Available via license: CC BY-NC 4.0
Content may be subject to copyright.
Volume 4, Number 1, 2017, 1–16 journal homepage: region.ersa.org
DOI: 10.18335/region.v4i1.95
Does the increase in house prices influence the creation
of business startups? The case of Sweden
Bj¨orn Berggren1, Andreas Fili2, Mats H˚akan Wilhelmsson3
1Royal Institute of Technology, Stockholm, Sweden (email: bjorn.berggren@abe.kth.se)
2Royal Institute of Technology, Stockholm, Sweden (email: andreas.fili@abe.kth.se)
3Royal Institute of Technology, Stockholm, Sweden (email: mats.wilhelmsson@abe.kth.se)
Received: 22 September 2015/Accepted: 22 November 2016
Abstract.
Entrepreneurs are at the core of economic development in that they start
new businesses or make existing firms grow. To fulfill this important role, entrepreneurs
need access to financing. Owing to information asymmetry and the relatively high risk
associated with business startups, many financiers shy away from engaging in relationships
with firms during the early stages of their development. Based on the existing body
of knowledge on the financing of entrepreneurship, we know that insider finance is
of paramount importance in the early stages of firms’ development. We expand this
knowledge base by analyzing the influence of house prices on business startups across
municipalities in Sweden. In our analysis, we include data from all municipalities in
Sweden. Our data on house prices and control variables are collected in period one, and
our data on the frequency of startups are collected in period two. We find that rising
house prices in a municipality lead to a higher frequency of startups. In our spatial
Durbin model, we find that a 1% increase in house prices leads to around 0.15% increase
in startups. Our findings are in line with the limited international research that has been
previously conducted, and therefore, our study might make a small but vital addition
to this growing body of knowledge within the area of entrepreneurship and regional
development.
JEL classification: R11, R31, M13
Key words: Business startups, entrepreneurship, financing, house prices, mortgages
1 Introduction
In the late 1970s, scholars concluded that small and medium-sized enterprises (SMEs)
create the majority of new jobs in the U.S. economy (Birch 1979). These findings
spurred great interest in research on the employment contribution of SMEs worldwide,
with scholars concluding that SMEs contribute to 70% to 90% of all new jobs that are
created (Davidsson et al. 1995,Armington, Acs 2002,Santarelli, Tran 2012). Previous
research has shown that in addition to playing a vital role in the creation of employment
opportunity, startups and SMEs are involved in creating industrial renewal, export income,
and innovation (Halilem et al. 2012,Agostini et al. 2015,Love, Roper 2015), as well as
acting as a dynamic influence to lagging areas (Keeble 1997,Gordon, McCann 2005,Doh,
1
2 B. Berggren, A. Fili, M. H. Wilhelmsson
Kim 2014)(Keeble 1997, Gordon, McCann 2005, Doh, Kim 2014). Therefore, governments
in most nations have developed different types of programs to support SMEs and startups
(Bateman 2000,Perren, Jennings 2005).
To fulfill the role of creators of new employment, export income, and innovations, new
firms must secure access to financial resources (Harding, Cowling 2006,Atherton 2012).
Previous research has shown that the most important source of finance for newly started
firms is insider finance (Cassar 2004,Gregory et al. 2005,Robb, Robinson 2014). Insider
finance includes the personal funds of the founder: personal savings, home mortgage, and
credit cards (Storey, Greene 2013). This implies that a booming housing market would
enhance the ability of entrepreneurs to increase their home mortgage, or at least enhance
their ability to take on new loans by virtue of the increase in collateral they can offer the
bank, and, thus, finance their newly started firms – a causality that has been established
by Jin et al. (2012).
In this paper, we build on the findings of Jin et al. (2012) and use Sweden as our
empirical case. More specifically, we attempt to identify and estimate an empirical
model of the relationship between business startups and house prices in all 290 Swedish
municipalities.
The remainder of the paper is organized as follows: In section 2, we present some
key findings from previous studies on the financing of entrepreneurial ventures and the
relationship between house prices and business startups. Next, we present the empirical
model in Section 3and describe our data in Section 4. We then present our results in
Section 5and conclude the paper with a discussion of our findings in Section 6.
2 Theoretical points of departure
Financing entrepreneurship and business startups have received a great deal of attention
from researchers and policymakers for over 100 years (see MacMillan 1931). According to
conventional wisdom, small and medium-sized firms have problems accessing finance at
reasonable terms (see Storey, Greene 2013). Whether SMEs are, in fact, subject to credit
rationing is a question that has been asked in numerous research studies but is one that
remains unanswered. A consensus, however, has been reached regarding the dependency
of entrepreneurial ventures on insider finance in their earlier stages of development as well
as regarding the higher degree of financial constraints experienced by SMEs in comparison
to larger organizations (Cassar 2004,Revest, Sapio 2012). As a consequence, there has
been considerable research interest in the financing of new firms. Three major theories
have been used to illustrate financing patterns: life-cycle theory, pecking order theory,
and agency theory (Berger, Udell 1998,Johnsen, McMahon 2005). The first two theories
focus on the perspective of the firm that receives financing, whereas agency theory takes
the perspective of the investor that provides financing.
2.1 Different perspectives on the financing of entrepreneurial ventures
Life-cycle theory suggests that firms, depending on what stage of development they have
reached, follow similar financing patterns (Weston, Brigham 1981). Research has revealed
that firms in certain stages of development seek certain types of financing and that firms
have similar financing needs and financing behavior – no matter the cultural differences
across countries (cf. Psillaki, Daskalakis 2009). Studies from all over the world – including
Europe (Psillaki, Daskalakis 2009), China (Newman et al. 2012), and Africa (Abor, Biekpe
2009) – have provided support for this perspective.
Pecking order theory is concerned with explaining why firms do not always prefer
the source of financing with the lowest interest rate. Research has shown that there
seems to be a stable preference order – a pecking order – whereby different sources of
financing are ranked (Donaldson 1984,Myers 1984). In essence, the theory states that
there is a general mistrust of outsiders: the more a firm is likely to lose control to external
financiers, the less likely it is to submit to that type of financing. Internally generated
funds are preferred to bank loans, which in turn are preferred to new equity. Although
initial studies were conducted in large companies, several studies have shown that the
pecking order framework is a fruitful approach to studying financial decision-making in
REGION : Volume 4, Number 1, 2017
B. Berggren, A. Fili, M. H. Wilhelmsson 3
small firms as well (Vanacker, Manigart 2010,Degryse et al. 2012,Alon, Rottig 2013) and
that SMEs follow a financial pecking order when they seek external financing (Berggren
et al. 2000). Like life-cycle theory, pecking order theory is a well-established theory, and
numerous studies have found support for it (cf. Davidsson et al. 2009,Mac an Bhaird,
Lucey 2011).
