German financial state aid during Covid-19 pandemic: Higher impact among digitalized self-employed

ABSTRACT In response to strong revenue and income losses facing a large share of self-employed individuals during the COVID-19 pandemic, the German federal government introduced a €50bn emergency-aid program. Based on real-time online-survey data comprising more than 20,000 observations, we analyze the impact of this program on the confidence to survive the crisis. We investigate how the digitalization level of self-employed individuals influences the program’s effectiveness. Employing propensity score matching, we find that the emergency-aid program had only moderately positive effects on the confidence of self-employed to survive the crisis. However, self-employed whose businesses were highly digitalized, benefitted much more from the state aid than those whose businesses were less digitalized. This only holds true for those self-employed, who started the digitalization processes already before the crisis. Taking a regional perspective, we find suggestive evidence that the quality of the regional broadband infrastructure matters in the sense that it increases the effectiveness of the emergency-aid program. Our findings show the interplay between governmental support programs, the digitalization levels of entrepreneurs, and the regional digital infrastructure. The study helps public policy to improve the impact of crisis-related policy instruments, ultimately increasing the resilience of small firms in times of crises.


Introduction
The lockdowns to contain the spread of COVID-19 significantly affected the economy, especially the service sector.The strict containment measures paralyzed many self-employed individuals.Accordingly, empirical evidence shows that entrepreneurs, self-employed individuals, and microbusinesses (referred to as 'self-employed individuals' from now on) were more likely to face income losses than employees and that income losses were also stronger among self-employed individuals than among employees (see inter alia Graeber, Kritikos, and Seebauer 2021;Kalenkoski and Pabilonia 2022).In response, governments in many countries introduced ad-hoc emergency aid programs to financially support self-employed individuals get through the crisis.In Germany, the Minister for Economic Affairs announced a €50bn emergency aid program (Soforthilfe) at the end of March 2020, offering one-off lump-sum payments of up to €15,000 to those facing substantial revenue declines and stated at the program announcement on 10th of March 2020, 'that we will not let any firm owner down', thus aiming to increase the resilience of self-employment to survive the crisis. 1 The program received a wide response with more than 2.6 million applications out of 4.2 million self-employed persons in Germany (BMWi 2021).Given the huge amount of taxpayer money that was made available for the program, it is important to understand whether, and, in particular, under what conditions, this support program was effective and achieved its goal.
At the same time, a public discussion started about the extent to which the digitalization level of self-employed individuals increases their ability to survive such a crisis, in that self-employed persons with higher levels of digitalization are better able to react flexibly to exogenous shocks and take advantage of new entrepreneurial opportunities.Prior literature points to positive effects of digital entrepreneurship on firm performance in crises times (Papadopoulos, Baltas, and Balta 2020).For instance, during the 2008/09 financial crisis, highly digitalized firms were hit less hard by the consequences of the crisis than firms with a low degree of digitalization, thus emphasizing the importance of digital technologies for firm resilience (Bertschek, Polder, and Schulte 2019).
This article combines the public policy perspective with the digital entrepreneurship perspective.We investigate how the effectiveness of the support program depends on the degree, type, and stage of digitalization among entrepreneurs.Further, we assess the role of the regional digital infrastructure.We argue that digitalization helps self-employed individuals react to the business restrictions introduced through lockdown and containment measures, perhaps also taking advantage of new entrepreneurial opportunities.In doing so, we reveal the interplay between the support program and the digitalization level of entrepreneurs impacting their resilience and confidence to survive the crisis. 2 In our analysis, we proceed in four steps.We start by investigating the overall effect of the emergency aid program on all self-employed individuals, and examine whether the program increased the confidence of self-employed individuals to get through the COVID-19 pandemic without having to quit their business.In our main analysis, we then study the role of digitalization in leveraging the benefits of government support programs in times of crises.In this regard, we investigate (i) the relationship between the confidence to survive the crisis and the degree of digitalization of self-employed individuals; and (ii) how this relationship depends on digitalization area (digitalization of the product, process, customer relations and distribution channels).Third, we also examine whether the timing of investment into digital technologies matters.Some selfemployed individuals had invested into such technologies before the crisis.Others spontaneously purchased digital technologies right after the start of the pandemic.The underlying question of the third part of the analysis is whether the adoption of digital technologies is a long-term strategy that needs complementary investments in order to improve resilience.Fourth, we investigate whether the kind of geographical area where firms are located, matters for the policy effectiveness.As disparities in digital infrastructure may exist between urban and rural areas, these disparities may influence the impact of the emergency aid.
Our empirical analysis is based on a survey comprising the answers from more than 20,000 selfemployed individuals in spring 2020.The data set contains information on crisis-related sales losses, resulting liquidity constraints, and the willingness to apply for financial support from the emergency fund.It also includes information about whether financial support was actually received.Finally, the respondents provided information on several individual-and firm-related characteristics relevant for self-employment, including information on the state of digitalization in the business.Our outcome variable is a measure based on the confidence of the individuals about the probability that they will have to exit from self-employment due to the COVID-19 pandemic in the following 12 months.This measure informs us about the assessment of the self-employed individuals to get their businesses through the pandemic.As such confidence is the crucial condition for self-employed individuals to actively run and further invest in their firms, we interpret this measure as an indicator for resilience (Ayala and Manzano 2014), which is needed for entrepreneurial survival and success in challenging times (Chadwick and Raver 2020;Hartmann et al. 2022).
To study the role of digitalization in the self-employed individuals' ability to resist economic crises and how it complements a state support program, we estimate a treatment effects model with propensity score matching.We hypothesize that self-employed individuals whose businesses have a higher degree of digitalization -either prior to the crisis or obtained through investments in the first months of the crisis -are more flexible and better able than other selfemployed individuals to react to new entrepreneurial opportunities, which allows them to benefit more from the support program.
The results of our study support this notion.We find that the program was able to achieve its goals, but particularly increased the confidence to survive the crisis among those self-employed individuals whose businesses were at advanced digitalization stages.Moreover, taking a regional perspective, we find suggestive evidence that regional digital infrastructure matters and that the quality of the broadband infrastructure increases the effectiveness of the emergency aid program.With our study, we contribute to the growing literature on digitalization and digital entrepreneurship (Berger et al. 2021;Nambisan 2017;Steininger 2019;Steininger, Kathryn Brohman, and Block 2022) as well as on how digitalization helps entrepreneurs achieve greater flexibility and resilience (Kreuzer et al. 2022;Ojala, Evers, and Rialp 2018;von Briel, Davidsson, and Recker 2018).Our study also contributes to the literature on entrepreneurship in crisis situations (Doern, Williams, and Vorley 2019;Belitski et al. 2022;Newman, Obschonka, and Block 2022) and how state aid can help entrepreneurs to master such challenging situations.In the context of public support for entrepreneurship, our study suggests that the effect of state aid depends also on regional digital infrastructure.This aspect of our study connects to a discussion on how the regional digital context and infrastructure shapes the mindset, actions, and resilience of entrepreneurs and their regions (Sussan and Acs 2017;Naldi, Larsson, and Westlund 2020;Belitski, Korosteleva, and Piscitello 2021;Biru, Gilbert, and Arenius 2021).

