Entrepreneurs from low-skilled immigrant groups in knowledge-intensive industries: company characteristics, survival and innovative performance

This paper analyzes how companies of immigrant entrepreneurs in knowledge-intensive industries differ from companies of native entrepreneurs with respect to start-up characteristics, company survival and innovative performance. I focus on immigrants from the “recruitment countries” of south and southeast Europe, who arrived in Germany mainly in the 1970s to fill labor shortages. They are the largest immigrant group in Germany and can be reliably identified via ethnic name coding. Companies owned exclusively by immigrants tend to be smaller and have higher exit rates. After controlling for size and other company characteristics, I find no differences in patenting activity compared to companies owned exclusively by natives.


Introduction
Technology entrepreneurship supports the introduction of new products and the diffusion of innovations throughout the economy. It is therefore an important contributor to long-term growth. For a country to improve its national innovative capacity, it is important that contributions to innovation by the country's population are facilitated (Furman et al. 2002). In that context, the present paper will take a comparative look at technology entrepreneurs with and without immigrant background. A specific focus is on start-up characteristics, company survival and innovative performance.
Several studies analyze the contributions of immigrants to technology entrepreneurship in the United States (see, for example, Saxenian 1999). This literature typically focuses on immigrant groups with a high level of education, such as the Chinese and Indians. However, since immigrant groups whose levels of education have been historically lower than those of natives are significant components of the overall populations of many developed countries, it is important to study their contributions to innovation as well. For example, 12.5 % of the US population over the age of 25 is of Hispanic origin. In this subpopulation the share with bachelor's degree is at 13 % substantially lower than in the overall population (28 %); for high school degrees the shares are 61 and 85 %, respectively (US Census Bureau 2009, p. 3).
Germany is an appropriate country to study the participation of low-skilled immigrant groups in technology entrepreneurship. In order to fill labor shortages, Germany had a large influx of immigrants from the so-called recruitment countries of south and southeast Europe in the 1960s and 1970s. These immigrants are a quite homogenous group. They came to Germany to work in dependent employment in the industrial sector and had typically a low level of education at the point of arrival in Germany. Immigrants from recruitment countries are the largest immigrant group in Germany, representing 7 % of the population.
The present study uses information on German companies from the Centre for European Economic Research (ZEW), which is based on data from Creditreform, Germany's largest credit rating agency. The company information was matched with information on patent applications from the European Patent Office. The company owners were identified as immigrants from recruitment countries or natives on the basis of ethnic name-coding performed by the market research company Acxiom.
The results show that compared to companies in exclusively native ownership, companies owned exclusively by immigrants have a smaller start-up size and their founders are younger when they start their company. Furthermore, these immigrant companies have a shorter survival span. Once detailed controls for size and other company characteristics have been included, there is no difference in filing patent applications between these two company types.
In order to develop an efficient economic policy, it is crucial to understand whether there are inefficiencies in the participation of immigrant groups in technology entrepreneurship that could be remedied. For example, smaller company sizes of immigrants point to more limited access to capital. This calls for an investigation into whether immigrants face specific problems with regard to access to capital.
The paper is structured as follows. Section 2 describes the related literature and highlights the paper's contribution. Section 3 provides background information on immigrants in Germany. The data that this study draws on is described in Sect. 4. Section 5 presents the paper's findings on the characteristics and performance of companies owned by immigrants and natives. The conclusions are presented in Sect. 6.

Related literature
This paper is related to the literature exploring the contribution of immigrant entrepreneurs to the innovative performance of the host country. This topic has been analyzed mainly with respect to the United States. One quarter of Silicon Valley companies that were started between 1980 and 1998 are headed by a CEO of either Chinese or Indian ethnic origin (Saxenian 1999). It is likely that those CEOs have also founded the respective companies. With regard to the United States as a whole, Wadhwa et al. (2007) found that at least one immigrant was a key founder in one quarter of engineering and technology companies started between 1995 and 2005. Hunt (2011) differentiates the immigrants' contribution to innovation on the basis of their visa status at entry into the United States. She finds that immigrants with at least a bachelor's degree are more likely to found a company than natives with at least a bachelor's degree. Since many immigrants are highly qualified with a master's or doctoral degree, Hunt concludes that this ''suggests a niche for immigrants in founding companies using specialized academic knowledge'' (Hunt 2011, p. 435). Hsu et al. (2007) also find higher new company formation rates for MIT alumni who are not US citizens compared to US citizens.
