector employment, white-collar versus blue- collar occupations, and commuters versus non-commuters), we can only investigate the heterogeneous effects of the pandemic on job gain by demographic characteristics since, by definition, individuals who gain jobs are either unemployed or out of the labor force at baseline. Therefore, we cannot explore the heterogeneity of the effects with respect to individuals’ employment types at baseline. The heterogeneous effects on job gain by demographic characteristics are reported in the Figure (10). In 2020Q2, men residing in refugee camps in the West Bank do not seem to witness any effect on job gain compared to the pre-pandemic period. On the other hand, those residing outside of refugee camps in the West Bank have a 6 percentage points lower likelihood of job gain in 2020Q2 compared to the pre- pandemic. Similarly, non-residents in refugee camps in Gaza have a 4 percentage 23 points lower likelihood of job gain in 2020Q2 relative to a 5 percentage points lower likelihood among those residing in refugee camps. During the recovery phase, those residing in refugee camps seem to fare worse compared to those residing outside refugee camps. The educational attainment of men also appears to have varying effects on job gain in the West Bank and Gaza. In Gaza, we find that both the less educated and the most educated individuals were equally affected in terms of the decline in job gain in the first three quarters after the onset of the pandemic and both equally recovered by 2021Q1. In the West Bank, the story is different. The less educated individuals are found to have a lower probability of job gain (1 percentage point) in 2020Q2 but then their job gain rate quickly bounced back in 2020Q3 and 2020Q4. On the other hand, the most educated individuals in the West Bank, who seem not to be initially affected in terms of their job gain rate in 2020Q2, witnessed a small decline in job gain in 2020Q3 and then increases in their job finding rates thereafter. 24 Figure 10: Heterogeneous impact of the pandemic on job gain, by demographic characteristics Notes : The figure reports coefficients on period dummies and associated 95% confidence intervals from Tables (A8) and (A9) in Appendix A2, where further information can be found. We run separate regressions for subsamples of men who have the respective demographic characteristics. If an individual is not employed in the previous period, but is employed in the current, then his job gain status in the current period is 1. If he is not employed in previous period and still is not in the current, then his job gain status is 0. If he is employed in the previous period, then his job gain status is undefined. The estimation sample is restricted to men aged 20-59. The control group is all job gain transitions between 2018Q2 and 2020Q1. The regression controls for seasonality fixed effects and the basic demographic characteristics of age and marital status. 6 Robustness Tests 6.1 Alternative specification This section reports a number of robustness checks. While the flow regression results presented in Section 5.2 and Section 5.3 rely on the panel dimension of the data in order to track changes in job loss and job gain for individuals observed over two consecutive quarters, this section exploits the panel dimension of the data in order to track the same individuals observed over 4 quarters (two quarters in the pre-pandemic period in 2019 and two quarters in the post-pandemic period in 2020). We are particularly interested in transitions between Q1 and Q2 of 2020 relative to Q1 and Q2 of 2019 in columns (1) and (3), as well as transitions between Q2 and Q3 of 2020 relative to transitions between Q2 and Q3 of 2019 in columns (2) and (4). 25 This allows us to compare individual transitions over the same quarters before and after the onset of the pandemic. To do so, we estimate Equation (3) below. Yiy = α + βDy + x′iy τ + ϵiy (3) Our dependent variable Yiy refers to individual transitions over two consecutive quarters. For example, for the Q1/Q2 job loss transitions, if an individual is employed in the previous period, but no longer is in the current, then the job loss status in the current period is 1. If the individual is employed in the previous period and still is in the current, then his job loss status is 0. If he is not employed in the previous period, then his job loss status is undefined. These transitions are defined in a similar way for job gain. x′iy refers to our vector of basic demographic characteristics, which includes age and marital status. The main coefficient of interest in Equation (3) is β , which captures the effect of the pandemic on individual transitions in the post-pandemic period relative to the baseline. The results are reported in Table (9) on job loss and Table (10) on job gain. In line with our benchmark results presented in the previous section, we consistently find evidence suggesting higher job loss rates between Q1 and Q2 in the West Bank and Gaza. Tracking the same individuals from before and after the onset of the pandemic, we find a 7 percentage point increase in job loss rate in the West Bank and a 10 percentage point increase in Gaza between Q1 and Q2 of 2020, with respect to the same quarters in 2019. As labor markets started to recover, we do not find any significant difference in job loss transitions between Q2 and Q3 of 2020 relative to the same quarters in 2019. 