Leon Boerop - Universiteit van Amsterdam - 07/30/2014
The focus of this report is on the personal motivation on a choice for entrepreneurship when a national disaster has struck. Through data on disasters, terrorist attacks and violence these motivations can be analysed. The research question of this report is: what is the role of extreme national events in one’s decision to become an entrepreneur? To answer these question two hypotheses will be examined, concerning an increase in entrepreneurship due to disasters, and the motivation (necessity- or opportunity-driven) behind the entrepreneurial choice. The literature studied indicated there might be a positive effect of war on entrepreneurship, and literature indicated that entrepreneurs become innovative by proposing unconventional manners in which they overcome the sub-optimal conditions for setting up a business. Data indicated that in case of natural disasters there is a minimal positive effect on entrepreneurship in a country. Violent disasters leave entrepreneurship unaffected. Data also indicated that the rise in entrepreneurial initiatives is caused more by opportunism than by necessity. In a violent disaster the amount of deaths and fatalities have a significantly positive effect on the percentage of entrepreneurs that took his decision out of opportunism rather than necessity. The effect of natural disasters does not have a significant effect on opportunism-related entrepreneurship. Data indicated that the observed effects occurred in the long and short run, indicating that those triggers that affect entrepreneurship influence households directly and indirectly.
The focus of this report is on the personal motivation on a choice for entrepreneurship when a national disaster has struck. Through data on disasters, terrorist attacks and violence these motivations can be analysed. The research question of this report is: what is the role of extreme national events in one’s decision to become an entrepreneur? To answer these question two hypotheses will be examined, concerning an increase in entrepreneurship due to disasters, and the motivation (necessity- or opportunity-driven) behind the entrepreneurial choice. The literature studied indicated there might be a positive effect of war on entrepreneurship, and literature indicated that entrepreneurs become innovative by proposing unconventional manners in which they overcome the sub-optimal conditions for setting up a business. Data indicated that in case
Leon Boerop - 10656480 (Master Student Organization Economics) Supervisor: Joeri Sol Pages: 82 Word Count: 24. 709
Entrepreneurship and
catastrophes: what triggers
the disaster-struck pioneer?
1
Table of Contents
1. INTRODUCTION ... 2 1.2MOTIVATION 4 2. LITERATURE REVIEW ... 7 2.1 HYPOTHESIS BUILDING 7 2.2 THE HOUSEHOLD SITUATION 9 2.3 ENTREPRENEURSHIP IN DEVELOPING ECONOMIES 10 2.3.1ACCESS TO FINANCE ... 112.3.2.INFRASTRUCTURE ... 12
2.3.3.THE GOVERNMENT... 13
2.3.4SKILLED LABOUR ... 14
2.3.5SOCIAL COHESIVENESS ... 15
2.4 ENTREPRENEURSHIP AND WAR 15 3. METHODOLOGY ... 18
3.1 DATA DESCRIPTION 18 3.1.1.ENTREPRENEURSHIP DATA ... 18
3.1.2.NATURAL DISASTERS DATA ... 18
3.1.3VIOLENCE DATA ... 19 3.1.4COUNTRIES USED ... 19 3.2 METHODOLOGY ANALYSIS 23 3.2.1DIRECT EFFECT ... 23 3.2.3.THE CUMULATIVE EFFECT ... 24 3.2.4.COMBINED EFFECT ... 26
4. RESULTS & ANALYSIS ... 28
5. CONCLUSION ... 34
5.2DISCUSSION 35 5.3FUTURE RESEARCH 37 6. BIBLIOGRAPHY ... 38
7. APPENDIX 1: MEASURES QUESTIONNAIRE ... 43
ENTREPRENEURIAL ACTIVITY ... 43
ENTREPRENEURIAL ASPIRATIONS... 43
ENTREPRENEURIAL ATTITUDES ... 44
8. APPENDIX 2: QUESTIONNAIRE ... 45
9. APPENDIX 3: DATA DIRECT REGRESSIONS ... 46
2 ‘The war made us tougher. It made us hungrier; more hungry for success’ (Novak Djokovic, 2012)
1. Introduction
A national disaster – caused by violence or nature – affects human lives on many levels. Disasters leave a trace of grief, tragedy and shattered dreams. Disasters however also give rise to drastic changes of faith. The quote above originates from a CBS-interview in which Novak Djokovic states - after his victory on Wimbledon - that the war in Yugoslavia and the bombings in Belgrade made him tougher and mentally shaped him into the champion he is today. Moreover, Djokovic explained that a side-effect of the bombings in Belgrade was endless free time. Due to the absence of employment or education opportunities (schools and factories were destroyed) many young Serbs chose alternative activities to spend time, mostly to kill boredom. This unexpected turn of faith caused Djokovic to make the many necessary practice hours in tennis.
Evidently a disaster has many terrible effects. However, beyond this dramatic and soul-stirring scope of tragedies the situation of Djokovic indicates a disaster might also bring unexpected outcomes. Either out of necessity (survival) or opportunity (due to changing circumstances) one might make life-changing decisions. In this report I will specifically analyse the hypothesis that disasters (natural disasters and war) positively influence one’s decision to become an entrepreneur. In addition, I will attempt to explore what triggers entrepreneurs in times of catastrophes.