The contribution of immigrants to the innovative performance of the host country has also been investigated. Stephan and Levin (2001) find that a disproportional percentage of researchers who have made exceptional contributions to science and engineering projects conducted in the United States are foreign-born or foreign-educated. In Silicon Valley, 32 % of the scientists and engineers in the hightechnology workforce are foreign-born (Saxenian 1999). Investigating the contribution of inventors of either Chinese or Indian ethnic origin to the overall patenting activity in the United States, Kerr (2008a) finds that these immigrant groups are more active than the average US population. In the case of Germany, Niebuhr (2010) finds that at the regional level there is a positive relationship between cultural diversity in the highly skilled workforce and R&D activity, measured as patents per capita.
More generally, this paper is also related to that section of the entrepreneurship literature which analyzes differences between native and immigrant entrepreneurs (see, for example, Waldinger et al. 1990). Such differences include, for example, the determinants of self-employment (Borooah and Hart 1999), the share of self-employed individuals in different ethnic groups (Fairlie 1999), motivations for self-employment (Shinnar and Young 2008), company survival rates (Bates 1999;Georgarakos and Tatsiramos 2009), the role of involvement in the ethnic community (Chaganti and Greene 2002), and the role of network ties back to the home country .
In contrast to the United States, Germany has no tradition of an influx of highly skilled immigrants on a large scale. I am not aware of any study on the contribution of immigrants to the innovative performance of Germany, however there are studies on immigrant entrepreneurs. Leicht et al. (2005b) investigate the self-employment activity of immigrants from recruitment countries in Germany. For immigrants, the possibility of earning more than they would as employees is a more important motive for deciding to start a company than it is for natives. Self-employment could partly be an escape from the discrimination faced in paid employment. Constant et al. (2007) compare native Germans with immigrants. Both groups achieve very similar earnings in self-employment, but interestingly, more years of education in Germany leads to a decrease in earnings for immigrants. This could indicate that highly qualified immigrants find that they have good options in paid employment. Constant and Zimmermann (2006) found that immigrants who feel discriminated against are more likely to opt for selfemployment than natives, but earn less than selfemployed natives. At a regional level, Audretsch et al. (2010) find a positive relationship between the cultural diversity of employees and the start-up rate of technology-oriented companies.
This paper contributes to the literature by studying the participation in the innovative activity of the host country of an immigrant group that has on average a lower level of education than the native population. Typically, the literature focuses on contributions to innovation by high-skilled immigrant groups (see, for example, Saxenian 1999). The paper also adds to the literature by directly investigating start-up characteristics, survival, and innovative output at the company level. Thus, it gives a comprehensive overview on how companies in native versus immigrant ownership differ. Finally, it also departs from other studies in that it uses a company dataset that covers an entire country and is not restricted to a specific region or to a limited number of companies covered by a survey (see, for example, Constant and Zimmermann 2006, for a survey-based approach).

Characteristics of the population with immigrant background
The household survey ''Mikrozensus 2009'' provides current information about the population of immigrant background in Germany. The survey is based on a 1 % household sample and the provision of information is mandatory for the selected households. The group of persons of immigrant background comprises: (a) all foreigners, whether born in Germany or not, (b) all persons who immigrated to the current area of the Federal Republic of Germany after 1949, and (c) all persons born in Germany as Germans, with at least one parent who immigrated to Germany or who was born in Germany but held a foreign nationality at birth.