26 Table 9: Effect of the pandemic on job loss, alternative specification West Bank Gaza (1) (2) (3) (4) Q1/Q2 Q2/Q3 Q1/Q2 Q2/Q3 Pandemic 0.0673*** -0.0231 0.101*** -0.0427 (0.0195) (0.0151) (0.0378) (0.0310) Constant 0.108*** 0.151*** 0.345*** 0.567*** (0.0354) (0.0280) (0.0759) (0.0701) Controls YES YES YES YES Observations 1,202 1,106 488 461 R-squared 0.011 0.026 0.048 0.082 # Ind 678 640 311 282 Mean base Y 0.0980 0.0787 0.174 0.158 *** p<0.01, ** p<0.05, * p<0.1 Standard errors clustered at the period level. Notes : The outcome of each regression is job loss, over periods denoted in the column titles. “Pandemic” is a dummy equalling 1 if the observed transition occurs in 2020, and 0 if in 2019. Controls are the basic demographic characteristics of age and marital status. “# Ind” refers to the number of unique individuals within the estimation sample. “Mean base Y” reports the pre-pandemic mean of the dependent variable within the estimation sample. The sample is demographically restricted to men aged 20 to 59, and constructed from a sample of individuals who are observed in both quarters of the transition, in both 2019 and 2020, for a total of four quarters. We thus construct a panel of quarterly labor market transitions over 2019 and 2020. Note that because the outcome is conditional on the initial labor market state, the panel is in general unbalanced. Hence the control group in the estimation sample is the transitions of approximately the same individuals over the same quarters in 2019. If an individual is employed in the previous period, but no longer in the current, then his job loss status in the current period is 1. If he is employed in previous period and still is in the current, then his job loss status is 0. If he is not employed in the previous period, then his job loss status is undefined. The results on job gain reported in Table (10) also confirm that job gain was mostly affected in the immediate aftermath of the pandemic between Q1 and Q2 before quickly recovering between Q2 and Q3. The decline in job gain between Q1 and Q2 of 2020 is estimated to be 10 percentage points in the West Bank relative to 6 percentage points in Gaza. 27 Table 10: Effect of the pandemic on job gain, alternative specification West Bank Gaza (1) (2) (3) (4) Q1/Q2 Q2/Q3 Q1/Q2 Q2/Q3 Pandemic -0.109** 0.0662 -0.0596** 0.0159 (0.0471) (0.0435) (0.0253) (0.0258) Constant 0.330*** 0.345*** 0.312*** 0.160*** (0.0843) (0.0729) (0.0412) (0.0438) Controls YES YES YES YESs Observations 366 419 715 684 R-squared 0.078 0.138 0.064 0.037 # Ind 261 295 431 406 Mean base Y 0.369 0.286 0.175 0.125 *** p<0.01, ** p<0.05, * p<0.1 Standard errors clustered at the period level. Notes : The outcome of each regression is job gain, over periods denoted in the column titles. “Pandemic” is a dummy equalling 1 if the observed transition occurs in 2020, and 0 if in 2019. Controls are the basic demographic characteristics of age and marital status. “# Ind” refers to the number of unique individuals within the estimation sample. “Mean base Y” reports the pre-pandemic mean of the dependent variable within the estimation sample. The sample is demographically restricted to men aged 20 to 59, and constructed from a sample of individuals who are observed in both quarters of the transition, in both 2019 and 2020, for a total of four quarters. We thus construct a panel of quarterly labor market transitions over 2019 and 2020. Note that because the outcome is conditional on the initial labor market state, the panel is in general unbalanced. Hence the control group in the estimation sample is the transitions of approximately the same individuals over the same quarters in 2019. If an individual is not employed in the previous period, but is employed in the current, then his job gain status in the current period is 1. If he is not employed in previous period and still is not in the current, then his job gain status is 0. If he is employed in the previous period, then his job gain status is undefined. High education refers to having a secondary degree or above. 6.2 Placebo test Finally, we performed a falsification test in order to make sure that we are capturing the effect of the pandemic and not any other sort of spurious correlation. We run these falsification tests both on the stock regressions (equation (1)) and the flow regressions (equation (2)). Our placebo test assumes that the pandemic took place in 2019Q2 instead of 2020Q2. Focusing on a sample of quarterly observations spanning the period be- tween 2018Q2 and 2020Q1, our post-pandemic period therefore refers to the quarters between 2019Q2 and 2020Q1. The placebo stock regressions reported in Figure (11) confirm that there are no significant differences in labor market stocks in the placebo post-pandemic period relative to the baseline. Indeed, if anything, we find a decline in the probability of dropping out of the labor force in 2019Q4 in Gaza and a small decline in hours worked in 2020Q1 in the West Bank. Both of these effects actually go in the opposite direction if we compare these results with the estimates in Figure (4). 