This hypothesis is based upon the concepts of the Occupational Choice Models (OCM’s) and behaviour following welfare maximization of economic actors. Occupational Choice Models describe the trade-off actors make when deciding to become an entrepreneur or follow regular employment. Holmes and Schmitz (1990) use one OCM to argue that the economy is in a permanent state of disequilibrium and individuals are constantly exposed to new opportunities (in the case of Holmes and Schmitz due to technological progress, in the case of this report due to changing circumstances). This changing environment alters the opportunity costs of labour or the opportunities on the labour market in general, and therefore the equilibrium between the entrepreneurial choice and regular employment. Applied to a war or natural hazard: regular employment disappears or becomes less attractive (as wages might decrease following lower economic performance) and therefore becoming an entrepreneur might become the more
3 attractive option. Note: the Occupational Choice Model is only a theoretical prediction. In this report the theory is used solely as an applicable line of reasoning for aspiring entrepreneurs. I will elaborate on the wealth hierarchy that the OCM’s imply in the literature review.
Historical data supports the assumption used in OCM’s that when employment is less attractive more people choose to become an entrepreneur, as low GDP-countries show higher levels of the Total early-stage Entrepreneurial Activity (TEA)1, suggesting that the deficient opportunities in the labour market are offset by more entrepreneurial incentives. As Figure 1 indicates countries that perform economically bad have a higher TEA and countries with a prosperous economic climate show less entrepreneurial activity.
A study by Wennekers et al. (2005) elaborates on this observation, as Wennekers et al. claim there is a U-shaped relation between entrepreneurial dynamics and the level of economic development of a country. This observation is in line with the interpretation that low-GDP countries give rise to entrepreneurship due to the lack of better alternatives, developed (high-GDP) countries show low entrepreneurial activity due to the presence of better alternatives and in the case of extremely high GDP (wealthy individuals) data shows high rates of entrepreneurship as these individuals have less to lose and can therefore take the risk of abandoning regular employment (Wennekers, van Stel, Thurik, & Reynolds, 2005). In addition to these results, one should not forget that the entrepreneurial choice can also be non-voluntarily. I will elaborate more on this situation in the literature review using the work of Tamvada (2005).
One specific drawback on the work of Wennekers et al. and Holmes and Schmidt is that these studies lack context, background and a sense of realism that makes them applicable to real-life situations. They solely provide the insight that there is a hierarchy in employment and that
1 The TEA -figure is a figure collected by GEM incorporating the percentage of 18-64 year old people who are a nascent
entrepreneur or an owner of a business younger than 3,5 years.
4 developing economies realize more entrepreneurial activity compared to developed economies. Further explanation on this observation unfortunately lacks and the same effect has never been studied during disasters. Disasters in this context are particularly relevant as they represent a specific point in time at which the welfare distribution (and following the welfare hierarchy) experiences a short shock and possibly changes. The situation before and after the shock can be compared. The effect of this shock possibly changes the equilibria but these effects have never been studied.
The focus of this research is on the personal motivation behind one’s decision to exchange regular employment for entrepreneurial ambitions. A questionnaire from the Global Entrepreneurship Monitor distinguishes between entrepreneurship by necessity and opportunity. I focus on entrepreneurship out of necessity or opportunity, as these are the major entrepreneurial triggers at work during a disaster. Through data on disasters, terrorist attacks and violence these motivations can be analysed in the event of a shock. The research question of this report is therefore: what is the role of extreme national events in one’s decision to become an entrepreneur? To answer this question two hypotheses will be examined.
Hypothesis 1: A natural disaster increases the probability one decides to become an entrepreneur
Hypothesis 2a: An aspiring entrepreneur experiences more opportunities for entrepreneurship when a disaster strikes
Hypothesis 2b: An aspiring entrepreneur is forced into the decision of entrepreneurship out of necessity when a disaster strikes.
1.2 Motivation
Concerning the relevance of this report, the direct relationship between a sudden change in one’s welfare distribution (in this case identified as a national disaster) and the choice to become an entrepreneur has not been studied in detail. Yet, it is interesting to examine why one makes the decision to become an entrepreneur. This study therefore fills a gap between literature on entrepreneurial motivation and economic circumstances that stimulate entrepreneurship, and literature that studies the economic consequences of disasters. Case studies indicate that a shift towards entrepreneurship in times of disaster is not as far-fetched as it seems. For example studies in Lebanon (Ciarli, Parto, & Savona, 2010) and Afghanistan (Fahed-Sreih & Pistrui, 2012) demonstrate that when safety disappears family businesses are used to control for risk, and regular employment takes a less dominant position in the welfare hierarchy. Case studies like these
5 indicate that the topic is of contemporary relevance, whilst literature leaves some relevant issues unexplored.