The COVID-19 pandemic and policy measures in Germany
Germany observed its first COVID-19 cases in January 2020, which increased steadily over the following months.At the time of the survey in April and May 2020, Germany was among the most affected countries.On 22nd of March 2020, two weeks before our data collection, the German government imposed severe restrictions.Among others, the government closed schools and daycare centers, public events were cancelled, and most shops, restaurants, and hotels, except for supermarkets, were shut down.In particular, the travel and hospitality industries were severely restricted.Only two individuals were allowed to meet in a public gathering and a distance of at least two meters had to be kept from each other in public spaces.Simultaneously, the Institut für Weltwirtschaft (IfW 2020) predicted that the GDP would decline by up to 9% in 2020.Self-employed individuals were among the most affected by the policy measures to stop the spreading of COVID-19.Many of them had to deal with a temporary inability to work, meaning they were not able to generate revenues to cover the operating costs of their business or their personal living costs.
To help the economy as well as to avoid massive unemployment and a severe recession, the German government initiated several policy measures to reduce the economic consequences of the pandemic.Targeting employees, Kurzarbeit -short-time work -was opened for the pandemic, with the Federal Employment Office covering a substantial portion of wage costs for affected employees (Cahuc 2014).Self-employed individuals, however, were not covered by this instrument.That is why the government launched the Soforthilfe, which was an emergency aid package comprising €50 billion.The package was accessible between end of March and end of May 2020.This ad-hoc program aimed to help solo self-employed individuals and owners of micro firms (with up to 10 employees) maintain their ventures and increase their chances of entrepreneurial survival.The applicants to this program could receive up to €15,000 as immediate financial assistance if they were able to demonstrate strong revenue declines.Its main limitation was that it could be used only for covering fixed operating costs, but not for living costs (Federal Ministry for Economic Affairs and Energy 2020).The analysis of this program's effectiveness is at the center of our analysis.

Existing research on self-employment during the COVID-19 pandemic
Crisis-related research on self-employment is considerable (see, e.g.Doern, Williams, and Vorley 2019;Laskovaia et al. 2019).Three strands of literature can be distinguished when analyzing relationships between the COVID-19 pandemic and entrepreneurship.The first strand documents that there is an important difference between the COVID-19 pandemic and earlier crises in the current century: the COVID-19 pandemic affected the majority of the self-employed population.Contemporaneous research shows how, and who among self-employed individuals suffered most strongly, in an economic sense, from the consequences of the COVID-19 pandemic (Bartik et al. 2020;Blundell and Machin 2020;Fairlie and Fossen 2022b;Graeber, Kritikos, and Seebauer 2021;Kalenkoski and Pabilonia 2022).Further research points to other effects of the pandemic on self-employed people beyond economic losses: First descriptive (Torrès et al. 2022) and causal (Caliendo et al. 2022) evidence reveals a worsening of mental health among self-employed individuals in the wake of pandemic-driven market distortions.
Beyond examining on how the pandemic affected self-employment, a second strand of research investigates how self-employed individuals coped with the early stages of the COVID-19 pandemic.Bertschek and Erdsiek (2020) show that self-employed individuals with a higher degree of digitalization are less affected by the crisis.Block, Fisch, and Hirschmann (2022) investigate how they tried to maintain their liquidity, particularly through the use of bootstrap financing, while Meurer et al. (2022) demonstrate how entrepreneurial online communities offered support to the affected entrepreneurs.
The third strand investigates the role of financial support and other government programs in responding to this economic disruption.Various public instruments were developed and provided by governments around the world in response to the financing needs of entrepreneurs and small businesses.Fairlie and Fossen (2022a) provide a disbursement analysis of two programs, the Paycheck Protection Program and the Economic Injury Disaster Loan Program, both in the US, that aimed to help disadvantaged groups.Atkins, Cook, and Seamans (2022) add to the understanding of the role of race in loans made through the Paycheck Protection Program.For China, Liu et al. (2022) show the supportive role of Chinese state-owned banks for small businesses' lines of credit where a broad policy mix was introduced consisting of loan guarantees, direct lending to small businesses, grants, and equity instruments.Block et al. (2022) for Germany and Belghitar, Moro, and Radić (2022) for the UK investigate the effects of governmental policies for SMEs during COVID-19 and examine whether these interventions affected the ability of SMEs to survive the pandemic.Dörr et al. (2022) point to the downside of such policies as they claim to have detected an insolvency gap that is characterized by firms that were already struggling before the pandemic.
The current study bridges these three strands of literature.While earlier impact studies of public support programs find only moderate effects and argue that such programs may simply postpone insolvencies, our analysis explores heterogeneities among self-employed individuals and shows which factors among the treated influence the effectiveness of such programs.In other words, by investigating which kind of activities among self-employed individuals themselves increase the impact of state aid, we are able to further the understanding how entrepreneurial resilience can be developed to master and survive a challenging crisis like COVID-19.This way, our study not only contributes to the entrepreneurship literature dealing with COVID-19 but also to the more general literature on the antecedents of entrepreneurial resilience (Hartmann et al. 2022) and the role of digital entrepreneurship therein.