28 Figure 11: Placebo effect on labor market stock Notes : The figure shows the output of a placebo test with a set-up analogous to Figure 4. We perform the same regression as specified in Equation (1). Our sample includes data from 2018Q2 to 2020Q1 and assumes that the pandemic started in 2019Q2. Therefore, the post-pandemic period refers to the quarters between 2019Q2 to 2020Q1. The analysis is restricted to men aged 20-59. Furthermore, overall the results reported in Figure (12) on labor market flows consistently show no significant change in the probability of job loss and job gain in 29 the West Bank and Gaza. Out of all the estimated coefficients for each period for both outcomes (job loss and job gain), we only find a negative effect on job gain in 2019Q4, which is very small in magnitude (1 percentage point). Taken altogether, the results presented in this section bolster our confidence that we are correctly identifying the effects of the pandemic shock on labor market outcomes. Figure 12: Placebo effect on labor market flows Notes : The figure shows the output of a placebo test with a set-up analogous to Figures 6 and 9. We perform the same regression as specified in Equation (2). Our sample includes data from 2018Q2 to 2020Q1 and assumes that the pandemic started in 2019Q2. Therefore, the post-pandemic period refers to the quarters between 2019Q2 to 2020Q1. The analysis is restricted to men aged 20-59. 7 Conclusion This paper examines the effect of the pandemic on labor markets in the West Bank and Gaza using quarterly labor market data provided by national labor force surveys. With a focus on men’s labor market outcomes, this paper sheds light on how labor markets in the West Bank and Gaza adjusted to the COVID-19 shock examining adjustments at the extensive (employment) and intensive (hours of work) margins. One of the main contributions of this paper is the use of panel data which allows us to examine the effect of the pandemic on labor market transitions—job loss and job gain rates—in addition to the effect on labor market stocks. Studying both stocks and flows provides a comprehensive framework to analyze the impact of the pandemic on labor markets and allows for a better understanding of the underlying mechanisms 30 behind the changes in labor market stocks. Our findings suggest that labor markets in the West Bank and Gaza adjusted to the COVID-19 shock through changes in the extensive and intensive margins of employ- ment. Indeed, we find that men in both the West Bank and Gaza were significantly more likely to drop out of the labor force after the onset of the pandemic. In addition to a decline in employment, we also find a sizable reduction in working hours. Our flows regressions show by how much fluctuations in labor market aggregates are driven by changes, respectively, in the probability of losing or finding a job. We find an increase in the probability of job loss among workers in the West Bank and Gaza and an equally sized decline in the probability of job gain. While labor markets in the West Bank and Gaza recovered quickly and aggregate labor market indicators reverted back to their pre-pandemic levels in the recovery phase, our results nonetheless underscore highly heterogeneous effects. Informal work- ers, those employed in blue-collar occupations, those residing in refugee camps, and the least educated individuals witnessed significantly larger job losses in the aftermath of the pandemic. These systematic differences highlight the necessity to provide targeted support for the most affected and most vulnerable segments of the workforce. They also confirm results found by the broader literature on the impact of the COVID-19 pandemic. First, economies characterized by a large informal sector experienced much higher job losses during the pandemic, but also relatively high job gains when the pandemic started to taper off. Second, disadvantaged groups lacking internet access were the most affected. And third, migration and cross-border crossing were crucial factors in magnifying the impact of the pandemic in the labor market. 31 References Adams-Prassl, A., Boneva, T., Golin, M., & Rauh, C. (2020, September). Inequality in the impact of the coronavirus shock: Evidence from real time surveys. Jour- nal of Public Economics , 189 , 104245. Retrieved 2023-06-14, from https:// www.sciencedirect.com/science/article/pii/S0047272720301092 doi: 10 .1016/j.jpubeco.2020.104245 Adnan, W., & Etkes, H. (2022). Undocumented Palestinian workers in Israel: Did the Israeli COVID-19 Policy Boost their Employment? Retrieved 2023-01-26, from https://www.inss.org.il/publication/illegal-workers/ Albanesi, S., & Kim, J. (2021, August). Effects of the COVID-19 Recession on the US Labor Market: Occupation, Family, and Gender. Journal of Economic Perspectives , 35 (3), 3–24. Retrieved 2022-09-27, from https://pubs.aeaweb .org/doi/10.1257/jep.35.3.3 doi: 10.1257/jep.35.3.3 Allard, J., Jagnani, M., Neggers, Y., Pande, R., Schaner, S., & Moore, C. T. (2022). Indian