Competing for Jobs: How COVID-19 Changes Search Behaviour in the Labour Market

1 Introduction

Economies around the world were suffering from the effects of measures to contain the corona virus. Particularly, the labour markets were hit hard. While unemployment had been rising, the stock of vacancies declined. To understand the impact of the Covid-19-crisis on the functioning of the labour market, it is not only important to look at unemployment and the stock of open vacancies, but also at the search behaviour of employees and employers. This has several advantages: On the one hand, it gives a more complete view on the actual state of the labour market, because not just (recently) unemployed people search for jobs but also people that feel threatened by unemployment. On the other hand, looking at applications is a more precise measure than unemployment because it summarizes two dimensions: Not just more workers search but also workers might search more intensively. Relatedly, the paper sheds light on questions such as: How do people on job search react to shifts in the labor market? To what extent did the Covid-19 crisis intensify competition for vacancies? Are those who have lost or were threatened to lose their jobs looking for employment in less affected industries? Do they apply below their seniority level?

Studies analyse the labour market effects of the Covid-19-crisis from various perspectives. For instance, Coibion, Gorodnichenko, and Weber (2020) use a U.S. household survey to document several facts on unemployment and participation. They find that the job loss in the U.S. was massive, but that people who lost their jobs did not search actively. Furthermore, they point out that the participation rate declined much more than in the Great Recession. Campello, Kankanhalli, and Muthukrishnan (2020) rather focus on the firms’ perspective and explore changes in vacancy posting behaviour in the U.S. Their paper shows that the pandemic leads to downskilling because less high skilled positions are posted, and that also hiring is hampered. Bauer and Weber (2020) assess the impact of the pandemic on labour market flows between employment and unemployment for Germany. The results indicate reduced hiring dynamics. The findings of these papers imply that the main challenge in the light of the pandemic was to strengthen the recovery by reinforcing vacancy creation, boosting search activities and thus stimulate hiring.

However, little is known so far about how the Covid-19-crisis changed search processes and behaviour in the labour market. The most related studies concerning such questions are by Hensvik, Le Barbanchon, and Rathelot (2021) for Sweden, Marinescu, Skandalis, and Zhao (2020) for the US and Hutter and Weber (2021) and Hartl, Christian, and Weber (2021) for Germany. While these papers use large online search platforms to evaluate the state of the labor markets, others like Carrillo-Tudela et al. (2023) for the UK or Balgova et al. (2022) for the Netherlands rely on survey data. Hensvik, Le Barbanchon, and Rathelot (2021) find that job search intensity dropped due to the Covid-19-crisis and that job-seekers redirect their search to occupations which are less affected. Marinescu, Skandalis, and Zhao (2020) evaluate a policy intervention that makes unemployment benefits more generous. In the US, the applications to vacancy ratio increased, because applications and job ads decreased unproportionately. Hutter and Weber (2021) use data of the job platform of the German Federal Employment Agency and find that search and placement activities dropped significantly and have not fully recovered yet. The same data is used in Hartl, Christian, and Weber (2021) to evaluate the impact of search activity on job findings over a longer time span. This paper exploits that changes in search intensity have explanatory power in understanding hiring dynamics. In contrast to our contribution, the data used in these papers is very different. Both papers derive search intensity by using click data, but do not have any structural information. The platform of the German Federal Employment Agency is visited to a large extent by people that are unemployed or have difficulties to find a job, while on the LinkedIn career network mainly employed people search. This also is reflected by the type of vacancies posted in these two different platforms. While on the platform of the Federal Employment Agency rather low skilled jobs are posted (see Bossler et al. 2020), the opposite is true for the LinkedIn platform. Carrillo-Tudela et al. (2023) focus on the question who changes jobs and find that “while workers adjust their job search in favour of expanding industries and occupations, a significant proportion target declining industries and occupations throughout the sample period.” (Carrillo-Tudela et al. 2023, p. 11). Balgova et al. (2022) also rely on survey data and show that employed rather than unemployed persons changed their job search behavior in the Covid-19-crises. While they also use application data, they do not show how the direction of search behaviour changed.