Regarding the literature on entrepreneurship, many studies have covered start-up rates, GDP growth, labour, and other macro-economic measures in relation to entrepreneurship, but the personal motivation behind one’s occupational choice in relation to distress is unknown. Why would one forbear on regular employment and choose for the perilous profession of an entrepreneur? Much about what makes entrepreneurs ‘tick’ is known, but what could be the trigger in times of disaster? Besides, most entrepreneurial research to date is carried out in developed economies, however economies suffering from conditions that are less economically favourable are relatively unexplored (Ciarli, Parto, & Savona, 2010). Concerning the relevance of results of these studies: reports on entrepreneurship in developed countries indicate that entrepreneurship and the entrepreneurial firm are a driving force of economic growth (Lazear, 2005), and a large amount of research by entities such as the United Nations and the World Bank concluded that entrepreneurship is even likely to be a necessary condition for generating economic growth (United Nations Development Programme, 2004). It is however less clear what kind of policy measures are effective in order to stimulate entrepreneurship. A second topic of interest of this report is therefore in what way entrepreneurship can contribute to conflict resolution and economic recovery in a disaster-struck region.
Concerning the literature describing disasters, there is a large amount of literature that studies either the effect natural disasters, wars or terrorist attacks on the economy, or that studies very case-specific disasters and their effects. There are however no studies that combine these insights with the entrepreneurship literature to provide a general conclusion. Literature from the fields described provides the starting point for this paper, as can be read in the literary review. An analysis with country-level data and specific details on disasters follows to give an informed conclusion.
Summarizing, the results of this report are as follows: the literature studied indicated there might be a positive effect of war on entrepreneurship. Besides, literature indicated that entrepreneurs during times of disaster become innovative by proposing unconventional manners in which they overcome the sub-optimal conditions for setting up a business. Data indicated that in case of natural disasters there is a small positive effect on entrepreneurship in a country. Violent disasters leave entrepreneurship unaffected. Data also indicated that the rise in entrepreneurial initiatives is caused more by opportunity than by necessity. In a violent disaster the amount of deaths and
6 wounded have a significantly positive effect on the percentage of entrepreneurs that took his decision out of improvement (meaning: opportunities opposed to no other option for work and increasing income). The effect of natural disasters does not have a significant effect on opportunity-related entrepreneurship. The observed effects occur on the long run (3 – 5 years) indicating that those triggers that affect entrepreneurship influence households indirectly.
This report is structured as follows. In the second chapter literature will be treated that elaborates on the choices of entrepreneurs and the conditions they face in countries experiencing war or natural disasters. Chapter three describes the data and methodology, chapter four follows with an analysis of the data, and chapter five provides a conclusion and a discussion on the results found.
7
2. Literature review
This literature review is structured as follows: I first explain why the Occupational Choice Models are not developed enough to be able to form an hypothesis. I will therefore - in addition to the OCM’s - elaborate on the welfare hierarchy and the psychological predictions of welfare economics. Afterwards I will treat empirical literature and I will summarize the necessary conditions that are required for entrepreneurial development, and I will relate these conditions to the situation observed in disaster-struck countries. At last the situations of households and business in times of war will be described. The topics treated are selected for their predictive capabilities on the studied subject.
2.1 Hypothesis building
The Occupational Choice Models represent a theoretical argument that uses the concept of the welfare hierarchy in occupations to explain one’s decision to become an entrepreneur, making them an ideal theoretical embedding for the hypothesis of this report. There are however a few objections against the OCM’s. First of all the OCM’s do not explain entrepreneurship out of necessity, the models solely view entrepreneurship as a voluntary choice. Second, the explanatory power of the OCM’s is limited as the models are of a very theoretical kind. I will therefore elaborate on the OCM’s using empirics of Tamvada (2009), I will present a psychological overview of the forces at work leading to a choice such as entrepreneurship and I ask for more research in the field of theory-building on the entrepreneurial choice in the discussion-chapter of this paper.
Where the occupational choice models assume that individuals opt for self-employment when they expect higher returns from doing so relative to wage-employment, Tamvada (2009) suggests that this is more of a selection-process in which the most productive individuals become entrepreneurs or very successful employees, and the lowest productivity workers (out of survival) become informal entrepreneurs or subsistence workers. Using collected data Tamvada proves a difference in theory and reality: models in economics assume that individuals become self-employed as they expect higher expected utility relative to wage employment whereas in developing economies people are forced into self-employment in the absence of viable economic opportunities. Besides, the data from Tamvada indicates that entrepreneurs have non-pecuniary motivations for their entrepreneurial decisions (being one’s own boss, income hedging, etc.) suggesting that they select into entrepreneurship even when the returns in terms of earnings are lower. Tamvada states that instead of being an entrepreneur or wage-worker one can select into five primary occupations:
8 being an employer, self-employed, salaried employee, casual labourer or unemployed. This result is relevant for this report, as it appears that in many cases the hierarchy of employments is built up different than expected and is therefore hard to compare. Besides, entrepreneurship does not always seem to be a better or worse option compared to employment, as it appears that self-employment or being an employer (both the result of entrepreneurial activities) have a different ranking in the hierarchy. I will elaborate on this statement in the discussion-chapter, as this remark on the occupational hierarchy is relevant for hypothesis one.