female migrants face greater barriers to post- covid recovery than males: Evidence from a panel study. EClinicalMedicine , 53 . Alon, T., Coskun, S., Doepke, M., Koll, D., & Tertilt, M. (2022, May). From Man- cession to Shecession: Women’s Employment in Regular and Pandemic Reces- sions. NBER Macroeconomics Annual , 36 , 83–151. Retrieved 2023-06-10, from https://www.journals.uchicago.edu/doi/abs/10.1086/718660 (Publisher: The University of Chicago Press) doi: 10.1086/718660 Alon, T., Doepke, M., Manysheva, K., & Tertilt, M. (2022). Gendered impacts of covid-19 in developing countries. In Aea papers and proceedings (Vol. 112, pp. 272–276). Baek, C., McCrory, P. B., Messer, T., & Mui, P. (2021, November). Unemployment Effects of Stay-at-Home Orders: Evidence from High-Frequency Claims Data. The Review of Economics and Statistics , 103 (5), 979–993. Retrieved 2023-06- 14, from https://doi.org/10.1162/rest a 00996 doi: 10.1162/rest a 00996 Barker, N., Davis, C. A., L´ opez-Pe textasciitildena, P., Mitchell, H., Mobarak, A. M., Naguib, K., . . . Vernot, C. (2020). Migration and the labour market impacts of covid-19 (WIDER Working Paper No. 2020/139). Borjas, G. J., & Cassidy, H. (2020, May). The Adverse Effect of the COVID-19 Labor Market Shock on Immigrant Employment [Working Paper]. National Bureau of Economic Research. Retrieved 2023-06-14, from https://www.nber.org/ papers/w27243 doi: 10.3386/w27243 Bottan, N., Hoffmann, B., & Vera-Cossio, D. (2020, October). The unequal impact of the coronavirus pandemic: Evidence from seventeen developing countries. PLOS ONE , 15 (10), e0239797. Retrieved 2023-06-14, from https://journals 32 .plos.org/plosone/article?id=10.1371/journal.pone.0239797 (Pub- lisher: Public Library of Science) doi: 10.1371/journal.pone.0239797 Chetty, R., Friedman, J. N., Stepner, M., & Team, T. O. I. (2020, June). The Economic Impacts of COVID-19: Evidence from a New Public Database Built Using Private Sector Data [Working Paper]. National Bureau of Economic Research. Retrieved 2023-06-12, from https://www.nber.org/papers/w27431 doi: 10.3386/w27431 Dang, H.-A., & Salehi-Isfahani, D. (2023). The impact of covid-19 on labor market outcomes in iran. Deng, J., Elmallakh, N., Flabbi, L., & Gatti, R. (2022). Labor Market Transitions in Egypt Post-Arab Spring (World Bank Policy Research Working Paper). Wash- ington, DC: World Bank. Retrieved from https://openknowledge.worldbank .org/handle/10986/37755 doi: 10.1596/1813-9450-10126 Fairlie, R., Couch, K., & Xu, H. (2021). The evolving impacts of the covid-19 pandemic on gender inequality in the u.s. labor market: The covid motherhood penalty. NBER Working Paper , 29426 . ILO. (2021). Impact of the COVID-19 Pandemic on the Labour Market in the Occupied Palestinian Territory: a Forecasting Model Assessment (Publication). Beirut, Lebanon. Retrieved from https://www.ilo.org/beirut/publications/WCMS 774731/lang--en/index.htm Krafft, C., Assaad, R., & Marouani, M. A. (2021). The impact of covid-19 on middle eastern and north african labor markets. In The economic research forum: Policy brief no. 55. Krafft, C., Assaad, R., & Marouani, M. A. (2022). ). the impact of covid-19 on middle eastern and north african labor markets: A focus on micro, small, and medium enterprises. In The economic research forum: Policy brief no. 60. Mabugu, R. E., Maisonnave, H., Henseler, M., Chitiga-Mabugu, M., & Makochekanwa, A. (2023). Implications of covid-19 and mitigation measures on gender and the zimbabwean economy. Economic Modelling , 121 , 106225. OCHA. (2020, March). COVID-19 Emergency Situation Report 1 (as of 1200 hrs, 24 March 2020). Retrieved 2023-05-23, from http://www.ochaopt.org/content/ covid-19-emergency-situation-report-1 Palestine - Labor Force Survey, LFS. (2021). Retrieved 2023-05-20, from http:// www.erfdataportal.com/index.php/catalog/206 Peluffo, C., & Viollaz, M. (2021). Implications of covid-19 and mitigation measures on gender and the zimbabwean economy. Review of Economics of the Household , a351. 19 , 327ˆ UNRWA. (2023). Palestine refugees. Retrieved 2023-05-31, from https://www.unrwa .org/palestine-refugees Viollaz, M., Salazar-Saenz, M., Flabbi, L., Bustelo, M., & Bosch, M. (2023). The 33 COVID-19 Pandemic in Latin American and Caribbean countries: Gender Dif- ferentials in Labor Market Dynamics. IZA Journal of Development and Migra- tion . Wahby, S., & Assaad, R. (2023). Job Finding and Separation among Syrian refugees in Jordan and their hosts during the COVID-19 pandemic [Working Paper]. 