Summarized, our study adds insights to the questions above by analysing data from the LinkedIn professional network for Germany. The data provides a comprehensive picture of which people were searching, what they were looking for and what they found.

It allows to analyse how the competition among workers for jobs changed at the onset of the Covid-19-crisis by looking at the overall application behaviour. In that respect, it gives a more comprehensive view on the state of the labour market as it overcomes certain issues classical measure – such as tightness – face (see Abraham, Haltiwanger, and Rendell 2020). In detail we ask whether the trend in applications per job posting changed and whether this is due to the fact that more people searched for jobs or because search intensity was rising.

We also assess whether the Covid-19-crisis might have changed the allocation of matches by having a look at the changes across industry sectors in applications. Such sectoral imbalances can cause rising aggregate unemployment (Lucas and Prescott 1974, Lilien 1982). As explained above, with the data at hand, we cover only a certain non-representative fraction of the overall allocation. Nonetheless, the imbalances can give indication whether we might expect the natural rate of unemployment to rise in the aftermath of the crisis. Indeed, while the labour market as a whole recovered from the crisis, Germany still faces higher long-term unemployment than before the pandemic.

Further we are able to observe whether workers applied for jobs at the same seniority level^[1] as before the crisis or whether the workers are willing to make concessions which might hamper their career progress. That can be linked to the questions whether match quality rises or declines in recessions, i.e. if the recent recession is cleansing or sullying (see, for instance, Foster, Grim, and Haltiwanger (2016)).

The paper is structured as follows. We elaborate on the situation in Germany during the Covid-19 crisis. In Section 3 we present the data used. Section 4 documents the development of labour market competition and sheds light on the driving forces. Section 5 highlights sectoral imbalances in application behaviour, and Section 6 takes a deeper look on intended reallocation processes. Section 7 illustrates possible changes in career progress. Section 8 concludes.

2 The Crisis in Germany

All in all, the Covid-19-crises left a clear mark on the labour market. Politicians in Germany reacted quickly by facilitating access to short-time work in order to keep as many people as possible in employment. In spring 2020, far more short-time work notices were received than during the Great Recession (Gehrke and Weber 2020). In May 2020 almost 6 million people had been on short-time work (see Figure 1), which is 4 times more than in the peak of the Great Recession. Nonetheless, the number of unemployed rose sharply between March and June, but far less than the number of short-time workers. In addition, the number of registered vacancies fell by a good 20 percent (see https://www.arbeitsagentur.de/en/press/en-2020-39-the-labour-market-in-august-2020). The overall stock even dropped by more than 35 percent in the second quarter of 2020 compared to 2019 and barely recovered in the third quarter (see Kubis 2020). The number of new vacancies fell as well-even if these have been recovering somewhat since May 2020. Figure 5 in the Appendix A displays the development of unemployment and vacancies over time.

Figure 1:

Short-time work over time. Source: Federal Employment Agency.

Some sectors are more affected in terms of the labour market reaction than others. Especially Accommodation and food service activities, Wholesale and service activities (e.g. tourism and recreation) were hit hard. Health care also shows a reaction in terms of unemployment inflows and short-time work, but is less affected than the other sectors. Given that, we will analyse the behaviour of applications in particular in those sectors that were most affected, positively or negatively. Figure 6 in the Appendix A provides information on short-time work and unemployment inflows by sector in the observation period.^[2] Short-time work is important for the understanding how sectors were affected by the crises as it gives a more comprehensive view than unemployment. In this paper, we can show that in sectors with high short-time work uptake, the application behaviour changed towards sectors with low uptake.

