THE ROLE OF STATISTICS IN SOCIETY NEW STATISTICS IN SOCIETY 2020 A Swedish database for studying entrepreneurship Fredrik Andersson Statistics Sweden Maria Håkansson Statistics Sweden Johan Karlsson Örebro University
A Swedish database for studying entrepreneurship Fredrik Andersson 1, Maria Håkansson 2, Johan Karlsson 3 Abstract There is an increasing demand among policymakers and researchers to study the composition and nature of entrepreneurship. There is, for example, a significant interest to study entrepreneurship from a gender or equity perspective. Moreover, policymakers and researchers are becoming increasingly interested in the dynamics of firm ownership, and family ownership in particular. We present a method for identifying entrepreneurs in administrative data, where entrepreneurs are defined as operative business leaders (OBL). OBLs are identified using a hierarchical decisionmaking procedure based on employer-employee data regarding occupation, income and age. We also present a method for identifying family firms in administrative data, where family firms are identified as firms which are controlled by one family. Family firms are identified by using a hierarchical decision-making procedure based on information on kinship, ownership and firm governance. We find that family firms are more likely to assign female OBLs compared to other firms; this likelihood is particularly great when the OBL is related to the controlling family. We also find that family firms have a higher labor turnover rate and that they contribute to a vast majority of all jobs created in Sweden during the studied time period. 1 Statistics Sweden, Senior Advisor, Örebro, Sweden, Email: fredrik.andersson@scb.se 2 Statistics Sweden, Örebro, Sweden, Email: maria.hakansson@scb.se 3 Örebro University, phd, Örebro, Sweden, Email: johan.karlsson@oru.se
1. Introduction Researchers, students, politicians and others who are interested in studies of entrepreneurship may derive benefit from the contents of the new Swedish Entrepreneurship Database, which now covers the years 2004-2014. The database provides a broader perspective on entrepreneurs and business leaders and studies of entrepreneurship can be performed from different angles, e.g. gender, country of birth, age, region or industry. A central aspect of the Entrepreneurship Database is being able to identify who is an operational business leader. We identify one individual per business (per corporate ID number) whom we consider to be the operational leader. This register-based identification is carried out in accordance with an algorithm based on a hierarchical decision-making procedure constructed with the help of information from the Swedish Companies Registration Office, the Occupational Register, statement of earnings details, the Business Partner Register and individual data on income and age. The Entrepreneurship Database has great potential and after discussions and request from external users Statistics Sweden will soon be able to measure family owned businesses in the Entrepreneurship Database. The identification of family owned businesses has also been made from an algorithm based on hierarchical decision making through use of different kind of data sources. Both the algorithms used to measure operational business leader and family owned businesses will be presented in this paper. 2. Background It all started in 2007 when the project Integrated database of enterprises, entrepreneurs and business leaders from a gender perspective was carried out. The project was funded by the Ministry of Industry, Employment and Communications. The aim of the project was to paint a more complete picture of entrepreneurs, business leaders and individuals with an entrepreneurial connection in the Swedish business sector. A statistical database, the Entrepreneurship Database, was created to highlight various issues concerning entrepreneurs and business leaders from a gender perspective in order to illuminate the gender distributions of entrepreneurs. Earlier this was a rather difficult category of people to identify and hence study various outcomes from policies changes. The database was based on information supplied by the Swedish Companies Registration Office, the Swedish Occupational Register, statement of earnings details, the Business Partner Register and individual data on income and age.