Whereas both life-cycle theory and pecking order theory focus on the perspective
of the firm, agency theory takes the perspective of the investor. Agency theory models
entrepreneurial finance in terms of contracts between a principal and an agent, where the
goals of the two parties diverge (Jensen, Meckling 1976). Thus, as soon as the investor
has supplied funding, the firm will try to use those resources for personal gains instead
of for the benefit of the investor. The fact that these goals differ implies that once the
funding has been supplied, the investor needs to ensure – through monitoring and control
– that the funds are used properly (Ross 1973).
2.2 Implications of the financial perspectives for entrepreneurial finance
The major explanatory construct in pecking order theory is the notion of control aversion.
The pecking order theory predicts that banks will be preferred to new shareholders.
However, this notion is also quite clearly linked to life-cycle theory, in that control aversion
is especially prevalent among young firms: overcoming control aversion is partly a matter
of reaching maturity in dealing with business associates, financiers, and presumptive
owners. Because control aversion is prevalent among young firms, they will often act
in accordance with pecking order theory by contacting banks, rather than new owners
(Howorth 2001,Paul et al. 2007).
The bank, in line with agency theory, will also demand some measures of control,
primarily through collateral provided personally by the founder(s). Thus, the very
foundational assumption of agency theory leads to a strong focus on control, which was
initially the reason for the firm’s decision not to seek other sources of financing. The lazy
bank hypothesis states that collateral is not an effective measure against bankruptcy, but
merely an easy way of handling SMEs (Manove et al. 2001). Still, as the bank retains the
right to cancel a loan at any time, collateral, performance data and legitimacy represent
significant obstacles to new enterprises (Bracke et al. 2013,De Clercq et al. 2013,Ramlall
2014). Because of the pivotal role of collateral, the size of the collateral also proves to be
important. Being able to offer a large portion of valuable collateral to the bank should
mean more access to financing.
Most young firms have nominal collateral and credibility to offer the bank: they
are not as transparent as older, larger firms (Robb, Robinson 2014). Moreover, service
firms have fewer tangible assets (machinery and inventory) to offer as collateral for loans
than do manufacturing firms. Instead, collateral is found in the personal property of the
founder and, therefore, main sources of collateral are the houses of the founders’ (Chaney
et al. 2012).
2.3 A model of higher house prices leading to an increase in business starts
Previous research shows that collateral is important for entrepreneurial activity. By
using variations in house prices, Schmalz et al. (2013) provide evidence that in regions
with house price appreciation homeowners are more likely to start a business; and the
firms started by homeowners are larger than those started by renters. That is, collateral
matters. Furthermore, Adelino et al. (2015) show, based on aggregated county data for
the period 1998-2010, that regions with larger rises in house prices experienced stronger
growth in employment in small firms, especially in industries with a limited need for
capital.
The model is explained theoretically in the following way. First, higher house prices
mean that nascent entrepreneurs have more collateral to offer to the bank when they
apply for a loan to start the business (cf. Bernanke et al. 1999,Greenspan, Kennedy
2008,Jin et al. 2012,Bracke et al. 2013). This implies that banks can grant more loans
to small businesses based on the fact that there is more private collateral on the part
of the owner-founder of the firm. The model is part of an argument that claims that
REGION : Volume 4, Number 1, 2017
4 B. Berggren, A. Fili, M. H. Wilhelmsson
entrepreneurs suffer from a lack of funding (cf. MacMillan 1931,Stiglitz, Weiss 1981) and
that sudden financial gains – “windfall gains” – have a positive effect on the number of
business starts (Sch¨afer et al. 2011).
A number of papers lend support to this model. In their examination of the number
of value added tax (VAT) registrations in the United Kingdom as well as other kinds of
aggregate data, Black et al. (1996) found evidence that collateral availability does have a
strong influence on firm formation and dissolution. By contrast, Hurst, Lusardi (2004)
found no support for this hypothesis except in the very richest segment of households.
Thus, they conclude that there is no general link between collateral and business starts.
Robson (1996) compares his results explicitly with those of Black et al. (1996) from the
same year and also in the United Kingdom. Although Robson (1996) finds no support for
a link between high house prices and an increase in business starts, he states that housing
wealth does appear to have a positive effect on entrepreneurship, in that it helps reduce
the regional rate of deregistration from VAT.
This implies that an increase in house prices would equal an increase in the pool of
liquidity available to entrepreneurs, which translates into a larger number of business
starts in a certain region. According to this logic, a causal link may exist between a rise
in house prices and the number of businesses started in a region as a result of greater
access to collateral. Thus, our null hypothesis is formulated as follows:
There is a positive relationship between rising house prices and the number of
business starts in a region.
3 Empirical model
To test our hypothesis that housing market conditions play an important role in explaining
new firm formation, we estimate a model where the number of new firms per capita for
the period 2007-2014 across Sweden is related to a number of determinants. Of main
interest here is the relationship between new firm formation and housing price growth.
Two major problems may arise in this type of regional model. The first problem
involves the issue of endogenous determinants. Because house price appreciation facilitates
lending and, ultimately, increases entrepreneurial activity and economic growth, higher
economic activity might result in more lending and higher house prices. That is to say,
the relationship between house price appreciation and formation of new business startups
is bidirectional (Iacoviello 2005,Adelino et al. 2015). The second problem involves the
issue of spatial dependence. Both problems are connected to the question of how to
interpret the estimated relationships as causality and not merely as correlations. That is,
the empirical challenge is to identify the causal direction of the house price growth effect.
We avoid these potential problems by estimating a model using determinants in the
preceding period (2007) when explaining the variation in firm formation in the subsequent
period (2007-2014). In contrast to Binet, Facchini (2015), we estimate three types of
spatial autoregressive models, a spatial lagged model and a spatial error model, as a
means to control spatial dependence. We also estimate a spatial Durbin model in order
to analyze if it could be simplified to a spatial error or lag model. Five different spatial
weight matrices are tested. Two nearest neighbor, two inverse distance based and, finally,
a spatial contiguity matrix. The distance is estimated using the centroid coordinates of
the labor market.