When we compare the Covid-19 crises to the Great recession, the following stands out. Short-time work was even used more frequently to save jobs than in the Great Recession. However, it was concentrated rather in service sectors during the Covid-19 crises as in manufacturing as during the Great Recession (see Gehrke and Weber 2020). The service sectors are usually low-wage sectors (see Bruttel 2019), had a high risk of contagion and only few possibilities of working from home (Bauer et al. 2021). These particularities gave rise to the question, whether the Covid-19 crises induced a reallocation shift (Barrero et al. 2021) to sectors with more favourable conditions. During the Great Recession, Bauer and King (2018) find no strong reallocation pattern in Germany. However, Garnadt, von Rueden, and Thiel (2021) show that the Covid-19 crises is atypical in terms of reallocation, but do not analyse sector changes. Looking at search behaviour can give indication on the (desired) direction of change. Furthermore, the Covid-19 crises is exceptional in terms of search activities, as there is indication that workers in the US decreased their search effort (Forsythe et al. 2020) while search behaviour typically is found to be countercyclical in the literature (Mukoyama, Patterson, and Şahin 2018). In that respect this paper can contribute by looking at the search behaviour of not just unemployed persons but also people that are employed.

3 Data

We build our analysis on comprehensive data of the career network LinkedIn, a platform that is used for professional networking by workers and firms. While firms are posting jobs, workers seeking for jobs might send applications. After registering on LinkedIn, members create a profile, which is similar to a resume and add information about work experience, competencies and education.

The number of LinkedIn members in the central European region (Germany, Austria, Switzerland) amounts to 15 million. In comparison to the German work force, Manufacturing, Professional, scientific and technical activities and Information and communication ore overrepresented on LinkedIn which reflects the increasing digitalization and the importance of networks in these sectors. LinkedIn members employed in Manufacturing are the largest group and employees in the Health and social services sector as well as in Wholesale and retail trade; repair of motor vehicles and motorcycles sector are underrepresented on LinkedIn. Though important industries are sufficiently represented in the LinkedIn data, the data is not representative for the overall economy. Persons that search on the LinkedIn platform are rather high-skilled, working in relatively large firms and higher-than-median income earners (Brenner, Aksin Sivrikaya, and Schwalbach 2019). From a firm’s perspective, about 75 percent of larger German firms have a LinkedIn profile and usage is quite common (see Bonsón and Bednárová 2013). In that respect, the data cannot be generalized for the overall population or the firm distribution in Germany but nonetheless provides a good impression on search processes. Tables 5 –7 in the Appendix A show further summary statistics of the data. Men’s share is ten percentage points higher than in the total labour force, entry positions – typically filled with younger workers – occur quite often, and the same holds true for white-collar jobs, e.g. in operations, education and business development.

Overall, the data provide an adequate though not representative basis for the analyses in the following sections, and in particular, allow for reasonable industry comparisons. An illustration of this comparison can be found in the Appendix A, Figure 7.

Access to the data is granted via cooperation with LinkedIn Economic Graph. Due to the high confidentiality demands, there is no Scientific Use File available and data are drawn from LinkedIn exclusively for certain research question.

We draw a sample of 3 million members on the 31st of May 2020. We constructed this sample such that a sampled person applies at least once in the observation period to a premium job posting. For our analysis, we rely on the longitudinal dimension of the data and track people back to January 2019.

This implies that the sampled members are either on job search for a longer time span or just joined LinkedIn recently. Then we investigate the behaviour of the members between January 1, 2019 and May 31, 2020 looking at applications. We use the sum of applications to premium job postings across the month and divide it by the stock of premium job postings. This statistic delivers insights on job competition. Furthermore, we take the number of applications over the stock of applicants, which we interpret as measure for search activity. Note that the applications sent via LinkedIn represent only a fraction of job searches within LinkedIn. It might be the case that members search for jobs but use a different mode for application than the application through LinkedIn. However, there is no indication that structural differences in the application behaviour between these groups exist.

Given the way the sample of LinkedIn members is drawn, we cannot distinguish between persons that just signed up to LinkedIn and persons that already searched for a longer time span. Hence, to check on this issue we refine the data even more by extracting people that have a full history from January 2019 onwards. This procedure rules out attrition and excludes possible distortions by new members flowing in over time. This leaves us with 2.43 million members.