34 A Online Appendix A.1 Additional Figures Figure A1 below provides an approximate view of the actual rotation in the dataset. An individual is assigned to a cohort defined by the quarter of his or her first observation. We then track the quarters in which individuals from each cohort are observed again. The first two cohorts follow the planned rotation described above. The cohort entering in 2018Q3 was already treated differently due to COVID-19, being re-interviewed again in 2020Q3. Yet the 2018Q4 cohort received no such treatment, while the 2019Q1 cohort was additionally re-interviewed after a gap of one, rather than two, quarter. The 2019Q2 cohort was helpfully retained for four consecutive quarters between 2020Q2 to 2021Q1, throughout the course of the pandemic. Yet from this cohort onwards, issues for short panel construction arise, as we can only observe at most one transition at regular intervals for all subsequent cohorts. For example, the cohort entering in 2019Q3 lacks a corresponding quarterly transition in 2020, and the cohort entering in 2020Q2 was only observed twice. Figure A1: Sample Rotation of West Bank and Gaza Labor Force Survey, 2018-2021 Notes : An individual is assigned to a cohort defined by the quarter of his or her first observation. The figure shows in which quarters individuals from each cohort are observed. 35 A.2 Regression outputs Table A1: Effect of the pandemic on labor market stocks West Bank Gaza (1) (2) (3) (4) (5) (6) (7) (8) OLF Unemployed Employed Hours worked OLF Unemployed Employed Hours worked 2020Q2 0.0798*** -0.0211** -0.0587*** -15.15*** 0.0674*** 0.00800* -0.0754*** -4.781*** (0.00610) (0.00916) (0.0153) (0.206) (0.000342) (0.00385) (0.00357) (0.0964) 2020Q3 0.0152*** 0.0268** -0.0420*** -1.811*** 0.0577*** 0.00704*** -0.0647*** -4.826*** (0.00295) (0.00944) (0.00649) (0.312) (0.00651) (0.00122) (0.00771) (0.111) 2020Q4 -0.00354 0.00674 -0.00320* -1.085*** 0.105*** -0.0295** -0.0754*** -3.146*** (0.00340) (0.00501) (0.00162) (0.119) (0.0170) (0.0120) (0.00495) (0.203) 2021Q1 -0.00243 0.00138 0.00104 -1.571 0.00168 0.0247*** -0.0264 0.694 (0.00218) (0.00197) (0.00370) (1.025) (0.0125) (0.00684) (0.0182) (0.484) Constant 0.0863*** 0.148*** 0.766*** 45.96*** 0.123*** 0.483*** 0.394*** 36.18*** (0.00548) (0.00842) (0.0102) (0.817) (0.0144) (0.0338) (0.0376) (1.858) Seasonality FE YES YES YES YES YES YES YES YES Controls YES YES YES YES YES YES YES YES Observations 55,609 55,609 55,609 41,826 34,993 34,993 34,993 15,584 R-squared 0.034 0.006 0.036 0.050 0.044 0.023 0.054 0.018 # Ind 21077 21077 21077 17390 12451 12451 12451 6897 Mean base Y 0.134 0.114 0.752 44.50 0.250 0.298 0.452 30.56 *** p<0.01, ** p<0.05, * p<0.1 Standard errors clustered at the period level. Notes : The labor market state outcomes - OLF, Unemployed, Employed - are defined as 1 if the individual is in said state and 0 if in any other state, and thus these samples include all observations within restrictions. Hours worked refer to the total hours worked in the past week. The sample for this outcome is restricted to the employed. Seasonality fixed effects are dummies for the recurring quarters of each year. Controls are the basic demographic characteristics of age and marital status. “# Ind” refers to the number of unique individuals within the estimation sample. “Mean base Y” reports the pre-pandemic mean of the dependent variable within the estimation sample. The sample is demographically restricted to men aged 20 to 59, and chronologically restricted to between 2018Q1 to 2021Q1. Each period dummy is 1 if an observation is made in that period, and 0 otherwise. Hence the control group in the estimation sample is all pre-pandemic periods. 36 Table A2: Effect of the pandemic on labor market stocks by type, West Bank (1) (2) (3) (4) (5) (6) (7) (8) Formal Informal Public Private White collar Blue collar Commuter Non-com. 2020Q2 0.0197 -0.0804*** -0.000962 -0.0598*** 0.00537 -0.0660*** -0.0411*** -0.0193 (0.0139) (0.000734) (0.00238) (0.0121) (0.00309) (0.0115) (0.00130) (0.0157) 2020Q3 0.00558 -0.0441*** -0.00414 -0.0344*** -0.00628*** -0.0323*** -0.0144*** -0.0255*** (0.0181) (0.0121) (0.00317) (0.00289) (0.000379) (0.00612) (0.000439) (0.00624) 2020Q4 0.0317*** -0.0341*** -5.88e-05 -0.00238 0.00328*** -0.00505 0.000749 -0.00337 (0.00496) (0.00140) (0.000985) (0.00259) (0.000809) (0.00437) (0.00122) (0.00443) 2021Q1 0.0319*** -0.0310*** -0.00114 0.00203 -3.90e-05 0.000939 0.0117** -0.0101 (0.00460) (0.000897) (0.00247) (0.00152) (0.00156) (0.00559) (0.00501) (0.00788) Constant 0.197*** 0.527*** 0.0792*** 0.645*** 0.0462*** 0.678*** 0.190*** 0.533*** (0.00942) (0.00728) (0.00686) (0.00601) (0.00774) (0.00913) (0.00762) (0.00976) Seasonality FE YES YES YES YES YES YES YES YES Controls YES YES YES YES YES YES YES YES Observations 51,525 51,525 51,525 51,525 51,525 51,525 51,525 51,525 R-squared 0.022 0.007 0.008 0.019 0.008 0.023 0.018 0.007 # Ind 20273 20273 20273 20273 20273 20273 20273 20273 Mean Base Y 0.282 0.470 0.115 0.636 0.114 0.638 0.169 0.583 *** p<0.01, ** p<0.05, * p<0.1 Standard errors clustered at the period level. Notes : Each labor market state outcome is defined as 1 if the individual is in said state and 0 if in any other state, and thus these samples include all observations within restrictions, not conditional on employment. Seasonality fixed effects are dummies for the recurring quarters of each year. Controls are the basic demographic characteristics of age and marital status. “# Ind” refers to the number of unique individuals within the estimation sample. “Mean base Y” reports the pre-pandemic mean of the dependent variable within the estimation sample. The sample is demographically restricted to men aged 20 to 59, and chronologically restricted to between 2018Q1 to 2021Q1. Each period dummy is 1 if an observation is