3. Operational business leader 3.1 Definition The person who runs the enterprise on a day-to-day basis. 3.2 Data sources To identify who the operational business leaders are, we use several data sources. These include the Companies Registration Office s limited liability company Board Member Register, the Occupational Register and the Register-based Labour Market Statistics (RAMS). 3.3 The Companies Registration Office s Business Register The Swedish Companies Act (ABL) requires all limited liability companies to be registered at the Companies Registration Office. Limited liability companies have four levels of governance. The owners appoint the board, which in turn appoints a president/managing director. The board is responsible for the administration of the company while the president runs the enterprise on a day-to-day basis. The annual shareholders meeting appoints the auditors who scrutinise how the administration of the enterprise has been handled by the board and the president. All these people are registered at the Companies Registration Office. If a limited liability company does not have a board, it may be forced to terminate its activities, ABL 13:4A. The president: Under the Companies Act, a public limited liability company with a share capital of at least SEK 500 000 must appoint a president who is to take care of the daily running of the company. In these cases, we know for certain that the president is the de facto operational business leader. Smaller limited companies may also have a president who takes care of the day-to-day running of the company, but adherence to the act in this context is optional.4 If the limited liability company has a president, the board may appoint one or more vice-presidents. Applicable parts of the Company Act shall also apply to vice-presidents. A president can only be one individual in a single limited liability company. If no president has been appointed, the board is solely responsible for the internal business management, i.e. responsible for taking care of the company s affairs. Board member: The Swedish Companies Act puts requirements on the number of board members and alternates. If the share capital exceeds SEK 500 000 (public company), the number of board members must be at least three. In private 4In large companies, the president may not be the chairperson. An external president is entitled to be present at board meetings.
companies, the number board members may amount to one or two if there is at least one alternate.5 In boards of more than one member, one of them must be the chairman. The terms in the Companies Act that apply to board members shall, where appropriate, also apply to alternates. If a board member cannot attend a meeting and there is an alternate who can replace him or her, the alternate must be given the opportunity to do so. Authorised signatory and registered agent: If it so wishes, the board may appoint an authorised signatory for the company. In the case where no board member or external authorized signatory is domiciled in Sweden (president not included), then at least one registered agent who is domiciled in Sweden must be appointed by the board. 3.4 The Occupational Register Statistics Sweden is responsible for producing an occupational register containing data on the occupations of the gainfully employed Swedish population. The Swedish Standard Classification of Occupations (SSCO) is based on two main concepts, type of work being performed and the level of qualification normally necessary. The Occupational Register is official statistics. Management work: The Occupational Register contains data on presidents, agency heads, etc., which we use in the algorithm to determine who the operational business leaders are. The two relevant occupational groups for us are, i) presidents, agency heads etc. for large enterprises, i.e. with ten or more employees, and ii) managers of smaller companies and units, for small enterprises, i.e. with fewer than ten employees. 3.5 Register-based Labour Market Statistics (RAMS) RAMS consists of a number of tables than can be matched with each other in different ways, depending on whether it is the individual or the enterprise that is in focus. The register is based on all statement of earnings data to do with gainfully employed persons and on sole proprietors standardised accounting statements. Employing the method used in RAMS, the professional status which the individuals have as their main activity is classified. The classifications are: employees, sole proprietors, entrepreneurs in own Limited Liability Company and combiners (employees associated with entrepreneurship as a secondary activity). 5 The terms in the Companies Act that apply to board members shall, where appropriate, also apply to alternates.