Theory and hypotheses
In this study, we aim to analyze how the emergency aid impacts the confidence of self-employed individuals to survive the crisis and to remain in this occupational form despite the COVID-19 pandemic, as well as how this impact is affected by digitalization.Our theoretical arguments are derived from the growing digitalization literature applied to the particular context of the COVID-19 pandemic.Digital technologies differ from other technologies in a number of important aspects.As general-purpose technologies (Bresnahan and Trajtenberg 1995), they are characterized by rapid change (technological dynamism), by general applicability (pervasiveness), and by innovational complementarities, i.e. firms applying these technologies are able to develop new products, services, new business models, or new distribution channels, thus making their processes more efficient.
There are different methods to measure firms' use of digital technologies and level of digitalization.A first approach employs general measures like investments in digital technologies or information and communication technologies (ICT) (see, for instance, Bertschek and Kaiser 2004;Bloom et al. 2014;Bresnahan, Brynjolfsson, and Hitt 2002).A second method focuses on the adoption and use of specific technologies, like enterprise software (e.g.Engelstätter 2012) or big data analytics (e.g.Niebel, Rasel, and Viete 2019).A third line refers to the digitalization of different business areas.For instance, Trischler and Li-Ying (2022) point out that products and services of self-employed individuals as well as production, distribution, and administrative support processes could be digitalized.Rosin et al. (2020), based on Hull et al. (2007), consider digital products and services, digital processes, digital customer interaction, and digital collaboration.In a similar vein, in our analysis, we consider the digitalization of (1) products and services, (2) internal business processes, and (3) customer relations and distribution channels.In our large-scale survey, self-employed individuals were asked to self-assess the degree of digitalization of these three business fields on a five-point Likert scale, which we use to construct three levels of digitalization: low, medium and high (see section 4.1 for details). 3

How digitalization boosts the impact of the emergency aid through entrepreneurial opportunities and flexibility (the digitalization effect)
The unique characteristics of digital technologies, particularly its general applicability and innovational complementarity, provide important benefits for its users in terms of new entrepreneurial opportunities and additional flexibility.We argue that these two benefits of digitalization help selfemployed individuals to master the COVID-19 pandemic and increase the impact of the emergency aid received.
Using digital technologies does not just affect efficiency, e.g. through more efficient communications or a reduction of production costs.It also creates new entrepreneurial opportunities for selfemployed individuals, e.g. through expansion into new markets and enabling new ways of value creation and value capture leading to new business models (Berger et al. 2021;Steininger 2019;Steininger, Brohman, and Block 2022).A self-employed individual with a primarily local customer base is able to use digital tools or networks to expand into new geographical areas.Research shows that digital entrepreneurs recognize and exploit emergent opportunities differently from non-digital entrepreneurs (Kreuzer et al. 2022).Moreover, they can use digital platforms to grow and internationalize more rapidly (Ojala, Evers, and Rialp 2018;Jean, Kim, and Cavusgil 2020).
Digitalization also provides self-employed individuals with greater flexibility to react to unforeseen situations, challenges, and opportunities.Adaptability through reprogramming and scalability makes it easy to adjust the products, services, internal processes, and business models to new situations (Lyytinen, Yoo, and Boland 2016).For instance, a self-employed individual offering consulting services normally face-to-face is able to offer these services online using videoconferencing tools.In line with this example, research from the 2008/09 economic and financial crisis shows that firms with a high level of digitalization were more flexible in the sense of being able to implement process innovation and to maintain a high level of productivity throughout the crisis (Bertschek, Polder, and Schulte 2019), thus being more resilient.
The advantages of digitalization described above put self-employed individuals with a high degree of digitalization into a better position to benefit from the emergency aid.Hence, the impact of the emergency aid on highly digitalized self-employed individuals should be larger.As pointed out above, we distinguish between the digitalization in three business areas, namely products/ services, internal processes, and the customer relations and distribution channels.We hypothesize for all three areas a positive digitalization effect.Our baseline hypothesis is:

H1 (the digitalization effect):
The impact of the emergency aid on the confidence of self-employed individuals to survive the crisis and remain in self-employment is stronger for self-employed individuals with a high degree of digitalization than for those with a low degree of digitalization.This digitalization effect should apply for digitalized products or services, digitalized internal processes, and digitalized customer distribution channels.
After having derived the basic digitalization effect that strengthens the impact of the emergency aid, we now dig deeper into the heterogeneous nature of this effect.Thereby, we consider the moderating effects of digitalization area, initial digitalization stage, and the regional digital infrastructure.Such a perspective is consistent with a process perspective on digital transformation comprising different stages (Matt, Hess, and Benlian 2015;Verhoef et al. 2021), as well as the ecosystem approach in the context of resilience among entrepreneurs (see Iacobucci and Perugini 2021) and digital entrepreneurship (Sussan and Acs 2017).