Our approach has been used recently by Andersson et al. (2014), and earlier by
Armington, Acs (2002). Our identification strategy is to relate the change in the number
of startups in a region with the change and level of house prices in the previous period.
That is, lagged house prices can have an effect on startups in a later period, but it is
unlikely that startups in the future have an effect on house prices in a previous period (cf.
Balasubramanyan, Coulson 2013).
New firm formation per capita varies to a great extent across municipalities in Sweden
for the period 2007-2014. We use five different types of determinants to explain this
variation, namely, (a) establishment structure, (b) labor market conditions, (c) human
capital, (d) income, and (e) housing market conditions.
REGION : Volume 4, Number 1, 2017
B. Berggren, A. Fili, M. H. Wilhelmsson 5
We use establishment size, measured as 2007 employment divided by the number
of establishments in 2007, as a proxy for the structure of the industry in the region
(see Armington, Acs 2002). The expected relationship is negative, as larger average
establishment size should be negatively related to regional startups rates (Armington,
Acs 2002).
Different measures are used here to measure labor market conditions. Both measures,
number of employees in 2007 and unemployment rate in 2007, are included in the model.
Number of employees measures the size of the market and is used instead of population
(Binet, Facchini 2015) or number of households (Adelino et al. 2015) as a measure of the
market size. It is a measure that captures some of the agglomeration effects observed
in the literature (see Acs et al. 1994). One of the most important determinants used in
previous research is the unemployment rate (see Armington, Acs 2002) as it was suggested
that unemployed workers were more likely to start new firms. The unemployment rate
is also used in more recent research (Adelino et al. 2015,Binet, Facchini 2015). The
third labor market indicator we use is the self-employment rate in 2007. This measure of
entrepreneurial culture (Johannisson 1984), has also been used in research by Armington,
Acs (2002).
We also include a variable measuring the percentage of the population that was born
outside of Sweden in 2007. Lee et al. (2004) used this diversity index, which they labeled
the “Melting Pot Index.” The hypothesis is that a positive relationship exists between this
index and new firm formation. The argument is that “immigrants lack skills, resources,
and networks” and, therefore, tend to be self-employed and to start new companies to
a greater extent than nonimmigrants. Migration data were also used by Adelino et al.
(2015).
As a measure of human capital, we use the rate of university degree completion in
the population in 2007. This measure is a proxy for the level of skill and knowledge in
the regional economy (see Armington, Acs 2002, and, more recently, Adelino et al. 2015,
Binet, Facchini 2015). The relationship between the rate of university degree completion
in the population and new firm formation is expected to be positive.
Income is measured as the average annual regional income level in 2007 and is
hypothesized to have a positive relationship with new firm formation in the subsequent
2007-2014 period. Based on the argument of Binet, Facchini (2015) that a high regional
income level broadens the market size and, therefore, increases the number of opportunities
for new firms, we would expect to observe more business startups in regions with higher
regional income levels. We have also tested the change in annual income, but the empirical
results suggest that it is not related to the number of new firms per capita. The level of
income seems to be more important.
The housing market is included with two different measures. The first variable
measures the annual growth in the seven years preceding 2007, that is, 2000-2006. The
same type of measure is used by Schmalz et al. (2013) even though they use individual
data. We also include a measure indicating whether the regional house price level in 2007
is above the average house price level.
4 Data
We use data on startups in Sweden from the period 2007 to 2014. The data are aggregated
and based on all 290 municipalities in Sweden. The dependent variable is the change in
the number of startups for the period 2007-2014. The independent variables all measured
in 2007 are: human capital measured as the proportion with a university degree, income,
employment, unemployment, and accessibility in the municipality as well as the change
in house prices in the seven years preceding 2007 (2001-2007) and the house price level in
2007. Some descriptive statistics of the data are shown in Table 1.
In Table 1we see that the average number of new firms per capita for the period is
0.05 (standard deviation: 0.01), which is equal to 5 new firms per 100 inhabitants. Almost
23% of the population has a university degree; the variation across the labor market is,
however, substantial. House price appreciation is measured for the period 2001-2007. The
average house price change is positive and equal to almost 0.7% (standard deviation: 0.3).
REGION : Volume 4, Number 1, 2017
6 B. Berggren, A. Fili, M. H. Wilhelmsson
Table 1: Descriptive statistics of the data
Variable Abbre- Period Average Standard
viation deviation
New firms per capita New 2007–2014 0.0505 0.0142
University degree (%) Univ 2007 0.229 0.078
High house prices (%) High-Hp 2007 0.33 0.47
Change house price (%) Dhp 2001–2007 0.6926 0.3216
Self-employment (%) Self 2007 9.7391 2.4726
Employment Emp 2007 15,203 30,722
Establishment size Estab 2007 8.4618 1.7561
Income (SEK 000) Inc 2007 210.426 18.208
Unemployment (%) Unemp 2007 6.6398 1.9628
Immigration outside EU/EFTA (%) Inm 2007 4.6474 3.3232
Stockholm (%) Sthlm 2007 0.1211 0.3268
Source: EFTA, European Free Trade Association; EU, European Union.
Table 2: Correlation matrix
New Univ High-hp Dhp Self Emp Estab Inc Unemp Inm
New 1
Univ 0.63* 1
High-Hp 0.71* 0.74* 1
Dhp 0.62* 0.51* 0.59* 1
Self 0.28* -0.23* -0.12* 0.03 1
Emp 0.49* 0.74* 0.61* 0.48* -0.42* 1
Estab -0.11 0.34* 0.26* 0.14* -0.85* 0.45* 1
Inc 0.34* 0.66* 0.66* 0.62* -0.42* 0.57* 0.63* 1
Unemp -0.32* -0.26* -0.35* -0.37* -0.17* -0.16* -0.09* -0.50* 1
Inm 0.22* 0.38* 0.41* 0.29* -0.47* 0.30* 0.47* 0.39* -0.25* 1
Note: * Statistical significance on a 5% level.