For the jobs, we examine premium job postings, which are postings that are paid for by the firm. This rules out duplicates and ensures up-to-dateness. Furthermore, we use applications from members within Germany, i.e. from members who state Germany as their base location and apply for jobs in Germany.

Because the data is available on a daily basis, it allows to precisely determining the periods of interest. The “crisis period” refers to the period between the 23rd of March 2020 and the 31st of May 2020. We chose the 23rd of March as this is the date when the lockdown came into force nationwide and comprehensively in Germany. The “pre-crisis period” refers to the same period in 2019.

Before we go ahead with the sample, we draw comparisons between Statistics of the Federal Employment Agency, the LinkedIn data overall which is labelled as “the universe”, and in particular our sample in the next section. Whenever we go into details on economic sectors, the results rely on the sample outlined above.

4 Labour Market Competition in the Crisis

To understand the overall trends during the Covid-19 crises, we first compare two measures of labour market competition. The first uses the “universe” of applications of all members and all job posting of the LinkedIn professional network,^[3] while the second uses registered unemployed over vacancies of the official statistics of the Federal Employment Agency. The latter is an inverted key statistic in modern macroeconomics called labour market tightness. Originally, it is defined as the ratio between vacancies and unemployed and is the central element in search and matching models (Pissarides 2000) as it reflects the state of the economy. In recessions, labour market tightness relaxes, indicating a slack labour market. In booms, more vacancies are created, and unemployment decreases, which tightens the labour market from a firm’s perspective. As we are interested in the state of the labour market from the workers perspective, we inverted this statistic.

In the left panel of Figure 2, the black line refers to the universe of data from LinkedIn. The line clearly shows that job competition started to rise particularly in March 2020. Comparing this measure to the inverted labour market tightness (grey dashed line), it is obvious that the LinkedIn data shows a much stronger increase in job competition. Later, we will see that the rise in competition in the universe of the LinkedIn data is mainly due to the fact that more persons joined for seeking for jobs, and not due the fact that members search more intensively.

Figure 2:

Number of applications per job posting over time, black line refers to all members, grey line to a random sample of members that remains constant over time. Dotted lines depict unemployment to vacancy ratio. Source: LinkedIn, Federal Employment Agency, own calculation.

Comparing the behaviour of job competition in the sample (solid grey line) to the inverted labour market tightness (dashed grey line) as in the left panel of Figure 2, three issues stand out. First, competition started to increase in both measures in March when Covid-19 hit, though not as strong as in the universe of the LinkedIn data (black line). Second, the unemployment per vacancy ratio shows a stronger increase than the applications per job posting during the crisis than in the LinkedIn sample, see right panel of Figure 2. Apart from that, the trends are similar and the data is highly correlated (0.8). Third, the level is different. The ratio between unemployed and vacancies is slightly higher compared to the applications per job posting in the sample (see right panel of Figure 2).

Labour market tightness (and so the inverted measure too) lacks to give a complete picture for several reasons as Abraham, Haltiwanger, and Rendell (2020) point out. The competition measure of LinkedIn overcomes several issues mentioned therein and hence allows for a more complete view on changes in labour market competition. On the one hand, it provides data for all people who search, independently of whether they are unemployed or not. And, according to Balgova et al. (2022), the search processes of employed workers are a key channel in the Covid-19-crises. A large part of the labour market reaction in Germany is mitigated by short-time work schemes as pointed out in Section 2. In that sense, the unemployed to vacancy measure is incomplete and reflects only partial effects as most of the people searching for jobs might be still employed but for instance in short-time work schemes.^[4] On the other hand, unemployed workers usually send more than just one application. Hence, counting applications instead of unemployed gives a more complete picture as well.

Relatedly, two possibilities might explain the rise in competition: more people searched for a job but less jobs are available, or workers intensified their search by sending more applications.