Besides welfare-related considerations the concept of risk aversion is relevant to this report, as regular employment is considered a safe choice and entrepreneurship a risky alternative. Results from psychological studies can be informative to speculate on risky behaviour in situations that have not been investigated empirically. Studies on households indicated that the discount rates are a subjective representation of one’s time and risk preferences and represent the cost-benefit analysis of a household for future decisions (Harrison, Morten, & Melonie, 2001). Therefore, the effect of a changing cost-benefit analysis and beliefs about future circumstances change one’s discount rates and following ones’ future actions. Studies confirm that a war or disruption can influence these beliefs and therefore one’s behaviour, towards a more short-term and risky preference (Voors et. al, 2010). These studies do not comment on and neither confirm whether this might lead to a change in entrepreneurial efforts.
Studies indicated that people appear to value their assets in an unexpected manner. Considering the utility on the future value of assets one prioritizes gains and losses rather than final assets. This means it isn’t the future value of wealth that people value the most but the change in future funds (Kahneman & Tversky, 1979). In the situation of a war or disaster, when one has lost a large share of valuable assets there is less to lose and more to gain, due to overestimation of low probabilities one might consider a risky option such as entrepreneurship.
An implication of these conclusions (that the effect of a changing cost-benefit analysis and beliefs about future circumstances change one’s discount rates and following ones’ future actions, and that people value gains and losses more than final assets) is examined by Raghunathan and Pham (1999) in a situation comparable to the one for example a warzone entrepreneur might face. Raghunathan and Pham studied how mood affects behaviour and how this influences the decision process. Three experiments were conducted. First, participants had to read a story outlining a hypothetical situation in which they represented a personage that was anxious, neutral or sad and this state of mind was verified by a questionnaire. Subsequently the subjects had to participate in
9 three experiments; one was a gambling exercise, one forced them to choose between two job options (a risky one with a high wage or a secure one with a lower wage) and the last exercise was a replicated version of the first exercise but with a few modifications to test for continuity and learning effects. Subjects appeared to behave different in terms of risk and consequently in the choice for the job contract. Sad individuals appear to be biased in favour of high risk and high reward options, anxious individuals are biased towards low risk / low reward options, compared to their neutral option. One explanation is that anxiety and sadness convey distinct types of information to the participant and prime different goals; anxiety primes an implicit goal of uncertainty reduction and sadness primes an implicit goal of reward replacement (Raghunathan & Pham, 1999). Studies in Israel concluded that violence causes both the emotions of sadness and fear, but the effects of fear are only weakly visible in people’s decisions indicating that sadness is the stronger emotion (Becker & Rubinstein, 2004). This indicates that Raghunathan & Pham’s conclusion with respect to sadness (high risk / high reward) is most applicable to violent periods. The entrepreneurial decision is compared to labour a situation of high risk and high reward.
Summarizing the theoretical and experimental results: according to the field of psychology, in the situation of job loss due to disaster in combination with fear, studies hint towards a risky choice such as entrepreneurship. Surprisingly however, a scarce collection of case-studies on entrepreneurial behaviour in times of disaster show a contrary effect compared to theoretical predictions. These studies mostly concerned farmers in ‘neutral’ weather regions and farmers in risky regions, comparing their allocation of assets. Farmers in regions that are often struck by disaster appear to show a more risk-averse approach to their investments and income. (Rosenzweig & Binswanger, 1992).
2.2 The household situation
From the macro-economic point of view several factors influence a country during and after a disaster, including: destroyed infrastructure, a decrease in social cohesion, environmental modifications, fiscal and foreign sector imbalances, price increases, modifications to demographic structures and changes in development priorities as lost or damaged assets need replacement (ECLAC, 2003). The most observed effects on a household-level are loss of life, loss of valuable assets and the loss of employment. Observed influences can be divided between direct and indirect effects (Justino & Verwimp, 2006). The direct effect is also referred to as the ‘shock component’ on the overall vulnerability and risk exposure of a household, the indirect effect is also referred to as the ‘institutional component’ (Binzel & Brück, 2007). Second to their effect, the direct and indirect
10 effect differ in their timespan. A direct effect is noticed immediately but an indirect effect takes at least nine quarters of a year to show an effect (Belasen & Polach, 2007).
The direct effects of disruptions are those related to the immediate destruction of assets, both physical and human. The impact of direct effects is likely to be negative, also concerning the amount of entrepreneurs in a country (Verwimp & Bundervoet, 2008; Justino P. , 2009). Indirect effects influence the household through welfare by influencing markets, politics and social networks. The likelihood of households engaging in entrepreneurial activities might also be influenced through welfare. Remarkably, one possible scenario is that through shifting abilities to alternative economic activities (exploiting predatory behaviour, rent-seeking and illegal activities) this might lead to entrepreneurial initiatives (Baumol, 1990). In this case entrepreneurship is driven by an instinct to survive (Naudé, 2007). As in this paper I also want to address the topic of reconstruction and conflict resolution therefore, this effect of rent-seeking and illegal activities will be treated later in this report. To summarize, the literature suggests that entrepreneurial activities might increase due to a shock, however this increase might not be beneficial to economic performance on the long run (Cooper, 2006; Kahn & Jomo, 2000).