The influence of digitalization area on the digitalization effect
We argue that the digitalization effect differs by digitalization area.This can be explained by the nature of the COVID-19 pandemic.The COVID-19 pandemic and its related business restrictions had a particularly strong effect on product distribution and customer interaction as it was not possible to meet customers in person or, for instance, visit trade fairs.Digital ways of product distribution and customer interactions became of high importance and those self-employed individuals with digital distribution channels and digital customer interactions were more flexible and had an advantage compared to other self-employed individuals.Digital marketing research has investigated the benefits of digital marketing in great detail.Digital tools, platforms, and marketplaces allow firms to apply dynamic pricing (e.g.Bigne, Nicolau, and William 2021;Wurman 2001), to involve their customers in the innovation and production processes, thus deriving more customized and individualized products and services (e.g.Franke and Piller 2004;Maier et al. 2021), as well as to reach a broader and more heterogeneous customer base (e.g.Dinner, Van Heerde, and Neslin 2014;Li and Kannan, 2014).Moreover, digital marketing tools help firms to keep contact with their existing customers, thus increasing brand and customer loyalty (Lou and Xie 2021), while also leveraging and creating opportunities for cross-selling and repeat purchases (Kumar et al. 2016).The latter two aspects should be of particular importance in the context of the COVID-19 pandemic.Such comparative advantage resulting from digitalization may be weaker for digital internal (administrative) processes as such processes could still be carried out in a non-digital way during the COVID-19 pandemic.Accordingly, the digitalization effect with regard to the impact of the emergency aid should be stronger for the digitalization of distribution channels and customer interaction as compared to the digitalization of internal processes.The following hypothesis applies: H2a: The digitalization effect is stronger for digitalized distribution channels or customer relations than for digitalized internal processes.
A similar argument is raised for the case of digital products and services or physical products with digital elements.The COVID-19 pandemic particularly affected those self-employed individuals with non-digital products and services.Sectors like restaurants and accommodation, events, and wellness were hit particularly hard (see, for example, Bertschek and Erdsiek 2020; Graeber, Kritikos, and Seebauer 2021; Zopiatis, Pericleous, and Theofanous 2021).Performing such non-digital services was impossible as physical customer interactions are an integral part of service provision.However, depending on their sector affiliation, some self-employed individuals intensified the digitalization of their products and services during the pandemic.Such digitalization efforts allowed them to still deliver their products and services to their customers despite not being able to meet their customers in person.For example, knowledge-intensive (professional) service firms involved in teaching, consulting, and coaching used the COVID-19 pandemic as an opportunity to deliver their services off-site (from home) instead of providing them on-site at the customer's premises.Walsman (2022), analyzing the operational adaptations that consulting firms made in response to COVID-19, shows that the pandemic was an important booster for service innovation in the activities of these professional service firms.
To summarize, we posit that self-employed individuals with a digital product or service or a high level of digitalization in their products or services had a comparative advantage compared to those with a lower level of digitalization in their products or services.Such self-employed individuals are better able to benefit from the emergency aid and use this aid to defend or improve their competitive position.This digitalization advantage should be weaker for self-employed individuals with digitalized internal processes that, despite the COVID-19 pandemic and the mandated restrictions, can only be carried out in a non-digital way.The following hypothesis is formulated: H2b: The digitalization effect is stronger for digitalized products or services than for digitalized internal processes.

The influence of initial digitalization stage on the digitalization effect
Digitalization is not a one-off event but rather a long and complex process (see, for example, Brynjolfsson and Hitt 2000;Brynjolfsson, Rock, and Syverson 2019).Several models of digitalization maturity or readiness exist in the literature (e.g.Lokuge et al. 2019;Mittal et al. 2018).The process involves substantial (organizational) set-up and adaptation costs, with the benefits of, for instance, new entrepreneurial opportunities and increased flexibility taking time to materialize.Those selfemployed individuals who only started to digitalize their products, processes, and distribution channels during the crisis are still at a relatively early stage of the digital transformation and, thus, may not yet have seen or experienced the advantages associated with digitalization.They are still in a situation of high uncertainty, primarily experiencing the psychological and economic costs associated with the digitalization process and what this means for the entrepreneurial business models and working life.
In turn, self-employed individuals who already started with the digitalization before the crisis are at a more advanced stage of the digitalization process and, hence, are more likely to realize the benefits and opportunities from digitalization.The uncertainty with regard to the transformation process and digitalization benefits is lower.Thus, public subsidies, like the emergency aid, may have a larger effect on their confidence to survive the crisis with their firms.Moreover, digital technologies are considered general purpose technologies, enabling new ways to create value for customers (von Briel, Davidsson, and Recker 2018).Only firms beyond a certain level of digitalization may benefit from the opportunities associated with digitalization as an enabling technology.Related to this reasoning, digitalization provides flexibility to firms that was of high value during the COVID-19 crisis.This additional flexibility only becomes possible beyond a certain level of digitalization.

H3:
The digitalization effect is stronger for those self-employed individuals who started the digitalization process already before as compared to during the crisis.