Around one third of households in the labor markets have house prices that are higher
than the average house prices. The average size of the labor markets, measured as the
number of employees, is only 15,000 persons but the variation is considerable (standard
deviation: 30,000 persons). The average unemployment rate is 6.6% with a variation of
2%. We measure the business set up in terms of establishment size. The average number
of employees per establishment is equal to 8 persons (standard deviation: 1.7 persons).
The entrepreneurial climate is measured with the self-employed variable. The average
rate of self-employed is 9.7% with a variation of 2.5%. The number of immigrants as a
percentage of the population is 4.6%, but the variation is substantial across the labor
markets. Around 12% of the population lives in the labor market of Stockholm (the
capital of Sweden). The correlation coefficients are shown in Table 2.
The correlations between the dependent variable (New) and the independent variables
are strong in most cases. The highest correlation among the dependent variable and
an independent variable is between new firms and high house prices, indicating the
importance of house prices as a channel of financing for startups. However, we can also
observe high correlations between high house prices and high proportion with a university
degree and between high house prices and high levels of income, with university degree
and income being positively correlated with startups. The house price appreciation for
2001-2007 is positively correlated with new firm formation for 2007-2014. We also notice
that the variable ‘employees per establishment’ and the self-employed variable are highly
negatively correlated, indicating that it can be difficult to differentiate the effects in the
empirical model. All correlations are statistically significant on a 5% level. In Section 5,
we present the results from our empirical model in which we relate new firm formation
REGION : Volume 4, Number 1, 2017
B. Berggren, A. Fili, M. H. Wilhelmsson 7
Source:Arena for Growth (2015)
Figure 1: Average number of startups per 1000 inhabitants in Sweden 2011-2014
per capita to all the determinants presented here.
4.1 Entrepreneurship in Sweden
Before we present and analyze the findings of our empirical model, we will present some of
the key characteristics of startups within the Swedish economy. The number of startups
in Sweden has increased over the past 20 years. Between 1994 and 2003, there were
34,000-39,000 startups per year, but in 2014 that number had increased to more than
71,000 startups per year (Statistics Sweden 2015). Two of the major reasons for the
increased number of startups are simplified rules for incorporating a business and reduced
capital requirements for limited liability companies. The number of bankruptcies has
been relatively stable over the past ten years and in 2014 6,000 Swedish firms filed for
bankruptcy (Statistics Sweden 2015).
In comparison with other European countries, the number of nascent entrepreneurs in
Sweden is relatively low (GEM 2012). One reason for the relatively low levels of startups
in the Swedish economy, in comparison with other nations, is lack of entrepreneurial spirit
owing to a tradition of large enterprises within the Swedish economy. Instead it seems as
though most entrepreneurial skills are distributed among established firms, a phenomenon
that has been labeled intrapreneurship (GEM 2012). Even though relatively few firms
are started each year in Sweden, the firms that do start have a higher survival rate than
firms started in comparable countries (Andersson, Klepper 2013). Regarding industries,
most startups are within the retailing and services industries. In 2014, more than 80% of
all new firms were started in these two industries.
In Sweden, there are relatively large regional differences in startup frequency, see
Figure 1. Among the hotspots for startups, as well as being the most dynamic regions, are
the three metropolitan areas of Stockholm, Gothenburg and Malmo. We can also find some
examples of municipalities outside these regions with relatively high frequencies of startups.
Among these are municipalities on the border to Norway, as well as various regional
centers where universities, hospitals, and other governmental agencies and institutions
are located.
REGION : Volume 4, Number 1, 2017
8 B. Berggren, A. Fili, M. H. Wilhelmsson
5 Empirical results
Our main hypothesis is that the growth in house prices play a major role in explaining
subsequent variation in new firms per capita. However, we also test the hypothesis that
business set up, labor market conditions, human capital, and entrepreneurial climate play
equally important roles when it comes to new firm formation. We test these hypotheses
using a regression model where the dependent variable is the number of new firms per
capita for the period 2007-2014. The independent variables used are those discussed in
the previous section.
Tables 3-5show the results from the ordinary least squares (OLS) estimation as well
as from three types of spatial econometric models: the SAR-model (spatial autoregressive
model), the SEM-model (spatial error model), and the spatial Durbin model (SDM).
For model comparison and selection of weight matrix we are following a specific-
to-general test procedure proposed by Elhorst (2010). First we estimate an OLS and
thereafter, using LM-tests (Anselin 1988,Anselin et al. 1996), testing for spatial depen-
dency. If the LM-tests on OLS-residuals are significant, then SDM is estimated. If these
LM-tests suggest that the SEM is the best spatial model, the log likelihood ratio test
(LR-test) is used in order to test the convenience of SDM against SEM and SAR.
The next step is to select the spatial weight matrix. We are using three different types
of weight matrices: nearest neighbor-based, distance-based and contiguity-based spatial
matrix (discussed in for example Chasco 2013). All of them have been used in the spatial
econometric literature (Elhorst 2010). Our selection of weight matrix is based on mobility
pattern between municipalities. For example, contiguity and inverse distance have been
used in Mendiola et al. (2015). We are using 2 and 10 nearest neighbor and 50 and 100
kilometers cut-off for the inverse distance matrix. The choice is somewhat arbitrary
1
.
However, as Elhorst (2010) says, the wrong choice of the spatial weight matrix can distort
the estimates, but “the probability that this really happens is small if spatial dependence
is strong” (Elhorst 2010). LeSage, Pace (2014) call the belief that the estimates are
sensitive for the choice of spatial weight matrix “the biggest myth in spatial econometrics”.
LeSage, Pace (2009), Stakhovych, Bijmolt (2009), and Halleck Vega, Elhorst (2013) are all
in favor of using goodness-of-fit measures to discriminate among different spatial matrix
specifications as there are no clear theoretical reasons for any specific form. We are using
the most widely used log-likelihood value in order to differentiate between spatial weight
matrices (Elhorst 2010). In order to test for the robustness of our coefficient estimates, we
are analyzing the coefficients in the final spatial model specification using all the different
spatial weight matrices.
As stated earlier, we are estimating the spatial Durbin model in order to test the
hypothesis if this more general specification can be simplified with a spatial error (SEM)
and/or autoregressive (lag) model (SAR). Here we are using the LR-test.
Table 3shows the result from OLS. We are also testing for normality (Jarque-Bera),
heteroscedasticity (Breusch-Pagan/Cook-Weisberg) and multicollinearity (Variance-of-
inflation, VIF).