However, a comparison of the universe of LinkedIn members to a sample of 2.43 million people, that were already members in the beginning of 2019, shows that people so far do not search more intensively.^[5] Figure 2 points out that applications per job posting for sample members (grey solid line) did not rise as much as for the universe of members (black solid line). Figure 3 shows that the sampled members do not search differently than the universe of members, which also includes persons who joined just at the onset of the Covid-19-crises. Though the latter send slightly more applications in general, which might be due to few persons that apply very frequently, the time trend does not differ.

Figure 3:

Applications per applicant (right panel) over time, black line refers to all members, grey line to a random sample of members that remains constant over time. Dotted line in left panel depicts unemployment to vacancy ratio.

Briefly, Figures 2 and 3 indicate that competition for jobs is higher when we consider all job seekers and not just the unemployed, and count applications instead of unemployed. This even amplified during the crisis. But job seekers do not search more intensively but more members joined for seeking for jobs.

We have to keep in mind that applications sent via LinkedIn are less time-consuming than applications via other search channels. Therefore, applicants may apply more often and to different kinds of positions. In comparison to data of the Panel Study Labour Market and Social Security (PASS),^[6] we see that 45 percent of the employed persons and 48 percent of the non-employed persons search for a job via social networks in 2020 (within the last four weeks prior to the interview). A brief comparison to the applications received by firms with the IAB Job Vacancy Survey (see Gürtzgen et al. 2022)^[7] reveals, that the firms received about 12 applications per filled vacancy in 2020. Given the LinkedIn data, where we see on average 6 applications before the crises (when we abstract from the effect that more people joined LinkedIn at the start of the crisis), the share of received applications by firms also lies at about 50 percent. That shows, that social networks are important for job search processes.

5 Sectoral Imbalances

Economic crises can lead to labour market reallocation depending on the distribution of recessionary effects. The theoretical and empirical literature, starting with seminal papers of Lucas and Prescott (1974) and Lilien (1982), show that sectoral shifts may lead to long lasting effects on aggregate unemployment.

To highlight the differences across sectors, we examine industries that are particularly affected by Covid-19, whether positively or negatively. We want to know, whether the sectoral imbalances induced by the Covid-crisis might lead to pressure to change the industry. We interpret an observed application as intention to change. Such changes are often accompanied by a loss in occupation and industry specific human capital (see Kambourov and Manovskii 2009; Neal 1995), which causes wage losses. Sectors that are particularly affected exhibit strong changes in the share of applications coming from and going to these industries.

Table 1 refers to the applications by origin and destination. Applications by origin are counted such that the number of applications is counted from the members that are or have been in a certain sector. Applications by destination are counted such that the sector of the firm who receives the application is the objective. Persons from the Recreation and Tourism sector applied 74.8 % more frequently, but this sector received about 60 % fewer applications. Recreation and Tourism is one of the sectors that suffered most which explains the fact that employees in these sectors apply more frequently. In light of Table 1, pressure has also arisen in the corporate services sector, in manufacturing and in the real estate industry. A certain tension is also apparent in construction, although so far this sector has only been slightly affected by the Covid-19-crises. Furthermore, this pattern is well in line with the inflows into unemployment per sector (see Figure 6 in the Appendix A) between April and May, and the overall change in unemployment across occupations in 2020 (see Figure 8 in the Appendix A). Inflows into unemployment were particularly high from the sectors of Accommodation food service activities and Wholesale and retail. The highest changes in unemployment were apparent in Transport and logistics occupations, Food and hospitality occupations, Trade occupations, Occupations in business management and organisation and Manufacturing occupations.^[8]

The situation is different in the Health Care sector, for example. This sector experienced increased demand due to the pandemic. Members from this sector are not only applying more frequently, this sector also receives more applications. The reason for this could be that people in unfavourable positions (temporary employment, commuting, poor pay, etc.) are looking for an improvement, but also people who work or have worked in severely impaired sectors saw good chances for employment here. However, another reason is the distribution of jobs across sectors. While there are more vacancies in the health care sector due to the increased demand, there are hardly any vacancies left in the recreation & tourism sector.