Here foreigners refers to persons without German nationality and Germans are persons with German nationality. Loosely speaking, the group of persons of immigrant background includes foreigners and immigrants of first or higher generation. According to this definition, a large share of the population in Germany has an immigrant background. Out of the overall population of 81.9 million, 19 %, or 15.7 million, have an immigrant background. The largest subgroup comprises immigrants from recruitment countries, who make up 34 % of the population of immigrant background (5.3 million). The most important single country of origin is Turkey, from which 16 % of the population of immigrant background originates. The other recruitment countries are former Yugoslavia (9 %), Italy (5 %), Greece (2 %), Spain (1 %), and Portugal (1 %). The second largest immigrant subgroup comes from the Former Soviet Union and east European countries. This group makes up 30 % of all immigrants in Germany (4.7 million). Here, the Former Soviet Union is the single most important country of origin, accounting for 16 % of all immigrants in Germany. 2 There are several important differences between the population of immigrants from recruitment countries and the native population. Table 1, which is also based on information drawn from the Mikrozensus 2009, highlights the principal differences. First, immigrants from recruitment countries are on average younger. Whereas 48 % of the native German population is under 45 years old, this share is 68 % in this immigrant subgroup. The differences are even greater in the age group of up to 19 years, from which potential entrepreneurs are most likely to stem in the future. Only 16 % of natives belong to this age group, compared to 24 % in the subgroup of immigrants from recruitment countries.
The education levels of the native population and of immigrants from recruitment countries also differ significantly. Overall, the native population is better educated, as 20 % have a high-school degree (i.e. a degree which allows them to study at a university), compared to only 11 % of this particular immigrant subgroup. Similarly, 9 % of the native population, but only 4 % of the immigrant population, have a university degree. Differentiating immigrants from recruitment countries according to country of origin reveals a high degree of homogeneity with respect to education. The only exception is the Greek population whose share of high-school degree holders is similar to that of the native population, while the share of university degree holders is only 1.4 % points below that of natives. The unemployment rate among immigrants as a whole and per country of origin is much higher than among natives. In fact, the unemployment rate in the overall immigrant population is more than double than in the native population.
Unlike immigrants from recruitment countries, immigrants from the Former Soviet Union and east Europe have a level of education comparable to that of the native population. Nevertheless, the unemployment rate of this subgroup is higher than that of natives (Statistisches Bundesamt 2010).

Entrepreneurs with immigrant background in Germany
When it comes to the likelihood of founding a company, Table 1 shows a lower propensity to found a company for the immigrant population. However, the immigrant group from recruitment countries is not homogeneous. Whereas Italians and Greeks are more likely than the native population to found a company, this propensity is lower among immigrants from Turkey and former Yugoslavia. The legal framework for founding a company in Germany is the same for all immigrant groups. Since 1991, when the law concerning foreigners was revised, every person who has the right to temporary or permanent residence in Germany has the right to found a company (Leicht et al. 2006, p. 32). Consequently, concerning the recruitment countries, there is no difference between Turkey, which is not a member of the European Union, and the remaining countries, which belong to the European Union. Data from the Mikrozensus also allows gaining an overview of the broad sectoral distribution of companies founded by native Germans and by immigrants from recruitment countries (see Table 2). Whereas 6 % of native entrepreneurs are active in knowledge-intensive industries, the share of immigrant entrepreneurs is only 2 %. The latter group's lower participation rate is observed both in knowledge-intensive manufacturing and in technology-intensive services. Entrepreneurs of migration background are overrepresented in non-knowledgeintensive industries, such as restaurants and trade. These sectors do not require a tertiary education and are characterized by relatively low initial investments.
The possibility to start a company was not a consideration for the immigration decision because the legal framework of the bilateral recruitment agreements mandated work as dependent employee. Only in the 1980s, after the majority of the immigrants from the recruitment countries had already entered Germany, did the number of self-employed immigrants grow (Leicht et al. 2006, p. 11). The results of this analysis are therefore probably not influenced by selection effects, which could arise if persons with especially low or high entrepreneurial ability decide to emigrate.

Data source
The present analysis is based on company data from ZEW. The original company data has been provided by Creditreform, Germany's largest credit rating agency. The analysis is restricted to companies in knowledgeintensive industries, where most innovative activity takes place. Table 8 in the Appendix provides a list of the covered industries. As patents are used as an indicator for innovative activity, the analysis is restricted to knowledge-intensive manufacturing and to technology-intensive services. These are sectors in which patents are used as a mechanism to appropriate returns from innovation. The knowledge-intensive service industries of non-technical consulting services are disregarded. One has to acknowledge that patents have limitations as an indicator for inventive activity, for example, because they do not cover all inventions (Grilliches 1990;Rentocchini 2011). However, their use as indicator is well-established in the literature and given the industry selection one can be confident that this indicator provides meaningful results.