Sole proprietors: Sole proprietor means a person who takes decisions on, or has influence over the daily running and the strategic questions in an enterprise. This refers to a close company or one-man company, in accordance with Swedish taxation terminology, i.e. sole proprietorship, partnership or limited partnership. In labour market statistics, one-man entrepreneurs, i.e. those who run sole proprietorships and those who run partnerships, are amalgamated into Sole proprietors. According to the definition of Sole proprietors in the labour market statistics, the person must have reported a surplus/deficit from active business activities in his/her normal tax return as a sole proprietorship or in partnerships or limited partnerships. Entrepreneurs in their own limited companies: According to the definition of Entrepreneur in own limited liability company, the individual must have specified ownership in a closely held company. If the person is given the professional status of Entrepreneur in own limited liability company, the person s chosen statement of earnings for the month of November refers to the company of which the person is a partner. The partnership is identified via the yearly tax information. Combiners: We call people associated with more than one professional status combiners. Professional status can be employee, entrepreneur or entrepreneur in own limited liability company The status in which the individual has his/her highest income in November will also be his/her main professional status. 3.6 Method The register-based method we use to determine operational business leaders is a hierarchical decision-making procedure, the underlying algorithm for which has been constructed using information from the Swedish Companies Registration Office, the Occupational Register, statement of earnings data, as well as individual income and age data. The hierarchical decision-making procedure: To be able to use a hierarchical decisionmaking procedure, a point function has been constructed based on the functions/positions of the individuals in the enterprises. Table 1 shows how many points an individual receives depending on his or her functions in the enterprise. If an individual is the CEO/president/managing director according to the Companies Registration Office, (s-)he receives 1 000 000 points, president according to the Occupational Register (for a company with 10 or more employees) 100 000 points, board member 400 points, and has the highest salary in the company, 10 points. Such an individual would therefore receive 1 100 410 points. All individuals receive points that correspond to their functions in the enterprise. Individuals who are just
employees (without being oldest or having the highest salary) receive 0 points. We then assume that the person with the highest points per company is that company s operational leader. Position Points CEO according to Companies Registration Office 1000000 CEO according to the Occupational Register >= 10 employees 100000 CEO according to the Occupational Register < 10 employees 10000 Partner (professional status) 1000 Vice President 800 Chairperson 700 Deputy chairperson 600 Executive board member 500 Board member 400 External authorised signatory 300 Legally responsible person 200 Alternate board member 100 Highest income in the company 10 Highest age in the company 1 Table 1. The function of the individuals in the companies with corresponding points If we have two partners in a partnership, the individual with the highest income is chosen as the operational leader in accordance with the decision-making procedure. If both have exactly the same earnings, we choose the individual who is the oldest. Every individual s total points score and source are reported in the Entrepreneurship Database - a kind of quality declaration. Nearly 6 per cent of the operational leaders are chosen based on the two lowest criteria, income and age and their businesses covers just under 4 per cent of the total number of gainfully employed persons in the business sector. 4. Family owned business 4.1 Definition A business is defined as family controlled if a) a family possess at least fifty per cent of the firm s decision making rights and b) at least one family member is involved in the governance of the firm. Listed firms fulfil the condition of a) if a family possess at least twenty-five per cent of the firm s decision making rights. 4.2 Data sources Our method of identifying family firms is based on several statistical registers. The register principally used is the Register Based Labor Market Statistics which includes information from the Business Register and the Population Register, among others.
This has been complemented with data from the Ownership Register the Financial Supervisory Authority s Central Registers of Investments and Investor Alerts, the Companies Registration Office s Limited liability Company Board Member Register, the Tax Authorities statistics of earnings and deductions and the Multiplegeneration Register. All information is observed over the period of 2004-2010, the first and last year where data is available. 4.3 Register-based Labour Market Statistics (RAMS) The Register-based Labour Market Statistics (RAMS) contains information on all Swedish residents and labor market participants, as measured by December 31, and all firms which are active in Sweden and have registered employees by November 30th. By using RAMS, we are able to identify all active Swedish firms and their employees. 4.4 The Ownership Register The Ownership Register contains information all holdings in listed firms in Sweden. The information is supplied by Euroclear Sweden AB which is the authorized central deposit for financial securities in Sweden which receives notifications of purchases and sales of stocks in listed firms in Sweden under the Law of Accounting of Financial Instruments (SFS1998:1479). The information supplied contains the name of the holder, the size of its holdings in terms of number of votes and share capital and the identity number of the holder. 4.5 The Financial Supervisory Authority s Central Register for Investments and Investor Alerts The Financial Supervisory Authority s Central Registers of Investments and Investor Alerts contain supplementary information to that of the Ownership Register. The registers contain, in addition to the latter, information on the ultimate owner(s) of any legal person (firm) which holds stocks in any listed firm in Sweden. A person is obliged to report control over a legal person (firm) if (s)he controls at least 10 per cent of the person s; a) share capital; b) votes or; c) cash flow rights (SFS 2000:1087 5). 4.6 The Companies Registration Office s Limited liability Company Board Member Register The Companies Registration Office s Limited Company Board Member Register contains information about all executive board members in limited liability firms in Sweden during a given year. The information contained in the Board Member Register is complete; all limited liability firms in Sweden must register an executive