In order to disentangle these effects, we proceed with a logistic regression. As dependent variable, we use a dummy that equals 1 when a member applies to a job in the same industry as of current employment. The propensity to apply to the same industry is regressed onto a dummy that equals 1 if the application was sent during the crisis period (Covid), a variable that holds the job distribution on the day the application was sent (i.e. the share of job postings of the industry of the members latest employment), and controls for current industry group of employment. Furthermore, we add an interaction effect of the Covid dummy and the current industry group. This interaction term holds the treatment effect of Covid-19 on the propensity of applying within the same industry. As we control for changes in the job distribution, this treatment effect reflects changes in application behaviour.

The regression results, displayed in Table 2, show that, the Health Care sector and Recreation and Tourism sector exhibit the strongest reactions. We see that for workers in the Health Care sector the propensity to apply within health care increases by about 40 percent^[9] in the crisis period, while it decreases more than 60 percent in Recreation and Tourism. Hence, not just the disproportionate decline in jobs is driving the results but there is a true change in the application behaviour of the workers.

Table 3:

Heat map on the changes in cross-sectoral applications before and after the crisis.

1. Note: Top 5 positive changes; Changes of same magnitude have the same colour. Source: LinkedIn, own calculation.

Furthermore, we checked whether the sectoral trends in applications are correlated with the depth of the crises in the different sectors using a scatter plot of the change in gross value added^[10] between first and second quarter of 2020 against the change in the applications by origin and objective. As Figure 4 in the Appendix A illustrates, there is a strong correlation between the origin of the application and the change in output. The correlation between the change in destination and the change in output is less pronounced. This figure underpins that those workers in the hardest hit sectors are especially concerned and search for jobs in hardly hit sectors.

6 Reallocation Across Sectors

In the last section we have learned that people who work, or have worked, in sectors that are strongly negatively influenced are less likely to apply within their sectors. The question that naturally arises is to which other sector they are applying to.

For this purpose, we analyse the changes in the industries by means of a heat map (Table 3). This table shows in which sector the persons are or were last employed, and in which sectors they applied more often as a result of the Covid-19-crisis. In order to filter the Covid-19-effect out of these transitions, the proportions of applications across the sectors were measured at two different points in time, once before and once during the lockdown, and put in relation to the change in the previous year’s period.^[11] The heat map therefore shows the change in the shares in the course of Covid-19. The strength of the change is represented by the intensity of the colour. The darker the area, the stronger the increase. Only the five strongest increases in an industry were considered. The changes range between 0.1 and 3.0 percentage points.

A look at the heat map shows that people in negatively impacted industries are more likely to apply in industries where demand has picked up during the Covid-19-crisis, such as health, retail, and software and IT service providers. Although the majority of the retail sector was closed during the lockdown, the additional demand in food or online retailing could be behind this development. Persons from the recreation & tourism sector not only applied more frequently in the health sector and retail trade, but also in the IT sectors (hardware & networks, software & IT services) as well as media & communication, the financial sector and the entertainment sector. Overall, however, it is also clear that members also frequently apply in sectors with certain overlaps with their current sector. For example, people from the consumer goods sector often apply in retail or manufacturing, or people from the business services sector apply to software and IT service providers.

7 Job Ladder in the Crisis

Recessions are often thought to have a cleansing effect, destroying relatively unproductive matches. However, match quality could also be impaired via so-called sullying effects.