Around 83% of the variation in the total number of startups per capita between 2007
and 2014 can be explained by business set up, labor market conditions, and entrepreneurial
climate, as well as by income and house price appreciation. The R
2
value is considerably
higher than that reported by Lee et al. (2004), for example, but is of the same magnitude
as reported by Armington, Acs (2002). All coefficients have the expected signs and
are statistically significant on a 95% level. The t ratios are calculated using White
heteroscedasticity-robust standard deviations as the Breusch-Pagan/Cook-Weisberg test
shows presence of heteroscedasticity. The Jarque-Bera test shows that residuals are
normally distributed, that is, Maximum Likelihood (ML) is a suitable estimation method
of the spatial error and spatial lag models. See also Figure 2where the kernel density
estimate is compared to the normal distribution. In the OLS model, we also present
the variance of influence (VIF) so as to analyze potential multicollinearity issues. High
correlation among the independent variables does not seem to create a problem of
1
We have also tested different numbers of nearest neighbors and different cut-offs, but that did not
change the results and the overall conclusion.
REGION : Volume 4, Number 1, 2017
B. Berggren, A. Fili, M. H. Wilhelmsson 9
Table 3: Regression analysis (OLS-results)
Model 1 (OLS)
Coefficient t value VIF
High house prices 0.1244 6.22 3.04
Change in house prices 0.1944 7.04 2.01
Human capital 0.1795 3.64 4.54
Establishments -0.1809 -2.51 7.06
Self-employed 0.3136 5.71 6.94
Employment 0.0271 2.01 3.71
Income 0.4419 2.5 4.96
Unemployment 0.0978 3.08 2.28
In-migration 0.0217 1.44 2.25
Stockholm 0.1577 6.87 1.9
Constant -6.0839 -5.97
R20.832
R2-adjusted 0.826
Jarque-Bera test (Prob >z) 0.145
Breusch-Pagan/Cook-Weisberg test 0.007
(Prob >chi2)
Observations 289
Notes: OLS, ordinary least squares; VIF, variance of influence. White heteroscedasticity-robust standard
deviation.
Table 4: Diagnostic tests (statistics and p-values)
Nearest neighbor Inverse-distance Contiguity
Tests 2 10 50 km 200 km
cut-off cut-off
Spatial Error:
Moran’s I 4.985 9.244 5.155 5.155 7.385
(0.000) (0.000) (0.000) (0.000) (0.000)
LM 21.379 62.846 39.584 29.281 44.261
(0.000) (0.000) (0.000) (0.000) (0.000)
Spatial lag:
LM 25.306 39.813 44.176 9.776 33.167
(0.000) (0.000) (0.000) (0.000) (0.000)
multicollinearity as VIF values are below 10.
Testing for spatial dependence (Moran’s I and the LM-tests with different spatial
weight matrices) reveals that we do have a problem with spatial autocorrelation and/or
spatial heteroscedasticity (Wilhelmsson 2002). Table 4shows the results of the diagnostic
tests for spatial dependence in OLS regression. Five different weight matrices are tested:
two nearest neighbor (2 and 10), different inverse-distances based matrices with different
cut-offs (50 and 200 kilometers) and one spatial contiguity based matrix.
At least two conclusions can be drawn from the diagnostic tests. First, spatial
dependence is present. All LM-tests are significant which indicates the presence of spatial
dependence. Second, the diagnostic tests are in favor of the spatial error model as the
LM-test concerning the spatial error model has a higher value compared to the value
concerning the spatial lag model.
We continue our analysis to estimate a spatial Durbin model with all the different
spatial weight matrices. We do that in order to test if the more general spatial Durbin
model is preferred compared to the spatial error model and the spatial lag model. The
test is carried out with a LR-test. The test statistics (33.24 and 46.25) are all higher than
the critical value (18.31), which indicates that the hypothesis can be rejected. That is
REGION : Volume 4, Number 1, 2017
10 B. Berggren, A. Fili, M. H. Wilhelmsson
Figure 2: Kernel density estimate
to say, the spatial Durbin model describes the data best
2
. The results from the spatial
Durbin model are presented in Table 5. The spatial weight matrix with the highest log
likelihood is the contiguity matrix (see Table 6), and consequently, only the results from
this specification are presented in Table 5below. We have also estimated two diagnostic
statistics concerning spatial autocorrelation in the residuals from the spatial Durbin
model, namely LM-test and Moran’s I. The Moran’s I and LM-test is based on Anselin
(2005) definition. Both of them show no indication of spatial dependency in the residuals.
By interpreting the coefficients we have the following results. If the proportion of the
population with a university degree increases, the number of startups increases. That
is, human capital is important as, for example, Armington, Acs (2002) and Lee et al.
(2004) have shown. The same is true for an increase in the variables number of employees,
income level, as well as unemployment rate. Our results are consistent with the findings
of Armington, Acs (2002) and, more recently, Adelino et al. (2015) and Binet, Facchini
(2015). The spillover effect (indirect effect) concerning human capital is negative indicating
that lower human capital in neighboring municipalities is associated with fewer startups.
However, the indirect effect concerning income and unemployment is positive, indicating
spillover effect.
However, if the ratio between employment and establishment increases (the proxy
for business set up), then the number of startups decreases. That is, regions with many
SMEs are more likely to foster new startups. Our results concerning the Swedish market
can confirm the results of Audretsch, Fritsch (1994) and those of Armington, Acs (2002),
among others.
We can also observe that the number of self-employed persons in a labor market in
2007 leads to an increase in the number of startups per capita in the next seven years.
Hence, the entrepreneurial climate seems to have an effect. The Melting Pot Index
indicated that diversity is not significant in all models. Our results support the findings
of Lee et al. (2004).
We can also notice a Stockholm effect. Being the capital, the Stockholm labor market
fosters more startups per capita than the rest of Sweden. However, this effect alone
cannot explain our findings. In fact, even if we exclude Stockholm, we can observe more
or less the same results.
Housing market conditions represent the key determinant here. Two variables are
used to measure housing prices. The first variable measures house price levels. It is a
binary variable indicating whether the specific labor market has a price level that is above
the national average house price level. The second variable measures house price growth.
Both coefficients concerning housing market conditions are statistically significant and
positive, indicating that a positive relationship exists between new firm formation and
2
We have also tested SDM against SAR and the same conclusion can be drawn, that is the spatial
Durbin model describes the data best.