The company data includes basic information such as number of employees, year of founding and legal form, as well as the names of the owners. The ZEW data includes companies founded in 1991 or later. Almost all companies founded in Germany in 1998 or Recruitment countries include Turkey, former Yugoslavia, Italy and Greece. Information on Spain and Portugal is not available. ''School education'' and ''professional qualification'' are calculated for the population with completed education phase. ''Propensity to found a company'' is number of self-employed divided by labor force of the specific population group

Source: Statistisches Bundesamt (2010) reporting results from Mikrozensus 2009
Entrepreneurs from low-skilled immigrant groups 875 later are covered, whereas for start-up years 1991-1997 the data coverage is less complete.
Unfortunately it is not entirely clear according to which criteria companies with start-up years between 1991 and 1997 were included in the dataset, but there is an overrepresentation of companies from East Germany. The dataset comprises annual observations up to and including 2007. In order to capture only ''de novo'' companies, the sample was restricted to companies with a start-up size of up to 50 employees and at least one natural person as an owner. An additional restriction was that at least one natural person should be an owner throughout each company's history. These restrictions aimed to ensure that new companies that were merely the result of reorganizations of existing companies were excluded from the analysis, and to make the classification into companies in immigrant versus nonimmigrant ownership possible. The entire ownership history of the companies is observed. The names of the current owners are recorded at an annual basis, so it is possible to keep track of ownership changes.
The company data was matched with information on patent applications from the European Patent Office. Each match was manually checked.

Name coding
In the ZEW database, the migration background of owners is identified according to their first and last names. An alternative would be to use information on nationality, but this information is not available in the data. The owners' ethnic background was coded by the German subsidiary of the global market research company Acxiom (www.acxiom.com) on the basis of a name list normally used to identify the ethnic background of potential customers. Initially entrepreneurs are allocated to an ethnic class according to their last name. Subsequently the results are refined and entrepreneurs may be reclassified if the combination of first and last name suggests a different ethnic origin. The data provider reports that the accuracy of identifying ethnic background lies between 90 and 95 %. 3 Nevertheless, it was not possible to code all combinations of first and last names, because of e.g., typos or because certain names had not been included in the Acxiom database. It was, however, possible to classify the ethnic identity of 94 % of company owners. Immigrants are probably slightly overrepresented in the group of unidentified owners. The name coding system differentiates between the following areas of origin: Turkey, former Yugoslavia, Italy, Greece, and Spain/Portugal/Latin America. These origins make up the recruitment countries. It is not possible to differentiate among owners from Spain, Portugal or Latin America on the basis of name and surname, since the names in all three regions are similar. For the purpose of this study this poses no problem, since few persons from Latin America actually live in Germany. Table 9 in the Appendix gives examples of names typical of each ethnic group according to country of origin.
Although it would be interesting to analyze the second largest immigrant group in Germany, which comprises immigrants from the Former Soviet Union and east Europe, it was not possible to do so using the ZEW database. Members of that subgroup are often ethnic Germans, with typical German first or last names, so identification of ethnic origin on that basis is less reliable. To analyze this particular subgroup it would be necessary to conduct a survey that requested explicit information on immigration background. Name coding has advantages as well as limitations. A big advantage is that name coding can be applied to large-scale datasets, whereas in tailor-made surveys that collect information specifically on migration background sample sizes are often limited. In the case of Germany, the Mikrozensus, which surveyed 1 % of the population, does include questions on migration background, however it has only very limited information on businesses. There is another large-scale household survey in Germany, the Socio-Economic Panel (SOEP), however, this includes only information on nationality and, like the Mikrozensus, offers only limited information on businesses. A further advantage of name coding is that it identifies ethnic origin and is thus not restricted to nationality. This is especially useful in the case of people who acquire German citizenship but may still be perceived as immigrants.