For instance, Bowlus and Neuman (2006) pointed out that job changes are usually associated with wage increases, i.e. with a certain upward dynamic, and are mainly used to climb up the job ladder. In recession, this career progress is hampered and the pace of reallocation slows down. Furthermore empirical evidence shows that the quality of matches might decline because matches that are created during recessions are rather low-paying and temporary (Barlevy 2002). This implies that match quality would decrease during recession. However, countervailing effects that increase match quality through cleansing exist (see Caballero and Hammour 1994; Shleifer 1986). In that case, match quality would increase leading to more productivity. Which effect dominates is an open question. A recent study of Foster, Grim, and Haltiwanger (2016) finds that the Great Recession differed from earlier recessions and was not cleansing. Though we cannot measure match quality directly, we can give some indication on the career progress and thus job ladder effects of Covid-19 from the application data.

Adjustment processes not only take place via sectors, but possibly also via qualification or experience levels. In order to shed light on this type of adjustment, we will analyse whether people apply for jobs with a corresponding level of seniority or whether they make cutbacks compared to their current level. We measure what proportion of applications are above, at or below the current experience level. This proportion is measured once during the Covid-19-crisis and once during the same period of the previous year.

It turns out that during the crisis as a whole, members applied significantly less often above their own seniority level (drop of 3.3 percentage points from 27.1 % pre-crisis to 23.8 during the crisis), but more often below their own seniority level (increase of 3.3 percentage points, from 29.2 % pre-crisis to 32.5 % during crises). In other words, people on LinkedIn try less likely to improve their own position or climb up the career ladder, but are more likely to make concessions in terms of seniority. The concessions are potentially made to avoid unemployment. Indeed, empirical evidence (Bauer 2016) shows that employees who change careers with an interim period of unemployment suffer a permanent loss of pay. This again illustrates the tension on the labour market due to the Covid-19-crisis.

We further break down the job ladder effects by sector. Table 4 shows a general decline in applications above the own level. Here it is certainly also noticeable that many people are applying across industries. For the increase in the willingness to apply below one’s own level of seniority, a more unequal distribution can be seen. People from the health and IT sector (software & IT services sector, hardware & networks) apply less frequently below their own level, as employment in these sectors is less affected by the crisis or even lifted. By the same token, employees from industries that were closed by the containment measures (education, retail, recreation & tourism, entertainment, wellness & fitness) do not apply more frequently below their level. This seems surprising at first. It is possible that workers anticipated that the lockdown is limited in time and waited accordingly. In industries that were not immediately closed (consumer goods industry, non-profit sector, manufacturing industry, transport & logistics, business services), but are clearly feeling the consequences of the Covid-19-crisis due to the overall economic situation, the proportion of applications below the current seniority level has increased. For example, manufacturing, business services or transport could be affected for a longer period and subject to strong structural change, which apparently increases the risk of job losses.

8 Conclusions

This paper contributes to the discussion on reallocation processes on the labour market induced by the onset of the Covid-19-crises (Barrero et al. 2021). While previous studies concentrate on unemployment, our paper takes a broader perspective on search processes by using data from the career network LinkedIn. This is especially interesting as during the Covid-19 crises reallocation through unemployment was muted by short-time work in Germany. Using LinkedIn data allows to analyse reallocation in a more comprehensive way by looking at applications and intensity of search. The shortcoming however is that the data might be selective to a certain extent in the sense that particular occupations, jobs and hierarchy levels (even within industries) are over- or underrepresented compared to the overall economy.

During the Covid-19-crisis, the number of jobs on offer has fallen significantly. We show that the number of applications per job has risen as a result, which means that competition has intensified. However, job search has not been intensified, as the number of applications per applicant has actually fallen slightly.

Furthermore, the LinkedIn data show that people from industries particularly affected by the crisis apply much more frequently and there has been a significant shift in the target industries for applications. This makes reallocation processes in the crisis evident. Finally, we find that applications are made significantly more often below and significantly less often above a person’s level of seniority than in the previous year. This shift from higher to lower quality applications shows that the crisis is affecting the functioning of the labour market and disables the potential of employees to develop. Besides unemployment hysteresis (Blanchard and Summer 1986) and scarring effects of the youth (e.g. Heckman and Borjas 1980), for this reason, too, the hiring dynamics must be revived as quickly as possible (compare Merkl and Weber 2020).