made in that period, and 0 otherwise. Hence the control group in the estimation sample is all pre-pandemic periods. A worker is a commuter if he is employed in Israel and its settlements. Informality is defined as either having no work contract, or having only a verbal contract and no health insurance. A worker is defined as white-collar if he is a senior official or senior manager, a professional, or a technician or associate professional. Table A3: Effect of the pandemic on labor market stocks by type, Gaza (1) (2) (3) (4) (5) (6) Formal Informal Public Private White collar Blue collar 2020Q2 -0.0211*** -0.0515*** -0.0124*** -0.0602*** -0.00773* -0.0649*** (0.00255) (0.00160) (0.00157) (0.000655) (0.00423) (0.00520) 2020Q3 -0.0115 -0.0522*** 0.000117 -0.0638*** -0.00902** -0.0548*** (0.00704) (0.000552) (0.00877) (0.00134) (0.00345) (0.00424) 2020Q4 -0.0251*** -0.0552*** -0.0144*** -0.0659*** -0.0134*** -0.0665*** (0.000448) (0.00551) (0.000408) (0.00477) (0.00240) (0.00273) 2021Q1 -0.0308*** 0.00588 -0.0178*** -0.00712 -0.00883** -0.0161 (0.00598) (0.0106) (0.00572) (0.00925) (0.00404) (0.0163) Constant 0.0443 0.319*** 0.0405 0.323*** 0.0144 0.349*** (0.0341) (0.00878) (0.0293) (0.00957) (0.0116) (0.0232) Seasonality FE YES YES YES YES YES YES Controls YES YES YES YES YES YES Observations 32,065 32,065 32,065 32,065 32,065 32,065 R-squared 0.062 0.018 0.047 0.019 0.019 0.039 # Ind 11992 11992 11992 11992 11992 11992 Mean Base Y 0.223 0.228 0.174 0.278 0.102 0.349 *** p<0.01, ** p<0.05, * p<0.1 Standard errors clustered at the period level. Notes : Each labor market state outcome is defined as 1 if the individual is in said state and 0 if in any other state, and thus these samples include all observations within restrictions, not conditional on employment. Seasonality fixed effects are dummies for the recurring quarters of each year. Controls are the basic demographic characteristics of age and marital status. “# Ind” refers to the number of unique individuals within the estimation sample. “Mean base Y” reports the pre-pandemic mean of the dependent variable within the estimation sample. The sample is demographically restricted to men aged 20 to 59, and chronologically restricted to between 2018Q1 to 2021Q1. Each period dummy is 1 if an observation is made in that period, and 0 otherwise. Hence the control group in the estimation sample is all pre-pandemic periods. Informality is defined as either having no work contract, or having only a verbal contract and no health insurance. A worker is defined as white-collar if he is a senior official or senior manager, a professional, or a technician or associate professional. 37 Table A4: Effect of the pandemic on job loss West Bank Gaza 2020Q2 0.0543*** 0.0365*** (0.00475) (0.00931) 2020Q3 -0.0106*** 0.0107*** (0.000841) (0.000798) 2020Q4 0.00172*** 0.0275*** (0.000425) (0.00224) 2021Q1 -0.00511 -0.0495*** (0.0131) (0.00701) Constant 0.146*** 0.461*** (0.0194) (0.0185) Seasonality FE YES YES Controls YES YES Observations 16,669 6,627 R-squared 0.014 0.048 # Ind 12,116 4,917 Mean base Y 0.0900 0.181 *** p<0.01, ** p<0.05, * p<0.1 Standard errors clustered at the period level. Notes : Seasonality fixed effects are dummies for the recurring quarters of each year. Controls are the basic demographic characteristics of age and marital status. “# Ind” refers to the number of unique individuals within the estimation sample. “Mean base Y” reports the pre-pandemic mean of the dependent variable within the estimation sample. The sample is demographically restricted to men aged 20 to 59, and chronologically restricted to between 2018Q2 to 2021Q1. We denote the time of an observation of a transition by the destination period. For example, a 2018 Q1/Q2 transition will be denoted as 2018Q2. Each period dummy is 1 if an observation is made in that period, and 0 otherwise. Hence the control group in the estimation sample is all pre-pandemic periods. If an individual is employed in the previous period, but no longer in the current, then his job loss status in the current period is 1. If he is employed in previous period and still is in the current, then his job loss status is 0. If he is not employed in the previous period, then his job loss status is undefined. 38 Table A5: Heterogeneous impact of the pandemic on job loss, West Bank (1) (2) (3) (4) (5) (6) Refugee camp Commuters Informal Public White collar High education 2020Q2 0.0454*** 0.0510*** 0.0238** 0.0604*** 0.0603*** 0.0855*** (0.00325) (0.00216) (0.0107) (0.00485) (0.00729) (0.00812) 2020Q3 -0.0112*** -0.00242*** 0.0189*** -0.0161*** -0.0125*** -0.0130*** (0.00212) (0.000716) (0.000244) (0.000830) (0.000304) (0.00189) 2020Q4 -0.000350 -0.00398*** 0.00746*** 0.000449 0.00209*** 0.00571*** (0.000544) (0.000973) (0.000627) (0.000549) (0.000259) (0.000350) 2021Q1 -0.00756 0.00503 0.00522 -0.0104 -0.00769 -0.00928 (0.0112) (0.0126) (0.0103) (0.0153) (0.0131) (0.0128) Heterogeneity in level 0.0420** 0.0662*** 0.0977*** -0.108*** -0.0753*** -0.0661*** (0.0188) (0.00229) (0.00428) (0.0154) (0.00138) (0.000430) 2020Q2#Heterogeneity 0.0744*** 0.0339*** 0.0488* -0.0409*** -0.0403* -0.0668*** (0.0190) (0.00611) (0.0231) (0.00707) (0.0218) (0.00989) 