The main disadvantage of name coding is the probabilistic nature of the results. This means that there is no certainty that a specific owner has been correctly classified. Nevertheless, name coding has been used in previous studies; for example, Kerr (2008b) identified US inventors according to their Chinese or Indian origin. A further disadvantage is the inability to differentiate between generations of immigrants. According to a calculation by the National Statistical Office, based on data from the 2009 Mikrozensus, 38 % of immigrants from recruitment countries are second generation or higher. In one important characteristic, namely education, first and second generation immigrants from recruitment countries still differ from Germans. With the exception of female second generation immigrants from Greece, first and second generation immigrants from the former recruitment countries leave full-time education statistically significantly earlier than Germans (Algan et al. 2010, p. F14). 4 In the present study, given the large number of observations and the still significant differences between second-generation immigrants and natives, I am confident that name coding provides a meaningful representation of the different population groups. Table 2 presents the descriptive statistics for the dataset that includes 795,190 observations with information on 129,466 companies. Observations for companies in exclusively immigrant ownership make up 2.0 % of all company-year observations. In these companies every owner is an immigrant from a recruitment country. 5 Companies in mixed immigrant and native ownership represent 1.3 % of all observations. The category 'immigrant participation in ownership' comprises observations with exclusively immigrant ownership as well as with mixed immigrant and native ownership and comprises 3.3 % of all observations. In absolute numbers, 3,146 companies in the sample were owned only by immigrants at start-up, 2,003 companies were in joint immigrant and non-immigrant ownership, and 124,317 companies were owned by natives. Determining the category of ownership is time variant, but the variable displays a high degree of persistence because ownership changes are not that common. Figure 1 shows an upward trend for the share of companies with immigrant participation in ownership at the time of start-up. In the 16-year period from 1991 to 2007 the share of companies in this ownership category increased from 2.6 to 3.8 %. 6 This upward trend is an encouraging sign of better integration of immigrants into the economic activity of Germany, the host country. 5 For simplification purposes, in the rest of the paper I'll refer to immigrant entrepreneurs from recruitment countries as ''immigrant entrepreneurs,'' unless otherwise stated. Also as a simplification, the term ''immigrant'' may refer to an immigrant of the first or of a higher generation. 6 Data from the Mikrozensus 2009 allows a very crude plausibility check for the name coding method. In knowledge-intensive industries there are 4,000 self-employed immigrants from recruitment countries (1.9% of a total of 216,000 self-employed immigrants from recruitment countries), compared to 192,000 self-employed native Germans (5.4% of a total of 3,537,000 selfemployed native Germans). Therefore, entrepreneurs of immigrant background represent 2.1% of all entrepreneurs in knowledge-intensive industries. The share of companies with immigrant participation in ownership in the ZEW data that was identified on the basis of name coding is 3.3%. The latter value is likely to be higher, because the share of immigrant participation is increasing over time and the ZEW data contains not the overall population of companies, but only start-ups.

Descriptive statistics
Entrepreneurs from low-skilled immigrant groups 877 The majority of companies are of small size, with 4.7 employees on average. 7 The company-year observations that concern companies with at least one patent application come to 1.4 %. The average size of the application stock for observations with at least one application is 2.3. The application stock is the cumulative sum of the patent applications filed with the EPO up to the year in question. The observations concerning companies with financing from at least one venture capitalist (VC) come to 0.3 and in 4.8 % of the observations at least one owner holds an academic title (''Dr.'' or ''Professor''). It is possible that this variable is an underrepresentation of the qualifications of immigrant entrepreneurs. If a doctorate has been obtained abroad it may not be recognized in Germany and the German habit of attaching academic titles to a name is not universal, so it is possible that the original data collector, Creditreform, will not be aware of recognized titles. In 38 % of the observations companies have at least two owners and in 6.6 % of the observations there is a corporate investor, meaning that at least one other company holds an ownership stake (Table 3). The average age of the owner(s) at start-up is lower for companies in immigrant ownership and in mixed ownership than for companies in native ownership. It is not influenced by the number of owners a company has. Age can be taken as a crude proxy for the sum of education and work experience and thus as a proxy for human capital. As the difference in average age between companies in native ownership and companies in immigrant ownership is about 3 years, immigrant entrepreneurs do have less human capital when they found their companies. Having said that, the age difference may reflect different start-up strategies or the overall younger age structure of the population of immigrant background. 8

Company survival
This subsection examines whether there are differences between immigrant and native companies with respect to survival. The dataset includes exact closing dates for forced closures (bankruptcies). In the case of voluntary closures, approximate dates are used wherever an exact closing date is not available. If the information that Creditreform holds on a particular company is not updated within a period of 5 years, it is assumed that the company was closed 1 year after the last update. With regard to company takeovers, it is not clear a priori whether they should be regarded as cases of closure or survival. For the purposes of this paper, takeovers are treated as survival, since the company continues to exist although under different ownership. 9 It should be noted that the results are almost identical when takeovers are regarded as closures, because takeovers constitute only 2.7 % of all closures. Of the 31,831 exits in the sample, 827 relate to companies in exclusively immigrant ownership at the time of exit, and 547 relate to companies in mixed immigrant-native ownership at the time of exit. 10  Table 10 in the appendix provides a further breakdown of company characteristics by ethnic group. The analysis with respect to size (at start-up and overall) and average age of the owner(s) at start-up confirms the patterns found in Table 4. Companies with at least one owner from a specific country of origin are similar to each other but differ significantly from the companies in purely native ownership. Companies in Turkish ownership tend to differ to a more extreme extent from companies in native ownership and differ significantly from companies with immigrant entrepreneurs from other countries for some measures. 9 Henkel et al. (2010) provide evidence for acquisitions as a positive outcome for entrepreneurs that can be part of an entry strategy. 10 I would like to thank my former colleague from ZEW, Sandra Gottschalk, for making the exit variable available to me.