The influence of regional digital infrastructure on the digitalization effect
In this section, we argue that regional disparities in digitalization may exist and that these regional disparities influence the impact of state subsidies for self-employed individuals, like the emergency aid.The main argument is that a minimum level of digital infrastructure, in terms of available wireline-or mobile-based fast broadband, is needed so that self-employed individuals are better able to benefit from their investments in digitalization (see, for example, Akerman, Gaarder, and Mogstad 2015, on the relevance of broadband availability for firm performance, or Duso, Nardotto, and Schiersch 2021, showing that advanced broadband applications increase firm productivity, and Barrero, Bloom, and Davis 2021, on the relevance of internet access on working from home during the COVID-19 pandemic).Digital technologies have strong (direct and indirect) network effects, meaning that the value of using digital technologies increases with a larger number of users, amount of available data, and the installed base (Katz and Shapiro 1985;Gregory et al. 2021).Oftentimes, these network effects are highly localized (Kim et al. 2021), meaning that it is not the total number of users of a digital tool or platform that matters but how many close collaborators are using the platform or tool.We suggest that, for the above described digitalization benefits to occur, it is not enough that self-employed individuals themselves have a good digital infrastructure, but their customers, suppliers, and other business partners also need to have a good digital infrastructure and a digital mindset so that a strong regional ecosystem of digital entrepreneurship and small businesses can emerge (Belitski, Korosteleva, and Piscitello 2021;Sussan and Acs 2017).This may become a problem for primarily locally-oriented self-employed individuals from regions with a weak digital infrastructure where negative spillover effects may occur.Thus, we argue that the impact of the emergency aid for self-employed individuals depends not only on the digitalization level of selfemployed individuals themselves but also on the digitalization level of the region in which they are located.Accordingly, we hypothesize: H4: The digitalization effect is stronger for those self-employed individuals located in regions with a strong digital infrastructure as compared to self-employed individuals located in regions with a weak digital infrastructure.
Figure 1 provides an overview of our conceptual framework displaying our baseline hypothesis about the digitalization effect as well as the moderating hypotheses relating to the effects of digitalization area, stage, and regional digital infrastructure.

Survey data
We empirically analyze the hypotheses using unique real-time online-survey data collected from more than 27,000 self-employed persons at the beginning of the COVID-19 pandemic between 7th of April and 4th of May 2020, in Germany. 4The data set contains rich information on the selfemployed persons' financial situation, subjective future prospects, their willingness to apply for government support programs, the application process and outcome regarding the emergency aid program, as well as socio-demographic and business characteristics including details on the ventures' digitalization level.The main aim of this investigation is to examine whether the emergencyaid program increased -depending on their digitalization level -the confidence of the selfemployed individuals in their ability to survive the crisis.To gather information on this outcome variable, respondents were asked in the survey to assess the likelihood of quitting self-employment within the coming year due to the pandemic.We use this information to construct our dependent variable, capturing the confidence of the respondents that their ventures will survive the crisis, ranging from 1 ('very unlikely') to 5 ('very likely').Block et al. (2022) provide a further description of the survey setup and present various descriptive statistics.Importantly, they show that applicants for the emergency aid program were severely affected by the crisis and that they expected financial hardship to continue for at least 6 months.
After cleaning the data of clearly erroneous entries (e.g.inconsistent entries reporting application dates before the policy measure was introduced etc.), missing values, and after excluding persons living outside of Germany, we obtain an estimation sample of 16,859 self-employed individuals.Half of the persons are less than 50 years old and 48% of the surveyed are male.The majority is highly educated, with 61% possessing a university degree.The regional dispersion is shown in Table A1 in the Appendix and follows the German population pattern, with most of the surveyed living in the most-populated states (North Rhine-Westphalia, Bavaria, and Baden-Württemberg) and in larger cities like Berlin, where the share of self-employed persons in total employment is particularly high (Fritsch, Kritikos, and Sorgner 2015;IfM, 2019).Our survey slightly oversamples self-employed persons in Berlin and Hamburg, while it underrepresents those in Baden-Württemberg, which we account for in the empirical strategy.At the time of the survey, 59% of the respondents had applied for the government support program and 34% had received the payment.The remaining persons were either waiting for the payment (4%) or for a decision (19%).Overall rejection rates were low.
Self-employed individuals were asked to assess their degree of digitalization on a five-point Likert scale (very low, rather low, medium, rather high, very high).We summarized the categories very low and rather low as well as rather high and very high leading to three categories.Having detailed information on the ventures' degree of digitalization prior to the crisis as well as changes in the degree of digitalization during the first months of the crisis, we uncover important differences in the stage of digitalization within and across ventures.As the first panel in Table 1 shows, 59% of the respondents had digitalized internal processes prior to the crisis (medium + high level), whereas a smaller share of respondents had made investments into digitalizing products and services, as well as customer relations.This changed to a certain extent during the first months of the crisis (March to May 2020), with about 20% of the respondents reporting to have increased the digitalization level of their products and services, or their customer relations (Table 1, second panel).The variation shows that when the COVID-19 pandemic started, entrepreneurs were at different stages in their digital transformation processes and, thus, had different prerequisites to meet the challenges posed by the containment measures and the economic recession.

Regional disparities in broadband infrastructure
To test whether the impact of the emergency aid is larger in regions with a strong digital infrastructure, we combine the survey data with information on the broadband infrastructure at the federal state level.Fast internet access is widespread in the German city states of Berlin, Bremen, and Hamburg, while federal states that are characterized by rural regions, in particular the federal states in eastern Germany, lag behind.Thus, whereas almost every household in the city states has the possibility of connecting to the internet at the minimum speed of one gigabit per second, only 12% of the households in Saxony-Anhalt can do the same (Figure 2).