board and give detail of the board s composition under the Law of Limited Liability Firms (SFS 2005:51). 4.7 Tax Authorities statistics of earnings and deductions The Tax Authorities statistics of earnings and deductions contains information on all taxable income during a given year. The information is complete for all firms, including both business owners and employees. 4.8 The Multiple-Generation Register The Multiple-generation Register contains compiled information from Swedish population registers where all individuals have been linked to their biological or adoptive parents. The Multiple-generation Register has complete coverage for individuals born 1932 or later who have been registered as Swedish citizens at some point since 1968. The register is updated annually in connection to the yearly publication of the Swedish Population Register. 4.9 Method Our method of identification is divided into three principal steps: 1) identifying the population containing family firms; 2) identifying families and; 3) identifying family firms. Identifying the population: Identification of the population of interest is achieved through method of elimination. Family firms can only take the form of limited liability firms, partnerships and sole proprietorships, by definition. Therefore, we delimit the population of potential family firms to firms within these legal forms. Identifying families: We apply a broad definition of the term family, referring to all individuals who are related through a common male ancestor. This is chosen, in oppose to the nuclear family, as many family firms are governed between relatives outside the nuclear family. Identification through a male ancestor in oppose of a female ancestor is made as most businesses have historically been governed by men. In practice, however, our method grants that the choice of strategy has little or no influence over the results. Identification of kinship and families is as such: First, all Swedish residents during a given year are selected from the Multiple-generation Register and their registered parents are identified (both biological and adoptive). Next, in a second step, all identified fathers are matched against the Multiple-generation Register in order to identify a given individual s grandparents. This makes it possible to identify family firms where up to three generations are active (children, parents and grandparents). Some family firms may, however, be composed of more distant relatives. Therefore,
in a third step, all grandfathers out of the registered grandparents are selected and matched against the Multiple-generation Register in order to identify their parents; and so the process continues. Once the process of identifying all known male ancestors is complete, the eldest known male ancestor is attributed as the forefather of all subsequent descendants. We then identify the families of all descendants (spouses, siblings and children) and attribute kinship between all identified individuals. By this, we capture governance and ownership structures involving, for example, siblings, parents, grandparents and cousins. At most, up to five generations can be identified using this method, although in practice we only identify families of up to three generations, about 98 per cent of the population. Fourth generation family members constitute approximately 2 per cent of the population, while fifth generation family members constitute less than.01 per cent of the population. Identifying family firms: Family firms are identified through use of the definitions and families identified above. As the definitions and data are different for listed and unlisted firms as well as between legal forms, the method of identifying each type is described separately. Identifying family firms among listed firms: The information contained in the Ownership Register has been matched against RAMS in order to acquire background information on all domestic owners. The method of identifying listed family firms is thereafter five-staged: 1) All holdings that are registered on a physical person are attributed to that person. Shares may have different decision-making rights, which our method takes into consideration. 2) All holdings that are registered on a legal person are attributed to the physical owner(s) of that legal person. 3) There are a handful of families exerting control of a substantial share of Swedish industry via family controlled foundations. These foundations are identified manually and all holdings of these foundations are attributed to the controlling family. 4) All substantial shareholdings6 without a known physical owner are examined manually through the use of the Swedish Financial Supervisory Authority s Central Register of Investments and Investor Alerts and the studied company s own financial reports and press statements. In the case where the process in 4) yields a positive identification of an ultimate, Swedish owner, then the holdings in question are attributed to that owner. All holdings which are not identified as privately held by Swedish residents are lastly denoted as foreign or government held. 5) In a final step, all identified Swedish holdings are consolidated to the families identified in the previous section. The families are then matched against the Swedish Companies 6 Substantial shareholdings are defined in listed firms as holdings which constitute at least 25 percent of the decision making rights within a given firm.