Entrepreneurs from low-skilled immigrant groups 879 Table 5 presents the results of Cox regressions. 11 The first results represent a parsimonious specification with only one dummy for limited liability and location in East Germany as controls. Companies with immigrant participation in ownership have with 1.34 a higher hazard rate of exit than native companies (column 1). The same is true when this category is broken down into companies in exclusively immigrant ownership and companies in mixed immigrant-native ownership with hazard ratios of 1.39 and 1.29, respectively (column 2). The hazard ratios are marginally lower when detailed controls for company characteristics are included (columns 3 and 4). In view of this, the higher exit rates of start-ups with any type of immigrant participation cannot be explained as a result of differences in company characteristics. 12 Companies survive over a longer period if they are larger, their owners are older and if at least one owner holds an academic title. 'Owner age' is used in a quadratic form to allow for a higher exit probability when owners near retirement age. The hazard of exit is higher for companies with VC financing, which is plausible, since venture capital is a financing source for start-ups with high risk and high expected return. The hazard of exit is also higher for companies owned by at least two persons or at least one other company. In addition to controlling for available funding, using the dummy for at least two owners makes it possible to differentiate the influence of immigrant ownership from that of size alone.
The higher exit rate of immigrant companies may be an indication of different start-up strategies. Immigrants start companies at a younger age, i.e. with less education and/or work experience. One reason is that they may be more willing to found a company at a young age, because this gives them more time to establish themselves in paid employment should the venture fail. However, as in the present study it was not possible to control for all company characteristics, there might be other explanations for the higher exit rates of immigrant companies; for example, lower capital intensities or more limited access to financial resources. Overall, the quite limited drop in the exit rate after controlling for owner age suggests that age is only a very crude proxy for education. Otherwise a larger influence on survival would be expected.
It could be that through the venture the immigrants are gaining work experience and then are able to get hired by established companies. Gimeno et al. (1997) shows that sometimes lower human capital individuals can persist in entrepreneurship, not due to their performance but due to a lack of outside options. It could also be that immigrants close successful companies in order to return to their countries of origin. However, the credit ratings included in the data set suggest that immigrant entrepreneurs are not closing successful companies. At the time of exit the ratings of immigrant companies are not better than those of   190 795,190 795,190 795,190  Based on ZEW data. The failure event is exit of the company. Hazard ratios are shown. All regressions contain industry dummies at the 2-digit SIC level, as well as dummies for the start-up periods 1991-1994, 1995-1997, 1998-2000, and 2001-2003. Standard errors are shown in brackets * Significant at 10 %, ** significant at 5 %, *** significant at 1 % Entrepreneurs from low-skilled immigrant groups 881 companies in native ownership and ratings markedly deteriorate for companies that are closed. The higher exit rate could also be partly explained by the higher unemployment rate that immigrants from recruitment countries face in Germany. Unemployment or the threat of unemployment can be a powerful push-factor for starting a company. However, companies that start under pressure may be less stable. Leicht et al. (2005b) found that escaping unemployment while living in Germany is a slightly more important motive for entrepreneurs with ethnic origin from Turkey than for native entrepreneurs.