2.3 Entrepreneurship in developing economies
On a global scale entrepreneurs differ in type. For example Europe and the USA startups are mainly innovation-driven, whereas in Africa, Asia and South America they mainly entail factor-driven activities (Desai & Hessels, 2008). These different businesses have different needs, and therefore Figure 1 on Total Entrepreneurial Activity does not show the complete entrepreneurial environment. To define the environment of entrepreneurship in developing economies I will now analyse the entrepreneurial conditions in these countries. I will use specific cases from countries that have been struck by violence or a natural disaster to provide case-specific insights, as the entrepreneurs in war-struck regions in general have not been studied yet.
The five ingredients of entrepreneurship that structure this paragraph are: access to finance, infrastructure, the mode of government, availability of skilled labour and social cohesiveness (Ciarli, Parto, & Savona, 2010). A study by Ali (2011) examining the war in Darfur, Sudan indicates how destructive a war can be on these entrepreneurial ingredients. First, the costs are visible in government expenditures on the war ($24,07 billion in this case, 162% of the Sudanic GDP), excluding $2,6 billion in lifetime earnings of the dead, $7,2 billion in productivity lost, $4,1 billion infrastructure damage and only 1,3% and 1,2% of the countries’ budget spent on public health and education. Direct effects are not mentioned by Ali but his conclusion shows that through
11 macroeconomic effects entrepreneurs are struck indirect at many aspects necessary for business growth (Ali, 2011).
The above described ingredients are less favourable in developing economies. However, and perhaps surprisingly, according to Figure 2 data by GEM indicates that the perceived opportunities to start a business are higher in developing economies. On average 69% of the entrepreneurs in Sub Saharan African Economies tend to see good opportunities to start a business in their region. The entrepreneurs in the European Union on average show a lower perception of
available opportunities (27,4% average). Figure 2 therefore indicates that the absence of resources in times of disaster might not be a major drawback for entrepreneurship. I discuss some case-specific examples of how entrepreneurs solve their lack of resources, which may explain perceived opportunities.
2.3.1 Access to finance
Figure 3 indicates that developing economies lack bank financing. Entrepreneurs in these economies might therefore experience problems in starting their business as one of the entrepreneurial ingredients clearly lacks. Besides, as Figure 4 indicates banks in developing countries fulfil a different role as they provide many state financial services, less cooperatives and less commercial activities, therefore acting less according to the demand of entrepreneurs.
A study by Kunt, Klapper and Panos (2008) shows that the active presence of a commercial bank strongly influences the survival rate of newly found businesses, however developing countries have a significantly smaller share of such banks compared to developed countries. Evidence shows that in developing countries working capital is available (however it is more expensive) and startup capital is very problematic (Kunt, Klapper, & Panos, 2008).
12
Figure 3 - Bank loans per 1.000 adults. Source: The World Bank (2010)
The most common solution to issues concerning finance are the use of Micro Finance Institutions (MFI), in which small loans supported by many individuals instead of banks provide credit and saving facilities to entrepreneurs (Benson & Clay, 2006). Besides MFI’s, a large amount of private savings (Ciarli, Parto, & Savona, 2010) or existing family capital (Fahed-Sreih & Pistrui, 2012) is used as investment capital. Also, banks show a learning effect and currently invest more in entrepreneurial activities in developing economies compared to a decade ago, as methods are developed to distinguish the profitable from the non-profitable enterprises (Townsend & Paulson, 2005). At last non-pecuniary methods of financing are observed like the use of livestock as collateral and risk-hedging instrument (Fafchamps, Udry, & Czukas, 1998) and the use of livestock as an income-smoothening instrument (Jacoby & Skoufias, 1995), and cross-border trade in a shadow economy that functions by trading goods rather than money (Dana & Galbraith, 2006).
2.3.2. Infrastructure
A countries’ infrastructure is one of the first and severe ‘victims’ of a natural disaster (Tierney & Webb, 2001; Ali, 2011). For a study in South Africa Isaacs & Friedrich (2006) interviewed organizations that support entrepreneurs in order to discover the success-factors of start-ups in
Figure 4 - Banking activities. Source: The World Bank (2010)
13 developing countries. The results indicate that a lack of infrastructure is perceived as the smallest hinder to success for entrepreneurs in developing economies, as Table 1 indicates:
Table 1 - Factors Inhibiting Success. Source: Isaacs & Friedrich (2006)
The ‘Mean values’ are the values (on a scale of 1 – 10) respondents gave the topics in terms of relevance in startup success. This table indicates that infrastructure is a relevant hurdle, but not the most difficult one for start-ups in developing economies (Isaacs & Friedrich, 2006).
Tierney and Webb (2001) indicate a second fact; if infrastructure becomes an issue, this is most likely also the case for competition. Only companies that control risk by securing operations in other (non-affected) regions have competitive advantages. Besides, entrepreneurs can prepare for disaster but the impact is always unpredictable. Data from Tierney and Webb indicates that infrastructural damage after earthquakes accounts for a loss of 65% in a business’ revenue and possibly even a temporary closing of the business for a short while, but the impact on continuity is minimized as all businesses in the region experience the same losses (Tierney & Webb, 2001).