Registrations Office s Executive Board Register to confirm that at least one family member is on the firm s executive board or holds the position of managing director. All listed firms in which a family controls at least 25 per cent of the decision-making rights are then identified as family firms. Identifying family firms among non-listed limited liability firms: Non-listed limited liability firms can be divided into closely and widely held firms, where closely held firms are limited liability firms where at least 50 per cent of the firm s decision making rights are controlled by a maximum of four owners, and where an individual and its family are regarded as one owner (SFS 1999). All other limited liability firms are classified as widely held. Among closely held firms, owners are obliged to submit their business income reports through a special form, the so-called K-10 form. This form is reserved to individuals who both own and operate a closely held business. Each owner (including individual family member) has to submit a separate form. This has been used to identify owners in closely held firms, where we have classified all firms where at least fifty per cent of the submitted K-10 forms are from members of the same family as family firms. By this, we assume that ownership is approximately equally distributed among owners. Among widely held, non-listed limited liability firms and among closely held firms which have not filed for business income, we have no information of their ownership structure. Instead, we use the information contained in the Companies Registration Office s Limited liability Company Board Member Register in order to identify all board members in these firms. When assume that ownership and board presence is closely related and classify all firms where at least fifty per cent of the executive board members are part of the same family as family firms. Identifying family firms among sole proprietorships: Swedish corporate law dictates that sole proprietors are firms which are registered on their founders who exhibit complete control over the firm and who are completely responsible for its activities. Sole proprietorships are thereby, by default, family controlled and we therefore classify all of these firms as family firms. This method of identification has been condoned by the European Commission Expert Group on Family Business, among others. Identification family firms among partnerships: Joint and limited partnerships do, by definition, not have any stocks. We therefore classify them as family controlled if at least half of the involved partners are related. In the case that the partnership is a daughter company, it is classified according to the ownership of the parent company.
5. Family owned business A novel data base like this gives us an opportunity to study questions like: a) is it more likely that family firms are hiring female as operational business leader (OBL) than non-family firms? b) do family firms have a lower labour turnover than nonfamily firms? c) do family firms create more jobs than non-family firms? With business leaders and family firms we mean the definitions as in this paper. 5.1 Female business leader Firm population To answer these three questions we are focusing only at limited liability companies in the private sector between the years 2004 2010 and only companies that have people who are gainfully employed consecutive years. When we are investigating the probability of new business leaders we also restrict the population to only allow firms that have changed business leaders between consecutive years. A yearly average on firms that change business leaders is around 12 800. Less than 20 per cent of the business leaders are females. Results We use a logit regression method to estimate the parameters of our model. The dependent variable is binary, and takes the value zero if the new business leader is a male and one if it is a female. The descriptive statistics give us a picture that only 12.5 per cent of the non-family firms have a new business leader who is a female compared to 20.6 per cent of the family firms. These figures are on the other hand not representative since most of the family firms are small and most of the nonfamily firms are large. In order to come a bit closer to the truth we need to control for some obvious skewness in the background factors. In this first attempt to answer this question we use tentative model where we chose to control for variables such as: industry, firms sizes and year and what gender the former business leader had. We think it is more likely to hire a new business leader who is female if the former business leader was female. We also need control if the new business leader is related to the owner family. The econometric result tells us that it is more than 32 per cent higher probability (risk) that family firms hire females as business leader than non-family firms does when these fact as controlled for. The large difference between the probabilities of family firms and non-family firms depends highly on the size of the firms. As shown in Figure 1, for smaller firms, up to 49 employees we can see significant difference in the probabilities, but for the large firms there are no significant differences. As we hypothesized, there is a 10 per