REGION : Volume 4, Number 1, 2017
B. Berggren, A. Fili, M. H. Wilhelmsson 11
Table 5: Spatial Durbin Model
Coefficients Direct Indirect Total
effect
High house prices 0.0839 0.0885 0.0814 0.1699
(4.25) (4.40) (1.33) (2.55)
Change in house prices 0.1456 0.1591 0.2305 0.3896
(5.30) (5.77) (2.84) (4.59)
Human capital 0.2487 0.2242 -0.4373 -0.2131
(6.40) (5.92) (-3.63) (-1.66)
Establishments -0.1683 -0.2882 -0.4229 -0.7111
(-2.39) (-3.98) (-1.70) (-2.68)
Self-employed 0.2111 0.1916 -0.3315 -0.1399
(3.54) (3.12) (-1.51) (-0.58)
Employment 0.0216 0.0248 0.0577 0.0825
(2.02) (2.32) (1.67) (2.22)
Income 0.2346 0.3191 1.6344 1.9535
(1.55) (2.09) (3.44) (3.76)
Unemployment 0.0708 0.0597 0.1805 0.2402
(2.37) (1.76) (2.21) (2.93)
In-migration 0.0226 0.0143 -0.031 -0.0164
(1.61) (1.00) (-0.84) (-0.42)
Stockholm 0.1891 0.1842 -0.0775 0.1068
(3.91) (3.87) (-1.09) (2.02)
Rho 0.422
(5.52)
Constant -7.7635
(-4.52)
Log-likelihood 405.449
R2-adjusted 0.8484
LR-test statistics (SDM vs SEM) 33.24
LR-test statistics (SDM vs SAR) 46.25
LM test for spatial autocorrelation 0.8898
(Prob >z)
Moran’s I (Prob >z) 0.7492
Note:t-values within brackets
both house price appreciation and house price level, respectively. Our finding supports
the results of Adelino et al. (2015) and of Balasubramanyan, Coulson (2013).
Hence, the change in startups per capita for the period 2007-2014 can be explained
by determinants that measure either the situation in 2007 or the change between 2001
and 2007. We argue that this is a causal relationship and not merely a correlation. If
the change in house prices increases by 1%, the expected change in startups is around
0.15%. We also observe that the number of startups per capita is higher if the house
prices are above average in 2007, indicating that both the level and the change in house
prices are of importance. If we consider the direct impacts, we can observe that these
are close to the spatial Durbin model coefficients, that is, the results indicate that the
feedback effects are very small and of no economic significance. Spatial spillover measured
by the indirect effect is positive and statistically significant. One interpretation could
be that this positive spillover effect reflects how changes in house prices in all regions
would impact startups in their own region. The spillover effect may be a result from
expectations of future house prices. The indirect effect concerning the variable high house
prices is, however, not statistically significant, indicating no regional spillover.
In order to test the robustness of our estimated coefficients concerning high house
prices and the change in house prices we have also estimated the spatial Durbin model
REGION : Volume 4, Number 1, 2017
12 B. Berggren, A. Fili, M. H. Wilhelmsson
Table 6: Spatial Durbin Model with different spatial weight matrices
Spatial weight Change in High house Log-likeli-
matrix house price price hood
Nearest neighbor 2 0.1556 0.0990 386.497
Nearest neighbor 10 0.1560 0.0817 399.016
Inverse distance 50 0.1465 0.0923 398.463
Inverse distance 200 0.2015 0.1268 390.913
Continuity 0.1456 0.0839 405.449
using different spatial weight matrices. The results are presented in Table 6. As observed,
the difference in estimated parameters is very small, that is our results are robust when
it comes to choice of spatial weight matrix. For example, the coefficient concerning the
variable high house prices ranges from 0.0817 to 0.1268 and the coefficient concerning
change in house price ranges from 0.1465 to 0.2015. Most of the estimates are of the same
magnitude. The exception is the spatial weight matrix based on inverse distance with a
cut-offs of 200 kilometers.
6 Conclusion and discussion
A number of interesting issues are highlighted in Section 5. For instance, most en-
trepreneurship theorists would agree that at the margin, there are nascent or potential
entrepreneurs who lack access to finance. There also exist entrepreneurs at the margin
who would remain self-employed longer if they had access to finance. However, assuming
that people who suddenly receive money would generally use some or all of this money to
start businesses would be an oversimplified view of the world. If potential entrepreneurs
do not perceive an opportunity, or if they do not possess the unique capabilities necessary
for exploiting a perceived opportunity, giving them money in itself is not enough (cf.
Shane 2000).
In the present study, we have provided evidence that higher house prices – at an
aggregate level – lead to an increase in business starts. A major contribution of our
analysis lies in our modeling approach. We control for both endogeneity and spatial
dependence of the entire population of Swedish municipalities. First, by separating
observations in time, where the observation of house prices is in period n and the
observation of entrepreneurship is in period n+1, we control for endogeneity and posit
a causal relationship where higher house prices lead to an increase in entrepreneurial
activity. Second, by employing a spatial Durbin model, we control for spatial dependence.
We contribute to current theory by providing evidence in support of studies where house
prices impact entrepreneurship. For policymakers, these results underline the paramount
importance of the public sector’s capacity for urban planning and the need for efficient
processes in the institutional framework regulating the housing sector.
In the future, we intend to conduct a more specific analysis of how increases in house
prices affect start-up frequencies in different sectors and regions. Today, the most common
type of start-up in Sweden is a service firm with other firms as customers, but we expect
to see regional differences in terms of the types of firms that are started.
One potential limitation of our paper is that it is based on an analysis of a single
national case, in our case Sweden. Sweden has a bank-oriented economy (along with
other countries such as Japan and Germany), whereas the United States and the United
Kingdom are examples of market-oriented economies (cf. Mayer 1988). It is difficult to
ascertain to what extent and in what way this orientation affects individual transactions
and relations, and while there surely exist different traditions in different countries in terms
of entrepreneurial activity, one could also make the opposite argument that some aspects
of economic activity are increasingly global in nature and not very different between
industrialized nations today. However, the difference in orientation has historically
pervaded all economic activity. Therefore, future research in this field should endeavor to
compare results between bank- and market-oriented economies.