2.3.3. The Government
Government-related issues mainly concern institutional capabilities and qualities in helping the entrepreneur with safety or legal-related issues (De Soto, 2006). De Soto (2006) explains that the government is a crucial partner in terms of subsidies, policy and legal structures. Obviously the law enforcement with respect to violence is crucial in creating a safe environment for entrepreneurs, but also the enhancement of a corporate legal system with for example a credible patent policy is a critical service. Other than examples from entrepreneurship in the informal (shadow) economy mentioned from Mozambique, there are no studies that examine the behaviour of entrepreneurs in
14 countries with poorly run governments. Therefore there appears to be no general solution to a bad functioning government.
Several studies state that collaboration with the government can provide an accelerator for aspiring entrepreneurs in the near future but can cause problems on the long term due to corruption-related issues (Leff, 1979). An example of this situation is Sierra Leone, where, according to Neil Cooper (2006) foreign firms and entrepreneurs on purpose kept elites in power and provided essential state services such as security, finding their interest in weakening local government capacity (Cooper, 2006). This is an implication of the warlord effect (which is further explained in section 2.4).
2.3.4 Skilled labour
Campbell (2011) compared the labour markets in developing and developed countries on pure characteristics. Campbell’s most important conclusion was that due to various effects the productivity of labour in developing countries is low, as is visible in Figure 5. This limits entrepreneurs in such a way that with comparable relative costs, the company will never be as competitive compared to peers in developed economy.
Entrepreneurs show various reactions to this difference in labour quality. First, data shows that in developing economies businesses make more use of family members than regular employees, compared to peers in developed economies, as can be seen in Figure 6. Cases from Lebanon and Afghanistan confirm these statements and imply that a network of family and friends is rather used than the regular labour market (Ciarli, Parto, & Savona, 2010) (Fahed-Sreih & Pistrui, 2012). Second,
management uses their own assets to perform tasks that employees could have done if they were available (Alesch, Holly , Mittler, & Nagy, 2001). Third, firms in areas with less skilled labour show more organizational inertia and less clear future strategy, indicating management is more focussed
Figure 5 - Share contribution to the world GDP per capita 1969 – 2009 Source: Campbell (2011)
Figure 6 - Contribution by family members in total employment of youths. Source: Campbell (2011)
15 on working in the business itself than managing it, in comparison to businesses in developing countries (Faircloth & Bronson, 2001).
2.3.5 Social Cohesiveness
The social cohesiveness of the network around an entrepreneur is important for the usability of this network. Social cohesiveness functions as a brokerage across structural holes in the network of an entrepreneur. This can cause two parties to collaborate (Burt, 2000), provide access to knowledge (Kogut, 2000), provide access to labour (Meyers & Schultz, 1991) and social cohesiveness appears to improve managerial performance (RodanSimon & Galunic, 2004).
With respect to social cohesiveness there is no general conclusion from literature imposing solutions to a decreasing network around the entrepreneur. There is evidence that an entrepreneurs’ network becomes smaller (Sweet, 1998; Chang, 2010), and that shortly after a disaster the social cohesiveness of the network increases (Chowdhury, 2011). There are methods of hedging one’s exposure to the risk of disappearing ties, such as being a member of a church (Clegg, 2008), having a large family (Fahed-Sreih & Pistrui, 2012), belonging to certain business clubs (Deare, 2004) and social groups (De Walque & Verwimp, 2010) but it is complicated to prove that being a member of these groups is used as a method to for example hedge risk.
2.4 Entrepreneurship and war
Various studies have been carried out in the field of war and entrepreneurship but no general conclusion on the effect of war on entrepreneurial ambitions and motivations can be drawn. Hereby I will give an overview of the relevant results.
According to Bruck, Naudé and Verwimp (2009) one can distinguish between effects of conflicts that deplete the capital stock of a country and it’s firms (influencing the country on a macroeconomic scale), and effects that affect the civilian population. The first has an ambiguous effect on entrepreneurship, the last a negative. Brück, Naudé and Verwimp (2009) state that negative effects of conflict are in general direct effects such as less security, the risk of death, forced migration, destroyed infrastructure, insecure property rights, falling consumer demand and diminishing productivity, all increasing the transaction costs and decreasing the ease of doing business. Death, displacement and the destruction of physical infrastructure (roads and railroads but also hospitals and schools) cause long-term negative impacts on an economy (Miguel & Roland, 2006). Concerning starting enterprises positive effects can be the warlord-effects (creating wealth, power and sometimes even patronage through network and business resulting from war-related
16 activities), lower prices of goods relatively often used or increased prices of goods sold by the entrepreneur, and decreased market entry requirements (Brück, Naudé, & Verwimp, 2009).
Empirical evidence shows that violence has a growth-limiting effect that is much larger in developing economies compared to developed economies. On average an additional terrorist incident per million inhabitants was followed by a reduction of GDP growth by 1,5%. Factors that limit growth the most are crowding in of government expenditures and a reduction in investments. Intrastate violence has a larger impact than domestic violence as this also decreases international trade (Gaibulloev & Sandler, 2008). The impact of violence can be limited by national and international aid focussed on rebuilding the economy and limiting the damage to only a regional scale (Miguel & Roland, 2006).