cent higher probability to hire a new female OBL if the former OBL was a female. The results also tell us that there is just over 40 per cent higher probability to hire a female OBL if the new OBL is in some way related to the owner family. A notable result is that the larger the firms are the smaller is the probability that the new BL is a female. Female as new BL Probability.05.1.15.2.25 1 2-4 5-9 10-19 20-49 50-149 150-249 250-499 500- Size Non-family firms Family firms Source: SCB Figure 1. Predicted probability that firms' new BL is a female, split by size of firm 5.2 Labour turnover Firm population We have to do one modification of the firm population which regards the time horizon. In this case we only have data available between the years 2007-2010, which implies three cohorts to use in our analyse. The yearly average of firm population is around 131 000 firms. Results If only descriptive statistics such as the means for each group are studied we can see that family firms have on average for this time period a 4.9 per cent labour turnover meanwhile corresponding number for non-family firms is 8.2 per cent. We see a u- shaped relation between labour turnover and year. The ratio is lowest for the year 2009, which most likely is a result of the financial crises that started at September 2008 and affected the Swedish labour market the year after. Since earlier we know that family firms are in general smaller and also in other industries compared to non-family firms. When we have controlled for firms characteristics the econometric result from figure 2 tells us that family firms have instead a significantly larger labour turnover than non-family firms. The results hold for almost all sizes of firms (except for firms with 1-4 employees).
Labour turnover Percentage 0 5 10 15 20 1 2-4 5-9 10-19 20-49 50-149 150-249 250-499 500- Size Non-family firms Family firms Source: SCB Figure 2. Predicted percentage labour turnover in family and non-family firms, split by size of firm 5.3 Job creation rate The last question we look into is: Do family firms create more jobs than non-family firms? The absolute number of job creation is often a function of firm size. Larger firms create often more jobs compared to smaller firms and vice versa when there is a contraction in the economy. In order or to normalize the job creation by firms size we divide job creation by the number of employed people in period t-1. Job creation rate is defined as the change in gainfully employed people between period t-1 and t divided by the number of gainfully employed people in period t-1. This may be a number which is large in both directions, positive or negative. Once more we use an OLS regression method when we estimate our parameters from the model. Firm population We are not studying the whole firm population which implies that we do not analyze the economies total job creation. Once again we are only looking at limited liability companies in the private sector. An additional requirement is now that the firms are not allowed to be newly started, since we need to know the change in firms labour force. The yearly average of firm population is around 141 000 firms. Results Given our firm population we are able to see from Table 1 that family firms have created many more jobs than non-family firms. The difference for the period 2004 2010 is almost 110 000 jobs. From table 1 we can also see the effects of the latest financial crises. In the year of 2009 more than 80 000 jobs disappeared in total, more jobs disappeared in non-family firms than family firms; otherwise have these firms created somewhere between 15 000 60 000 jobs each year.
Job creation Year Non-family Family Total 2004 5815 17452 23267 2005-1822 15605 13783 2006 11498 32109 43607 2007 22580 38418 60998 2008 10217 16740 26957 2009-45519 -34557-80076 2010 10945 35126 46071 Total 13 714 120 893 134 607 Table 2: Total number of job creation, split by family and non-family firms Once again we chose a tentative model. We control for factors such as year and size, industry and then a dummy variable if the firm is family or a non-family firm. We chose to present two figures, job creation rate by size and job creation by year. These two different figures also represent two versions of the model, since in each of these models is the family variable interacted with either size or year. The regression results show us that family firms have in principle for all years and firm sizes a higher job creation rate compared to non-family firms. Job creation rate Job creation rate Job creation rate -.2 -.1 0.1.2 Job creation rate -.05 0.05.1 1 2-4 5-9 10-19 20-49 50-149 150-249 250-499 500- Size 2004 2005 2006 2007 2008 2009 2010 year Non-family firms Family firms Non-family firms Family firms Source: SCB Source: SCB Figure 3. Predicted job creation rate in family and non-family firms, split by size and year