Estimation strategy
To address our main research question and to study the role of digital technologies in the ventures' ability to survive the pandemic crisis, we estimate a treatment effects model and analyze how a firm's digitalization status affects its future when taking advantage of the government support program.As the main outcome measure, we use the reported confidence of remaining in self-employment for one year after receiving financial support, where we compare between ventures that are in different stages of the digital transformation process.We estimate the average treatment effect of the treated based on the Roy (1951)-Rubin (1974) model with two potential outcomes, Y 1 and Y 0 , and a binary treatment variable D equal to one if the individual receives the treatment and equal to zero otherwise.Since we do not observe the counterfactual outcome of the treated had they not received the government payment (and vice versa), we cannot estimate the individual treatment effect but follow Roy (1951) and Rubin (1974) by relying on population averages and assuming conditional independence.The latter builds on the fact that the counterfactual outcome of the treated cannot be easily approximated by the observed outcome of the untreated since individuals might self-select into treatment and the treatment group might systematically differ from the control group.Conditional independence controls for these systematic differences and assumes that, once we condition on a set of observable characteristics X, the potential outcome is independent of treatment assignment, obtaining the following formula for the ATT: Thus, the individuals from the comparison group are matched to the treated units in such a way that the mean distribution of the covariates in the matched control group resembles that of the treatment group, which is formally expressed by the outer expectation E X ½:jD ¼ :� (Caliendo and Kopeining 2008).We apply propensity score matching to reduce the covariates distributions to a single balancing score P(X).We trim the sample to observations within the region of common support using the criteria max(min{P(X)|D = 1,P(X|D = 0)}) and min(max{P(X|D = 1),P(X|D = 0)}) at the tails of the propensity score distribution to ensure that all individuals with the same values of X have a positive probability of being treated and untreated.For the matching procedure, we use an Epanechnikov kernel to obtain a weighted average of control units whose outcome is compared to the treated unit, with the kernel bandwidth being chosen by cross-validation.
Our outcome variable is the confidence of the self-employed persons in spring 2020 to survive the crisis and to remain in this occupational form over the next 12 months despite the COVID-19 pandemic.We consider this outcome measure as useful: it allows us to measure psychological response in times of a crisis.The aim of the emergency aid program was to avert the firm closure or bankruptcy of selfemployed individuals and micro-sized ventures with up to 10 employees whose economic survival was threatened by the COVID-19 pandemic.In addition to financial support, the program was also intended to provide self-employed individuals with psychological support, reassuring them that the government would not abandon them and that they could continue their business despite the crisis.This psychological component is particularly important in the context of entrepreneurship and resilience: the personal assessment of self-employed individuals about their future prospects influences the effort they put into their venture and, thus, the likelihood of its survival.In fact, a high confidence to survive the crisis is a prerequisite for realized survival, since it is the entrepreneurs who make the decision whether or not to continue their business.We measure the outcome as a binary variable set to 1 if selfemployed individuals perceive continuation as likely and 0 otherwise, which allows us to interpret the estimated ATT as percentage point changes in survival probability.
Our treatment variable is the participation in the government program, where our focus is on comparing the treatment effect between ventures with different levels of digitalization.Possessing detailed information on each individual's application process and outcome, we consider all individuals who received the payment out of the emergency aid program (n = 5,743) as treated.The control group comprises all persons who were planning to apply for the program but have not yet done so at the time of the survey (n = 1,013).The advantage of this approach is that respondents who are inclined to apply are likely to share important characteristics with those who have already applied regarding their financial situation, their venture's characteristics, etc., compared to individuals who did not (and will not) apply.In addition, we apply propensity score matching to match treated and untreated individuals based on a rich set of covariates that are likely to affect the application for the emergency aid as well as the respondents' expectations about their ventures' prospects.Notably, we control for (i) personal characteristics such as age, gender, self-employment experience, education, and risk tolerance; (ii) for venture characteristics covering full-time vs. part-time self-employment, the number of employees, industry, the location of the business (federal state); and (iii) the self-employed person's financial situation measured by the private costs of living, support from the basic income scheme, solvency, and the revenue decline due to the crisis.Finally, we control for the expected duration of financial hardship and the calendar week in which each individual was surveyed.
We verify the quality of the propensity score matching by calculating standard statistics such as the mean absolute standardized bias (Rosenbaum and Rubin 1985), which is reduced to 2.1% after the matching compared to 6.8% before the matching.Caliendo and Kopeining (2008) argue that successful bias reduction should result in a mean absolute standardized bias of below 5%.Furthermore, we calculate Rubin's B, determining that the means of the covariate distributions in the treatment and control groups are 0.26 standard deviations away from each other after matching, which is close to the optimal value of one quarter suggested by Rubin (2001).The ratio between the propensity score's variances of both groups, commonly referred to as Rubin's R, is 1.28 and, thus, well inside of the optimal interval [.5;2] (Rubin 2001).To verify that there is sufficient overlap in the propensity score distributions between the treatment group and the control group, we plot the distributions in Figure 3.We find continuous common support within the interval [.11,.99], confirming that the min-max-trimming criterion is valid and unrestrictive.

Main results
Table 2 presents the main results: the emergency aid program positively affected self-employed individuals in their assessment of future prospects, with the confidence of self-employed individuals who received financial support to survive the crisis being moderately increased by 6.5% points.Turning to the underlying factors that make some entrepreneurs feel more confident that they will survive the crisis than others, we reveal important effect heterogeneities with respect to digitalization.Notably, positive effects only occur among those individuals whose ventures were highly digitalized prior to the crisis (11.0%points) -whereas in ventures with low or medium degrees of digitalization, the survival probability did not significantly increase despite their participation in the emergency program (hypothesis 1).We observe this pattern in all digitalization areas (cf.Tables 3 and 4), further confirming hypothesis 1.Moreover, Table 4 suggests that effects are particularly strong among those who had digitalized their customer relations as well as their products and services, with an increase in the confidence to survive the crisis of 13.9% points and 10.8% points, respectively.The latter finding is plausible given the fact that social distancing measures during the COVID-19 pandemic especially affected these areas (hypotheses 2a and 2b).However, due to the smaller sample sizes, the coefficients are less precisely estimated and are not significantly different from each other at the 5-percent-level.Consequently, while indicative evidence is suggested, we cannot confirm hypotheses 2a and 2b from a statistical perspective.
During the crisis, many self-employed individuals decided to make ad-hoc investments into digitalizing their distribution channels and customer interactions in order to catch up with respect to their digitalization strategy.Thus, we investigate in the empirical analysis whether it matters for the program effect when these investments happened.As a matter of fact, the emergency aid could only increase the confidence of self-employed individuals to master the crisis if they had made their digitalization investments before the crisis started.
For those who tried to catch up by spontaneously investing into digitalizing their businesses during the first months of the crisis, the support program had no effect (Table 2, last column) (hypothesis 3).Table 5 examines further heterogeneous program effects, distinguishing the German city states of Berlin, Bremen, and Hamburg, which have very high shares of households with access to high-speed Internet (cf. Figure 2).These cities contrast with the rest of Germany, including more rural regions.The results show that the government program was effective throughout Germany.Although the estimated coefficients suggest that the program was more effective in the urban areas (hypothesis 4), the difference between the two coefficients is not significantly different from zero.Hence, there is only first indicative evidence that differences in the broadband infrastructure between rural and urban areas affect the confidence of self-employed individuals to master the crisis with the help of the government program.To gain further insights and obtain more robust results, it would be necessary to classify each observation according to the population density of its county, which, unfortunately, is not possible due to missing information on the exact location of the surveyed persons and their business. 5  These results show that a high degree of digitalization is an important factor for the confidence to survive the crisis, making financial support more effective in terms of resilience and the assessment of future perspectives.However, this also implies that digitalization is a tedious process that cannot be made up in the short term.Therefore, developing a long-term strategy for digitalizing different business areas is a prerequisite for building confidence that one will be able to master the consequences of such a crisis.Further, digitalization should not be restricted to internal business processes, but should comprise the development of digital products and services as well as the relationships with customers and distribution channels.The latter were particularly decisive for getting through the COVID-19 crisis.