2020Q3#Heterogeneity 0.000791 -0.0334*** -0.0479*** 0.0324*** 0.0120* 0.00763*** (0.0163) (0.00671) (0.00335) (0.000336) (0.00665) (0.00190) 2020Q4#Heterogeneity 0.0196*** 0.0325*** -0.00435*** 0.00120 0.00491 -0.00754*** (0.00190) (0.00151) (0.000704) (0.00270) (0.00647) (0.000476) 2021Q1#Heterogeneity 0.0282 -0.0511*** -0.0135** 0.0201 0.00959*** 0.0125*** (0.0189) (0.00216) (0.00460) (0.0159) (0.00128) (0.000386) Constant 0.139*** 0.125*** 0.0670*** 0.154*** 0.145*** 0.170*** (0.0171) (0.0187) (0.0169) (0.0213) (0.0190) (0.0182) Seasonality FE YES YES YES YES YES YES Heterogeneity-specific seasonality YES YES YES YES YES YES Controls YES YES YES YES YES YES Observations 16,669 16,640 16,621 16,647 16,633 16,669 R-squared 0.015 0.024 0.031 0.026 0.021 0.022 # Ind 12116 12094 12083 12103 12089 12116 Heterogeneity mean 0.0966 0.230 0.612 0.155 0.155 0.420 Mean Base Y 0.0886 0.0763 0.0386 0.103 0.100 0.106 *** p<0.01, ** p<0.05, * p<0.1 Standard errors clustered at the period level. Notes : The outcome of each regression is job loss, and the heterogeneity in question is the title of each column. Seasonality fixed effects are dummies for the recurring quarters of each year. Heterogeneity-specific fixed effects are interactions between the heterogeneity dummies and the seasonality dummies. Controls are the basic demographic characteristics of age and marital status. “# Ind” refers to the number of unique individuals within the estimation sample. “Mean base Y” reports the pre-pandemic mean of the dependent variable within the estimation sample. The sample is demographically restricted to men aged 20 to 59, and chronologically restricted to between 2018Q2 to 2021Q1. We denote the time of an observation of a transition by the destination period. For example, a 2018 Q1/Q2 transition will be denoted as 2018Q2. Each period dummy is 1 if an observation is made in that period, and 0 otherwise. Hence the control group in the estimation sample is all pre-pandemic periods. If an individual is employed in the previous period, but no longer in the current, then his job loss status in the current period is 1. If he is employed in previous period and still is in the current, then his job loss status is 0. If he is not employed in the previous period, then his job loss status is undefined. A worker is a commuter if he is employed in Israel and its settlements. Informality is defined as either having no work contract, or having only a verbal contract and no health insurance. A worker is defined as white-collar if he is a senior official or senior manager, a professional, or a technician or associate professional. High education refers to having a secondary degree or above. 39 Table A6: Heterogeneous impact of the pandemic on job loss, Gaza (1) (2) (3) (4) (5) Refugee camp Informal Public White collar High education 2020Q2 0.0323*** -0.0202** 0.0712*** 0.0322** 0.0672*** (0.00648) (0.00847) (0.0144) (0.0115) (0.0179) 2020Q3 0.00291 -0.0295*** 0.0691*** 0.0296*** 0.0491*** (0.00544) (0.000787) (0.00567) (0.00309) (0.00130) 2020Q4 0.0327*** 0.000777 0.0824*** 0.0370*** 0.0334*** (0.00249) (0.00277) (0.00855) (0.00383) (0.00547) 2021Q1 -0.0514*** -0.0448*** -0.0546*** -0.0526*** -0.0451*** (0.00368) (0.00431) (0.00758) (0.0105) (0.00924) Heterogeneity in level -0.00743 0.215*** -0.178*** -0.0979*** -0.0597*** (0.0194) (0.00491) (0.00717) (0.0133) (0.00272) 2020Q2#Heterogeneity 0.0284** 0.107*** -0.0995*** 0.0349** -0.0527*** (0.0111) (0.00638) (0.00481) (0.0118) (0.0136) 2020Q3#Heterogeneity 0.0404 0.114*** -0.104*** -0.0640*** -0.0589*** (0.0351) (0.00563) (0.00534) (0.00247) (0.00217) 2020Q4#Heterogeneity -0.0335*** 0.0835*** -0.0840*** -0.0246*** -0.00660 (0.00257) (0.0147) (0.0155) (0.00240) (0.00589) 2021Q1#Heterogeneity 0.00705 -0.0193*** 0.0109* 0.0120 -0.0126*** (0.0194) (0.000697) (0.00512) (0.0123) (0.00254) Constant 0.460*** 0.185*** 0.420*** 0.447*** 0.471*** (0.0179) (0.0227) (0.0168) (0.0186) (0.0201) Seasonality FE YES YES YES YES YES Heterogeneity-specific seasonality YES YES YES YES YES Controls YES YES YES YES YES Observations 6,627 6,610 6,623 6,612 6,627 R-squared 0.050 0.117 0.104 0.063 0.061 # Ind 4917 4904 4913 4906 4917 Heterogeneity mean 0.164 0.490 0.400 0.231 0.547 Mean base Y 0.178 0.0677 0.261 0.211 0.230 *** p<0.01, ** p<0.05, * p<0.1 Standard errors clustered at the period level. Notes : The outcome of each regression is job loss, and the heterogeneity in question is the title of each column. Seasonality fixed effects are dummies for the recurring quarters of each year. Heterogeneity-specific fixed effects are interactions between the heterogeneity dummies and the seasonality dummies. Controls are the basic demographic characteristics of age and marital status. “# Ind” refers to the number of unique individuals within the estimation sample. “Mean base Y” reports the pre-pandemic mean of the dependent variable within the estimation sample. The sample is demographically restricted to men aged 20 to 59, and chronologically restricted to between 2018Q2 to 2021Q1. We denote the time of an observation of a transition by the destination period. For example, a 2018 Q1/Q2 transition will be denoted as 