Entrepreneurs and established businesses do not react equal to violence (Miguel & Roland, 2006; Ruttan, 2006). Ruttan (2006) questions the role of war in economic development as he states that military and defence related research, development and procurement have been major sources of technological development across a broad spectrum of industries in the US. Ruttan states that war has played a large role in the development of various industries (internet, space- and air travel, etc.) and thereby spurs economic growth. However, in context to this particular study Ruttan also argues that military R&D plays an ambiguous role with respect to entrepreneurship as innovations are costly and usually carried out by a specialized company. During WWII there has been a huge amount of war-related innovation and corresponding entrepreneurship however according to Ruttan in the current wars innovation is too complicated to show a large correlation with violence (Ruttan, 2006).
Various studies conclude that it is not necessarily due to the entrepreneurial climate after a war that a TEA might increase or decrease, but more due to the limitations that industries come across. War simply causes lower attractiveness for regular employment and following the occupational hierarchy this might lead to entrepreneurship (McDougal, 2010). However McDougal also argues that regular employment might cause a safer (and therefore more attractive) employment option as large companies can protect themselves against violence. Following a study in Liberia where there has been a war for 14 years, McDougal discovered four general reactions in which protection is possible: outsourcing of labour, (forced) increasing throughput of personnel, investments in strengthened property rights and safe accommodation. Also, some employment disappeared as staff often leaves the country and finds employment elsewhere, and people that lost their job also fled Liberia. These observations either make labour a more attractive option, or otherwise indicate
17 that labour that disappeared not necessarily shows up in entrepreneurial data as this also concerns people that left the labor population due to death or escaping the country, thereby influencing the results of this study.
McDougal also discovered a tactical component of firms in wars. Any party in a war appears to have a double relationship with the industrial sector as firms tend to have a certain ‘golden egg’; one particular reason why they are crucial in the domestic economy, providing them with a strategic importance not to be destroyed. A support base in war may thus erode with diminishing industrial performance, implying that the larger an economies’ industrial manufacturing sector is, the less destructive the conflict should be (the results from McDougal indicate this phenomenon only exists for manufacturing industries, not for service-related industries). To be precise, every percentage point rise in the value of manufactured goods as a fraction of total merchandise exports is associated with a significant 1.3% decrease in conflict related deaths experienced by the company (McDougal, 2010).
18
3. Methodology
3.1 Data description
The data used for this report is collected from different sources. In the following paragraphs the composition of these datasets will be explained.
3.1.1. Entrepreneurship data
The data on entrepreneurship is collected from the Global Entrepreneurship Monitor (GEM) database. GEM performs a yearly questionnaire on entrepreneurial behaviour on a worldwide scale; the Adult Population Survey. Data from this survey includes perception variables related to entrepreneurship such as perceived opportunities, perceived capabilities, fear of failure rate, entrepreneurship as a desirable career choice, and whether the choice to become an entrepreneur was necessity- or opportunity-driven. Table 1 describes the variables used for this report in more detail. All variables originating from the adult population survey and the questionnaire are described the appendix. In the analysis the variable TEA represents the rate of entrepreneurship.
Variable Purpose
1. Total early stage Entrepreneurial Activity (TEA)
Percentage of 18-64 populations who are either a nascent entrepreneur or owner-manager of a new business (younger than 3 years). This figure describes a countries’ entrepreneurial incentives. Various questions are used to identify individuals who are currently starting a new firm, as is included in the appendix. The variable is collected binary (yes/no) and is used as a countries’ percentage of TEA in the analysis.
2. Improvement-driven Percentage of those involved in TEA who (i) claim to be driven by opportunity instead of finding no other option for work; and (ii) who indicate the main driver for being involved in this opportunity is being independent or increasing their income, rather than just maintaining their income
3. Necessity-driven Percentage of those involved in TEA because they had no other option for work.
4. Desirable Entrepreneurship as a desirable career choice: Percentage of 18-64 year olds of the population who agree with the statement that in their country, most people consider starting a business as a desirable career choice.
5. Perceived Opportunities Percentage of 18-64 who see good opportunities to start a firm where they live.
6. Perceived Capabilities Percentage of 18-64 population who believe to have the required skills and knowledge to start a business. 7. Fear of Failure Percentage of 18-64 population with positive perceived opportunities who indicate that fear of failure
would prevent them from setting up a business. Table 2 - Variables used
3.1.2. Natural Disasters data
The database from CRED EM-DAT (Centre for Research on the Epidemiology of Disasters. Emergency Events Database) is used to collect data on natural disasters. The EM-DAT database contains yearly collected data giving an overview of the amount of disasters in a particular country,
19 the amount of casualties, the amount of injuries, the amount of people affected, the amount of homeless people, the total amount of affected people and the total damage in USD. The CRED EM-DAT database considers avalanches, landslides, droughts, famines, earthquakes, epidemics, floods, volcanic disasters and windstorms.
This particular dataset is used as CRED EM-DAT provides a world-wide objective basis for vulnerability assessment and a standard of observations. Government institutions and scientific studies use the database by CRED EM-DAT as the field of data-collection on disasters lacks standardization on definitions, methodologies, tools and sourcing whereas the database by CRED EM-DAT does provide these conventions. The database by CRED EM-DAT is however limited in its amount of countries and years in which data is collected.