Robustness checks
In the main analysis, we use an Epanechnikov kernel estimator to match the treated individuals to the control group.We aim to assess the robustness of these results and, therefore, re-estimate the treatment effects model using nearest-neighbor-matching with two neighbors and replacement to check whether our results depend on the choice of the matching algorithm.The results, listed in Table 6, are very similar to those obtained with the Epanechnikov kernel estimator in terms of both size and efficiency.The results for the digitalization area analysis are shown in the Appendix, Table A2 and A3.They confirm once again the pivotal role of digitalizing customer relations and distribution channels in weathering the crisis.

Discussion and conclusion
The COVID-19 pandemic and the related government-mandated lock-down measures severely affected self-employed individuals.In response, the German government introduced a financial support program, for which €50bn was budgeted and €13.7bn spent.It was intended to help selfemployed individuals bridge the negative economic consequences of the COVID-19 pandemic and to survive the crisis with their businesses.At the same time, during the early stage of the pandemic, it became clear that self-employed individuals would be better able to cope with the consequences of the lock-down measures if they had digitalized their businesses.Therefore, in this study we investigate how the German emergency aid program affected the confidence of self-employed individuals to get through the crisis and how the level of digitalization of the business of the self-employed individuals impacted the effect of the emergency aid.In other words: we are interested in the interplay between the support program and the digitalization strategy of entrepreneurs.For our analysis, we use a real-time online survey answered by more than 20,000 self-employed individuals in spring 2020, which captured a rich set of information on variables that influence the selection into the treatment as well as the outcome variable, namely the confidence to survive the crisis.We use these data to implement a propensity score matching approach and compare self-employed individuals with differing digitalization levels who received the grant versus those self-employed individuals with the same digitalization levels, but were planning to apply for this support.
Our analysis delivers three important findings: First, overall, the emergency aid program had only a moderately positive effect of 6.5% points on the confidence to survive the crisis among all observed self-employed individuals.However, self-employed individuals whose businesses were highly digitalized benefitted significantly more strongly from the state aid than self-employed individuals with less digitalized businesses.Effects are particularly strong among those who had digitalized their customer relations, as well as among those who digitalized their products and services, with an increase in the confidence to survive the crisis of 14% points in the former case and 11% points in the latter.Second, these particularly strong positive effects among those selfemployed individuals who were running highly digitalized businesses only hold if the digitalization processes had happened before the crisis.Third, at the regional level, there is indicative evidence that the quality of the digital infrastructure -an important prerequisite for introducing and using digital technologies that is generally better in urban areas -supports the effectiveness of the emergency aid program in the same direction: effects seem to be stronger in urban areas with a high quality of digital infrastructure than in rural areas.
Increasing the confidence in being able to survive such crises is an important condition for selfemployed individuals to further believe in their own business and for their willingness to further invest in it.Therefore, we interpret increases in the confidence to survive shocks like the pandemic as an increase in resilience.In that sense, our findings imply that having sufficiently digitalized their businesses ahead of the crisis was an important prerequisite for self-employed individuals to develop resilience to this very kind of unforeseen economic shock.The digitalization of businesses enables them to react flexibly to changes in demand and disruptive events affecting their business processes.The provided financial support then seems to be a crucial complement to such investments, as it significantly increased the confidence of these self-employed individuals to get through the COVID-19 pandemic, even though these businesses, despite their digitalization, had also suffered from the crisis.This psychological effect is important in the context of entrepreneurship, since the entrepreneurs' personal assessments of future prospects influence their efforts and decisions taken with respect to business development and, thus, business survival (Chadwick and Raver 2020).Our analysis reveals that the confidence of self-employed persons whose ventures were poorly digitalized did not increase despite financial support from the government.It shows that financial support from the government alone is insufficient and cannot fully take effect in times of crises if, at the same time, there is a lack of the necessary capacity among the ventures (Simón-Moya, Revuelto-Taboada, and Ribeiro-Soriano 2016) to exploit the financial support in a sensible way or to develop plans for alternative scenarios while the crisis was ongoing.
Of similar importance is the insight that the emergency aid program had no effect among those who invested into digitalizing their businesses during the crisis.It seems that only ventures that had previously developed long-term strategies to embed digital technologies into their business processes are able to face the crisis and remain confident.Short-term digital solutions, despite the financial support from the government, did not seem to inspire the necessary confidence to master the crisis.
Our analysis is not restricted to the effects of the support program at the individual level; it also looks into regional effects.Accordingly, our results referring to the regional perspective provide indicative evidence that strong broadband infrastructure can support the effectiveness of an emergency aid program, thus influencing resilience not only at the individual level but also at the regional level (Audretsch and Belitski 2017).Still, although this part of the analysis needs to be supplemented with further research, it suggests that a strong digital infrastructure is a prerequisite for adopting and using digital technologies as well as for being able to connect to suppliers, customers, and other business partners.