Innovative performance
The innovative performance of companies is approximated by patent applications to the European Patent Office (EPO). Unfortunately, there is no information on R&D expenditure, which would also be an interesting measure of innovative performance. The category of ownership tends to be very time-persistent, e.g. in the present data, only 0.6 % or 812 companies changed ownership category after starting up; therefore I chose not to use panel estimators. Standard errors have been adjusted for within-company correlation across time and for heteroscedasticity. Table 6 shows the results of probit regressions for the probability of having filed at least one patent application with the EPO. The sample comprises 1,836 companies that have filed patent applications. Twenty-one of the companies in exclusively immigrant ownership and 55 of the companies in mixed immigrant-native ownership have filed for patents. 13 When only basic control variables are used (column 1), there is no difference between start-ups with immigrant participation in ownership and start-ups in purely native ownership. Differentiating between the two immigrant ownership categories reveals a significant negative effect for start-ups in exclusively immigrant ownership and a significant positive effect for start-ups in mixed immigrant-native ownership (column 2). This, however, reflects differences in company characteristics, i.e. once company characteristics have been controlled for, the differences between the two ownership categories disappear (columns 3 and 4). Combined with the evidence on smaller start-up sizes for immigrant entrepreneurs, this result suggests that lack of access to capital could constitute a severe problem for immigrant entrepreneurs. The probability of patenting would increase if companies in immigrant ownership would work with the same size and the same other company characteristics as companies in native ownership.
All additional company characteristics have a positive and at least at the 5 % level positive significant influence on the probability of patenting. Here owner age is included in a logarithmic form because the experience of higher age should increase the patenting probability without a reversal of sign. It is plausible that the negative effect observed in companies in exclusively immigrant ownership vanishes once company characteristics are controlled for, because this type of company works with fewer employees and has also lower values for the other characteristics.
To gain a better understanding of the participation of natives and immigrants in innovative activity in Germany, it is useful to compare the relative importance of sector selection and the probability of patent application. Concerning sector selection, Table 2 shows that immigrant entrepreneurs are 50 % less likely than natives to found a company in a knowledgeintensive industry. Differences in the probability of patent applications can be inferred from the marginal effect observed for companies owned exclusively by immigrants (column 2 of Table 6). The marginal effect of -0.0028 implies a 0.28 % points smaller probability of having applied for a patent for this type of company. The overall probability of having at least one patent application is 1.4 % in the dataset. Thus, comparing the 0.28 % point reduction for companies owned exclusively by immigrants to the base value of 1.4 % it follows that companies owned exclusively by immigrants have a 20 % smaller probability of having a patent application. The calculation shows that the comparatively greater hurdle for immigrants seems to lie in sector selection (50 vs. 20 %). Once a company has been founded in a knowledge-intensive industry, the differences between companies owned exclusively by immigrants and those owned exclusively by natives are more limited. 14 13 The number of observations for companies with patent applications in immigrant ownership is unfortunately not sufficient to allow a breakdown of the analysis according to country of origin.
14 The lower probability of patenting for immigrant companies without controls for resources does not necessarily imply that Table 7 investigates the correlation between ownership category and size of the patent application stock. Poisson regressions are used for the analysis.
The findings confirm the results obtained for the probability of patenting. Without controls for company characteristics we find that companies in exclusively immigrant ownership have significantly fewer patent applications and companies in mixed ownership have significantly more patent applications compared to companies in native ownership (column 2). Once controls for company characteristics are employed (column 4), these differences disappear. It can therefore be concluded that, once company characteristics have been taken into account, innovative performance, as measured in the form of number   190 795,190 795,190 795,190 Companies 129,466 129,466 129,466 129,544 Based on ZEW data. The dependent variable is a dummy for at least one patent application. Marginal effects are shown. The regressions contain industry dummies at the 2-digit SIC level and year dummies. Standard errors in brackets allow for heteroscedasticity and for autocorrelation within companies * Significant at 10 %, ** significant at 5 %, *** significant at 1 % Footnote 14 continued immigrant companies are resource constrained. It is possible that immigrant entrepreneurs select to create less innovative ventures which require less investment in R&D and fewer resources. This leads to an endogeneity problem that does not allow this type of analysis to separate out whether the venture was resource constrained and could not undertake innovation or whether the venture was not trying to innovate and thus did not require additional resources for R&D.