REGION : Volume 4, Number 1, 2017
B. Berggren, A. Fili, M. H. Wilhelmsson 13
References
Abor J, Biekpe N (2009) How do we explain the capital structure of smes in sub-Saharan
Africa?: Evidence from Ghana. Journal of Economic Studies 36: 83–97. CrossRef.
Acs ZJ, Audretsch DB, Feldman MP (1994) R&D spillovers and recipient firm size. Review
of Economics and Statistics 76: 336–340
Adelino M, Schoar A, Severino F (2015) House prices, collateral, and self-employment.
Journal of Financial Economics 117: 288–306. CrossRef.
Agostini L, Caviggioli F, Filippini R, Nosella A (2015) Does patenting influence SME sales
performance? A quantity and quality analysis of patents in Northern Italy. European
Journal of Innovation Management 18: 238–257. CrossRef.
Alon I, Rottig D (2013) Entrepreneurship in emerging markets: New insights and directions
for future research. Thunderbird International Business Review 55: 487–492. CrossRef.
Andersson M, Klepper S (2013) Characteristics and performance of new firms and spinoffs
in sweden. Industrial and Corporate Change 22: 245–280. CrossRef.
Andersson R, Mandell S, Wilhelmsson M (2014) S˚a skapas attraktiva st¨ader. Reforminsti-
tutet, isbn 978-91-980600-8-9
Anselin L (1988) Spatial Econometrics: Methods and Models. Kluwer, Dordrecht. CrossRef.
Anselin L (2005) Exploring spatial data with geoDa
TM
: A Workbook. Center for spatially
integrated social science. revised version, march 6, 2005
Anselin L, Bera A, Florax R, Yoon MJ (1996) Simple diagnostic tests for spatial depen-
dence. Regional Science and Urban Economics 26: 77–104. CrossRef.
Arena for Growth (2015) The new geography of sweden. Arena for growth, Stockholm
Armington C, Acs ZJ (2002) The determinants of regional variation in new firm formation.
Regional Studies 36: 33–45. CrossRef.
Atherton A (2012) Cases of start-up financing: An analysis of new venture capitalisa-
tion structures and patterns. International Journal of Entrepreneurial Behavior &
Research 18: 28–47
Audretsch D, Fritsch M (1994) The geography of firm births in Germany. Regional
Studies 28: 359–365. CrossRef.
Balasubramanyan L, Coulson E (2013) Do house prices impact business starts? Journal
of Housing Economics 22: 36–44. CrossRef.
Bateman M (2000) Neo-liberalism, SME development and the role of business support
centres in the transition economies of Central and Eastern Europe. Small Business
Economics 14: 275–298
Berger AN, Udell GF (1998) The economics of small business finance: The roles of private
equity and debt markets in the financial growth cycle. Journal of Banking & Finance 22:
613–673
Berggren B, Olofsson C, Silver L (2000) Control aversion and the search for external
financing in swedish SMEs. Small Business Economics 15: 233–242
Bernanke BS, Gertler M, Gilchrist S (1999) The financial accelerator in a quantitative
business cycle framework. In: Taylor J, Woodford M (eds), Handbook of macroeconomics:
vol. 1. North Holland, Amsterdam. CrossRef.
Binet ME, Facchini F (2015) The factors determining firm start-ups in French regions and
the heterogeneity of regional labor markets. Annals of Regional Science 54: 251–268.
CrossRef.
REGION : Volume 4, Number 1, 2017
14 B. Berggren, A. Fili, M. H. Wilhelmsson
Birch DGW (1979) The job generation process. MIT Program on Neighborhood and
Regional Change 302
Black J, de Meza D, Jeffreys D (1996) House prices, the supply of collateral and the
enterprise economy. Economic Journal 106: 60–75. CrossRef.
Bracke P, Hilber C, Silva O (2013) Homeownership and entrepreneurship: The role of
commitment and mortgage debt. Institute for the Study of Labor (IZA) discussion
paper no. 7417, IZA, Bonn
Cassar G (2004) The financing of business start-ups. Journal of Business Venturing 19:
261–283. CrossRef.
Chaney T, Sraer D, Thesmar D (2012) The collateral channel: How real estate shocks
affect corporate investment. American Economic Review 102: 2381–2409. CrossRef.
Chasco C (2013) GeoDaSpace: A resource for teaching spatial regression models. React@,
Monogr´afico 4: 119–144
Davidsson P, Lindmark L, Olofsson C (1995) Small firms, business dynamics and dif-
ferential development of economic well-being. Small Business Economics 7: 301–315.
CrossRef.
Davidsson P, Steffens P, Fitzsimmons J (2009) Growing profitable or growing from profits:
Putting the horse in front of the cart? Journal of Business Venturing 24: 388–406.
CrossRef.
De Clercq D, Lim D, CH O (2013) Individual-level resources and new business activity:
The contingent role of institutional context. Entrepreneurship Theory and Practice 37:
303–330. CrossRef.
Degryse H, de Goeij P, Kappert P (2012) The impact of firm and industry characteristics
on small firms’ capital structure. Smal l Business Economics 38: 431–447. CrossRef.
Doh S, Kim B (2014) Government support for SME innovations in the regional industries:
The case of government financial support program in South Korea. Research Policy 43:
1557–1569. CrossRef.
Donaldson G (1984) Managing corporate wealth: The operation of a comprehensive
financial goals system. Praeger, New York
Elhorst JP (2010) Applied spatial econometrics. Spatial Economic Analysis 5: 9–28.
CrossRef.
GEM (2012) Entrepreneurship in Sweden: National Report. Entrepreneurship Forum,
Stockholm
Gordon IR, McCann P (2005) Innovation, agglomeration, and regional development.
Journal of Economic Geography 5: 523–543. CrossRef.
Greenspan A, Kennedy J (2008) Sources and uses of equity extracted from homes. Oxford
Review of Economic Policy 24: 120–144. CrossRef.
Gregory BT, Rutherford MW, Oswald S, Gardiner L (2005) An empirical investiga-
tion of the growth cycle theory of small firm financing. Journal of Small Business
Management 43: 382–392. CrossRef.
Halilem N, Bertrand C, Cloutier JS, Landry R, Amara N (2012) The knowledge value
chain as an SME innovation policy instrument framework: An analytical exploration
of SMEs public innovation support in OECD countries. International Journal of
Technology Management 58: 236–260. CrossRef.