2018Q2. Each period dummy is 1 if an observation is made in that period, and 0 otherwise. Hence the control group in the estimation sample is all pre-pandemic periods. If an individual is employed in the previous period, but no longer in the current, then his job loss status in the current period is 1. If he is employed in previous period and still is in the current, then his job loss status is 0. If he is not employed in the previous period, then his job loss status is undefined. Informality is defined as either having no work contract, or having only a verbal contract and no health insurance. A worker is defined as white-collar if he is a senior official or senior manager, a professional, or a technician or associate professional. High education refers to having a secondary degree or above. 40 Table A7: Effect of the pandemic on job gain West Bank Gaza 2020Q2 -0.0497** -0.0378*** (0.0180) (0.00376) 2020Q3 0.000125 -0.0209*** (0.000812) (0.00642) 2020Q4 0.0538 -0.0818*** (0.0315) (0.00536) 2021Q1 -0.0209*** 0.00229 (0.00154) (0.00246) Constant 0.404*** 0.208*** (0.0259) (0.00953) Seasonality FE YES YES Controls YES YES Observations 5,542 8,340 R-squared 0.049 0.026 # Ind 4,520 6,089 Mean base Y 0.305 0.156 *** p<0.01, ** p<0.05, * p<0.1 Standard errors clustered at the period level. Notes : Seasonality fixed effects are dummies for the recurring quarters of each year. Controls are the basic demographic characteristics of age and marital status. “# Ind” refers to the number of unique individuals within the estimation sample. “Mean base Y” reports the pre-pandemic mean of the dependent variable within the estimation sample. The sample is demographically restricted to men aged 20 to 59, and chronologically restricted to between 2018Q2 to 2021Q1. We denote the time of an observation of a transition by the destination period. For example, a 2018 Q1/Q2 transition will be denoted as 2018Q2. Each period dummy is 1 if an observation is made in that period, and 0 otherwise. Hence the control group in the estimation sample is all pre-pandemic periods. If an individual is not employed in the previous period, but is employed in the current, then his job gain status in the current period is 1. If he is not employed in previous period and still is not in the current, then his job gain status is 0. If he is employed in the previous period, then his job gain status is undefined. 41 Table A8: Heterogeneous impact of the pandemic in West Bank and Gaza West Bank Gaza (1) (2) (3) (4) Refugee camp High education Refugee camp High education 2020Q2 -0.0577*** -0.0818*** -0.0359*** -0.0359** (0.0176) (0.0261) (0.00220) (0.0130) 2020Q3 -0.00152 0.0390*** -0.0194*** -0.0162** (0.00663) (0.00270) (0.00546) (0.00617) 2020Q4 0.0712 0.0347 -0.0729*** -0.0794*** (0.0405) (0.0262) (0.00805) (0.00760) 2021Q1 -0.0236*** -0.0607*** 0.00510 0.00334 (0.00468) (0.0167) (0.00618) (0.00659) Heterogeneity in level -0.0467** -0.129*** 0.0131 -0.0179* (0.0184) (0.0380) (0.0423) (0.00835) 2020Q2#Heterogeneity 0.0547*** 0.0763*** -0.0121 -0.00117 (0.00356) (0.0184) (0.00905) (0.0365) 2020Q3#Heterogeneity 0.0164 -0.0968*** -0.0101* -0.00601*** (0.0447) (0.00568) (0.00539) (0.00183) 2020Q4#Heterogeneity -0.125 0.0516*** -0.0459*** 0.00251 (0.0724) (0.00748) (0.0145) (0.00578) 2021Q1#Heterogeneity 0.0149 0.0938** -0.0143 -0.00138 (0.0185) (0.0366) (0.0424) (0.00852) Constant 0.413*** 0.494*** 0.205*** 0.221*** (0.0249) (0.0317) (0.0118) (0.0106) Seasonality FE YES YES YES YES Heterogeneity-specific seasonality YES YES YES YES Controls YES YES YES YES Observations 5,542 5,542 8,340 8,340 R-squared 0.050 0.064 0.027 0.029 # Ind 4520 4520 6089 6089 Heterogeneity mean 0.138 0.434 0.189 0.485 Mean base Y 0.310 0.367 0.157 0.177 *** p<0.01, ** p<0.05, * p<0.1 Standard errors clustered at the period level. Notes : The outcome of each regression is job gain, and the heterogeneity in question is the title of each column. Seasonality fixed effects are dummies for the recurring quarters of each year. Heterogeneity-specific fixed effects are interactions between the heterogeneity dummies and the seasonality dummies. Controls are the basic demographic characteristics of age and marital status. “# Ind” refers to the number of unique individuals within the estimation sample. “Mean base Y” reports the pre-pandemic mean of the dependent variable within the estimation sample. The sample is demographically restricted to men aged 20 to 59, and chronologically restricted to between 2018Q2 to 2021Q1. We denote the time of an observation of a transition by the destination period. For example, a 2018 Q1/Q2 transition will be denoted as 2018Q2. Each period dummy is 1 if an observation is made in that period, and 0 otherwise. Hence the control group in the estimation sample is all pre-pandemic periods. If an individual is not employed in the previous period, but is employed in the current, then his job gain status in the current period is 1. If he is not employed in previous period and still is not in the current, then his job gain status is 0. If he is employed in the previous period, then his job gain status is undefined. High education refers to having a secondary degree or above. 42