Entrepreneurs from low-skilled immigrant groups 883 of patent applications, is not lower for companies in purely immigrant ownership than for companies in other ownership categories. This result is to be expected because patent applications, as an output measure of innovation, depend mainly on the input factors used. Owners typically only engage in innovation if their companies have sufficient financial resources to sustain the innovation projects over several years. Peters (2009) analyses persistence in innovation and points out that innovators incur sunk costs of establishing an R&D department or hiring R&D staff to start an innovation program. To summarize, the main problem of producing patentable innovations is the availability of resources. As we have seen in the descriptive statistics, immigrant companies work with fewer resources as they are smaller in size. Once immigrant companies have the same resources as native companies, no difference in innovative performance remains. In contrast, the results of the previous subsection do not show a similar reduction in the exit probability after company characteristics are controlled for. The controls are probably less encompassing Based on ZEW data. The dependent variable is the size of the patent application stock. Coefficients are shown. The regressions contain industry dummies at the 2-digit SIC level and year dummies. Standard errors in brackets allow for heteroscedasticity and for autocorrelation within companies * Significant at 10 %, ** significant at 5 %, *** significant at 1 % in explaining exit than in explaining innovation. For example, different motivations to found a company are relevant for survival but motivations are not included in the data. As immigrants have a higher unemployment rate than natives, the push-motivation of escaping unemployment may be more relevant and may lead to the founding of less stable companies. The results do not show a positive influence of the ethnic diversity of mixed immigrant-native ownership on innovation. This is in line with Østergaard et al. (2011) who did not find a significant relationship between the ethnic diversity of employees and the probability of introducing an innovation for Danish companies. Also, Hart and Acs (2011), who investigate high-growth companies in the US high-tech sector, did not find differences in technological performance for immigrant-and native-founded companies.

Conclusion
This paper analyses differences in company characteristics and performance between native entrepreneurs and immigrant entrepreneurs for Germany. The results show that companies in exclusively immigrant ownership have on average a smaller start-up size and are led by younger owners compared to companies in native ownership. Companies with immigrant participation in ownership have a significantly shorter survival probability. However, there are no differences between immigrants and natives in the probability of filing a patent application and in the size of the patent application stock, when company characteristics, such as size, are controlled for.
From the smaller start-up size of immigrant companies it follows as policy implication that lack of access to capital could constitute a problem for immigrant entrepreneurs. The ability of immigrant entrepreneurs to raise capital in order to finance their companies is probably more limited than that of natives for two reasons. Firstly, the first generation typically found low-paid jobs when they arrived in Germany, therefore immigrant parents are likely to have relatively limited means to support their children. Secondly, immigrants from recruitment countries have on average more children than natives (Statistisches Bundesamt 2010), which means that within a family the potential financial support per child is proportionally smaller. Since size is an important determinant for company survival and patenting, it would be of great interest to establish whether immigrants face specific problems with regard to access to capital. In the United States, for example, Blanchflower et al. (2003) found that companies owned by black entrepreneurs are more likely to be refused credit than can be explained by their economic situation alone.
A limitation of this study is that no detailed information about the educational level of the owners is available. It would also be useful to have information about the education level below academic titles. Furthermore, the dataset does not allow for an identification of corporate spin-offs. They have been identified as a particularly successful form of entry (see, for example, Klepper (2001) for evidence for the United States). If persons with immigrant background have less opportunities for good jobs as employees, they are also presumably less likely to produce good spin-offs. Spin-offs are a prevalent form of entry in Germany. Lejpras and Stephan (2011) find for knowledge-intensive industries in Eastern Germany that 27 % of companies started off as corporate spinoffs.
A worthwhile subject of future research would be the specific challenges that immigrant entrepreneurs face. This could be investigated with the help of tailormade surveys focusing on access to capital, familiarity with German and European institutions, and the degree of fluency in the German language.    Grupp and Legler (2000) and Nerlinger and Berger (1995). The first column gives the industry code according to the German industry classification WZ (Klassifikation der Wirtschaftszweige) from 1993 Entrepreneurs from low-skilled immigrant groups 887 The category ''Germany'' covers companies with exclusively native owners. For the other categories at least one owner from the respective country is required. The comparison group in column 3 is the country with the median value of the respective country means. The p-values for the two-sided t tests for comparison of means are in brackets