Halleck Vega S, Elhorst J (2013) On spatial econometric models, spillover effects, and W.
Working paper, Faculty of Economics and Business, University of Groningen
REGION : Volume 4, Number 1, 2017
B. Berggren, A. Fili, M. H. Wilhelmsson 15
Harding R, Cowling M (2006) Assessing the scale of the equity gap. Journal of Small
Business and Enterprise Development 13: 115–132. CrossRef.
Howorth CA (2001) Small firms’ demand for finance: A research note. International
Small Business Journal 19: 78–86. CrossRef.
Hurst E, Lusardi A (2004) Liquidity constraints, household wealth, and entrepreneurship.
Journal of Political Economy 112: 319–347. CrossRef.
Iacoviello M (2005) House prices, borrowing constraints, and monetary policy in the
business cycle. American Economic Review 95: 739–764. CrossRef.
Jensen MC, Meckling WH (1976) Theory of the firm: Managerial behavior, agency costs
and ownership structure. Journal of Financial Economics 3: 305–360. CrossRef.
Jin Y, Leung CKY, Zeng Z (2012) Real estate, the external finance premium and
business investment: A quantitative dynamic general equilibrium analysis. Real Estate
Economics 40: 167–195. CrossRef.
Johannisson B (1984) A cultural perspective on small business – local business climate.
International Small Business Journal 2: 32–43
Johnsen GJ, McMahon RGP (2005) Owner-manager gender, financial performance and
business growth amongst SMEs from Australia’s business longitudinal survey. Interna-
tional Small Business Journal 23: 115–142. CrossRef.
Keeble D (1997) Small firms, innovation and regional development in Britain in the 1990s.
Regional Studies 31: 281–293. CrossRef.
Lee SY, Florida R, Acs Z (2004) Creativity and entrepreneurship: A regional analysis of
new firm formation. Regional Studies 38: 879–891. CrossRef.
LeSage J, Pace RK (2009) Introduction to Spatial Econometrics. Chapman & Hall/CRC,
Boca Raton. CrossRef.
LeSage J, Pace RK (2014) The biggest myth in spatial econometrics. Econometrics 2:
217–249. CrossRef.
Love JH, Roper S (2015) SME innovation, exporting and growth: A review of existing
evidence. International Small Business Journal 33: 28–48. CrossRef.
Mac an Bhaird C, Lucey B (2011) An empirical investigation of the financial growth
lifecycle. Journal of Small Business and Enterprise Development 18: 715–731. CrossRef.
MacMillan HP (1931) Committee on finance and industry: report. Cmnd 4811, His
Majesty’s Stationery Office (hmso), London
Manove M, Padilla J, Pagano M (2001) Collateral versus project screening: A model of
lazy banks. RAND Journal of Economics 32: 726–744. CrossRef.
Mayer C (1988) New issues in corporate finance. European Economic Review 32: 1167–
1183. CrossRef.
Mendiola L, Gonz´ales P, Cebollada A (2015) The relationship between urban development
and the environmental impact mobility: A local case study. Land Use Policy 43:
119–128. CrossRef.
Myers SC (1984) The capital structure puzzle. Journal of Finance 39: 574–592. CrossRef.
Newman A, Gunessee S, Hilton B (2012) The applicability of financial theories of cap-
ital structure to the Chinese cultural context: A study of privately owned SMEs.
International Small Business Journal 30: 65–83. CrossRef.
REGION : Volume 4, Number 1, 2017
16 B. Berggren, A. Fili, M. H. Wilhelmsson
Paul S, Whittam G, Wyper J (2007) The pecking order hypothesis: Does it apply to
start-up firms? Journal of Small Business and Enterprise Development 14: 8–21.
CrossRef.
Perren L, Jennings PL (2005) Government discourses on entrepreneurship: Issues of
legitimization, subjugation, and power. Entrepreneurship Theory and Practice 29:
173–184. CrossRef.
Psillaki M, Daskalakis N (2009) Are the determinants of capital structure country or firm
specific? Small Business Economics 33: 319–333. CrossRef.
Ramlall I (2014) Is there a pecking order in the demand for financial services in Mauritius?
Journal of African Business 15: 49–63. CrossRef.
Revest V, Sapio A (2012) Financing technology-based small firms in Europe: What do
we know? Small Business Economics 39: 179–205. CrossRef.
Robb AM, Robinson DT (2014) The capital structure decisions of new firms. Review of
Financial Studies 27: 153–179. CrossRef.
Robson MT (1996) Housing wealth, business creation and dissolution in the UK regions.
Small Business Economics 8: 39–48. CrossRef.
Ross SA (1973) The economic theory of agency: The principal’s problem. American
Economic Review 63: 134–139
Santarelli E, Tran HT (2012) Growth of incumbent firms and entrepreneurship in Vietnam.
Growth and Change 43: 638–666. CrossRef.
Sch¨afer D, Talavera O, Weir C (2011) Entrepreneurship, windfall gains and financial
constraints: Evidence from Germany. Economic Modelling 28: 2174–2180. CrossRef.
Schmalz MC, Sraer DA, Thesmar D (2013) Housing collateral and entrepreneurship.
Working paper 19680, National Bureau of Economic Research (NBER), Cambridge,
MA. CrossRef.
Shane SA (2000) A general theory of entrepreneurship: The individual-opportunity nexus.
Edward Elgar, Cheltenham
Stakhovych S, Bijmolt THA (2009) Specification of spatial models: A simulation study
on weights matrices. Papers in Regional Science 88: 389–408. CrossRef.
Statistics Sweden (2015) Business activities. http://www.scb.se/en /Finding-
statistics/Statistics-by-subject-area/Business-activities/
Stiglitz JE, Weiss A (1981) Credit rationing in markets with imperfect information.
American Economic Review 71: 393–410
Storey DJ, Greene FJ (2013) Small business and entrepreneurship. Pearson, Harlow
Vanacker TR, Manigart S (2010) Pecking order and debt capacity considerations for high-
growth companies seeking financing. Small Business Economics 35: 53–69. CrossRef.
Weston JF, Brigham EF (1981) Managerial finance (7th edition ed.). Dryden Press,
Hinsdale, IL
Wilhelmsson M (2002) Spatial models in real estate economics. Housing, Theory and
Society 19: 92–101. CrossRef.
REGION : Volume 4, Number 1, 2017