From a more general point of view, our findings imply that it is important to consider the interplay between a support program and factors of heterogeneity that are connected to the treated individuals, in this case the level of business digitalization.In this specific case, digital technologies do not just shape how self-employed individuals engage with customers but also the effectiveness of policy instruments meant to support the very same self-employed individuals.This is an important policy conclusion of our analysis for the design of future policy measures, particularly if public policy is aiming to increase the resilience of self-employed individuals during a crisis and to avoid deadweight losses or the financing of firms that have no real perspective ahead.Moreover, it is clear that developing resilience as an entrepreneur appears to be a long-term process and part of a more complex strategy in which digitalization represents an important component.Such investments need to be well-prepared to unfold positive effects in times of unforeseen shocks and economic crises.In that sense, developing resilience strategies has become a key issue for entrepreneurs.Against this background, public programs targeting the digitalization of small and medium-sized enterprises (SMEs), like the German 'go digital' program, might play an important role in supporting SMEs' or self-employed individuals' long-term digitalization and, thus, long-term resilience in times of crises.Designing such programs in a way that allows their proper evaluation with respect to SMEs realizing higher degrees of digitalization as well as their contribution to a better resilience is an important task for policymakers as well as for researchers.Developing a digitalization strategy and investing in digitalization remains, however, first and foremost, the task of self-employed individuals.
Moreover, supporting the expansion of a strong digital infrastructure, especially in rural areas, is important if self-employed individuals should become able to digitalize their own products, services, and customer relations, thus improving their resilience in times of crises at the individual and regional levels.Although the important role of broadband infrastructure for economic growth is the subject of existing research, more evidence is needed on the regional contribution of broadband to businesses' resilience in times of crises.
Our analysis has some limitations with respect to the data used and the methodology applied.As with all cross-sectional studies relying on propensity score matching, we are not able to account for unobserved characteristics that might explain differences in resilience or in the fact that an entrepreneur receives emergency aid.For instance, although we know the propensity among selfemployed individuals to take risks, we do not have information on further dimensions of entrepreneurial orientation such as innovativeness and proactiveness.The degree of proactiveness could affect not only the likelihood of applying for the program (treatment), but also correlate with the impression of having control over the situation and, thus, influencing optimism about weathering a crisis.Likewise, innovativeness could help in adapting the business model to new circumstances during a crisis.Using a rich dataset comprising very detailed information on self-employed individuals, their venture, and the application process -including, for instance, information about the expectations of self-employed individuals regarding the duration of the pandemic and the financial hardship it will cause -we are confident to have minimized any biases from omitting relevant aspects from the analysis.However, we cannot rule out the possibility that some bias remains.In addition, we are not able to deepen the regional analysis since we only have information about the federal state that each responding self-employed individual is located in.We can neither control for broadband availability in the exact municipality of the self-employed individuals nor for sociodemographic differences between urban and rural regions.To tackle these challenges, panel data and more precise information on the location of self-employed individuals is necessary.
Overall, we conclude that the program had strong positive effects on the confidence of selfemployed individuals to survive the crisis, but only if they had digitalized their businesses prior to its outbreak, while it had no measurable impact on non-digitalized businesses.This points to potential cleansing effects among these businesses in the sense that, thanks to this government support program, self-employed individuals with digitalized businesses will have survived the crisis with higher probability than those who had not yet digitalized their businesses, even if the latter were also financially supported.Thus, the program, although costly, was helpful, in that it strengthens those entrepreneurs who use digital technologies.Our results refer to the German economy, which is characterized by a large share of SMEs compared to other European countries.However, German SMEs have a relatively low level of digitalization compared to many other countries (European Commission 2022).This implies that our results represent a kind of lower limit of the effects of an emergency aid program in such a context.Future research needs to follow-up on this and analyze the long-term impact of the pandemic, government measures, and digitalization strategies on firm survival.

Notes
1. Minister for Economic Affairs press conference on 10th of March 2020 (see (https://www.businessinsider.de/wirtschaft/handel/altmaiers-versprechen-in-deutschland-soll-kein-unternehmen-wegen-des-coronavirus-pleitegehen/, accessed 21st of May 2021).2. We apply the psychological approach of resilience, i.e. the ability of self-employed individuals to master a shock and to be able to continue to serve their customers despite the significant changes in customer demand that were induced by the pandemic and its containment measures (Bhamra, Dani, and Burnard 2011;Fletcher and Sarkar 2013).3. Some authors differentiate between digitization and digitalization (see, for instance, Ritter and Pedersen 2020).
We refer to digitalization as using digital technologies in order to change products and services, internal business processes or customer relations and the way of distribution.The term digitization as the process of converting information from a physical format into a digital one is not applied here.4. The survey was administered via the Verband der Gründer und Selbstständigen Deutschland e.V. (VGSD) and other self-employment associations.5. We only observe the federal state where the business is located in without any further information on the county or municipality.

Table 1 .
Digitalization level among German self-employed persons in 2020.

Table 2 .
ATT by digitalization level before and during the crisis.

Table 3 .
ATT for ventures with low digitalization level by digitalization area.

Table 4 .
ATT for ventures with high digitalization level by digitalization area.

Table 5 .
ATT by region.