The variables Damage in USD and Total amount of people affected (both in absolute figures and relative figures) are used as an indication of the impact of a natural disaster in a particular country at a particular time.
3.1.3 Violence data
Data from the Uppsala University department of Peace and Conflict Research gives a yearly overview of violence in a country through the amount of deaths and the amount of wounded people. These figures represent all possible events related to violence including (civil) wars, protests and events leading to revolutions. Other natural and non-natural causes of deaths such as crime or accidents are not included in the dataset.
The UCDP Conflict Encyclopaedia used is a detailed online database containing data on armed conflicts in terms of parties involved, weapons used and the origin of the conflict. Coverage is global with information from 1946 onwards. For this report the variables Deaths (describing the total amount of deaths due to violence in a country in one year) and Wounded (describing the total amount of wounded people due to violence in a country in one year) are used to give an indication of the amount of violence in a particular country at a particular time.
3.1.4 Countries used
65 countries have been analysed with 15.372 observations in total collected from 2000 to 2013. Some countries experienced disasters and some did not, making the last category useful as a comparison group. A selection of countries took place to pick those countries from the dataset that showed the most elaborate data. The countries have been divided upon developing and developed
20 countries to distinguish for this division in the analysis (a dummy ddveloping is used)2. The distinction between developed and developing has been made according to the division the IMF makes, updated on the 15th of February 2014 (Nielsen, 2014).
A large amount of data has been collected that was after all not all used for the regression. Robustness checks indicated that using the variables on damage, deaths, wounded people and affected people provided the best fit for the model. The correlation matrix is used to decide upon this fit as the chosen explanatory variables show high significant correlations with the dependent variables. Besides, regressions have been carried out and coefficients, Z-scores and R-Squared values lead to a selection of variables with the best explanatory power. As is visible from Table 3 some variables are incomplete. In some developing economies a few years in the dataset lack, however of all the years used there is a complete range of the variables. The data showed the following descriptive statistics:
Statistics Observations Mean Median Range Standard Deviation
Developing 845 0,65 1 1 0,48 TEA 650 10,53 8,20 40 7,44 Desirable 440 65,46 66 71 13,56 Fear Failure 512 33,61 33 55 8,51 Imprd. Dr. 392 48,27 47 72 13,68 Necc. Dr. 510 24,13 23 66 12,69 Perc. Skills 513 47,92 48 79 15,55 Perc. Opp 525 37,52 37 81 16,49 #Disasters 845 3,35 2 42 5,17 #Killed 845 584,95 7 88.450 4.798,71 #Injured 845 3.733,37 0 1.800.006 63.954,13 #Affected 845 2.748.431 1375 342.000.000 20.800.000 #Homeless 845 28.909,78 0 5.003.500 262.298,30 Damage ($) 845 2.781.074 1629 342.000.000 20.900.000 Aff. % pop 845 0,01 0 0,49 0,04 Dam. % GDP 845 0 0 0 0 Deaths 845 80,96 0 6.594 428,87 Wounded 845 55,52 0 4.044 276,38 Dummy Developed 845 0,65 1 1 0,48
Table 3 - Descriptive statistics
2 The countries treated are the developed countries Australia, Belgium, Canada, Croatia, Denmark, France, Germany,
Greece, Iceland, Ireland, Italy, Japan, Latvia, the Netherlands, Norway, Poland, Singapore, Spain, Sweden, Switzerland, the UEA, the UK, and the US and the developing countries Algeria, Angola, Argentina, Bosnia, Botswana, Brazil, Chile, China, Colombia, Costa Rica, the Dominican Republic, Ecuador, Egypt, Ghana, Guatemala, Hungary, India, Iran, Israel, Jamaica, South Korea, Macedonia, Malaysia, Mexico, Nigeria, Pakistan, Palestine, Peru, Romania, Russia, the Slovak Republic, Slovenia, South Africa, Taiwan, Thailand, Trinidad, Tunisia, Uganda, Uruguay, Venezuela and Zambia.
21 As can be seen the TEA is on average relatively low (10,53% of the 18-64 populations are either a nascent entrepreneur or owner-manager of a business younger than three years) but the range and standard deviation imply large fluctuations between countries. Remarkable is that although the TEA is around 10, the variable Desirable implies that this figure could be much higher as in 65% of the countries observed entrepreneurship is perceived a desirable career choice. On average three natural disasters occurred in the countries observed, 584 people got killed and 3.733 got injured, almost 3 million people were affected by these disasters and almost 29.000 people became homeless. On average the damage of these disasters was $2,7 million annually. As can be seen from the range the disaster-related variables might have large outliers as the mean is low and median and range are high, implying that most countries experience little disasters (keeping the mean low) and a few countries experience terrible disasters (making the median and range higher). On average 80 people were killed by violence and 55 people became wounded. This indicates the data lacks some observations as one might assume that in a war more people get wounded than killed. For violence-related disasters the same holds as for natural disasters; only a few countries experienced these disasters. The variables used show the following correlation coefficients:
Developing TEA Desirable F. of Fail. Impdr. Dr. Necc. Dr. Perc. Sk. Perc. Opp.
Developing 1 TEA 0,4944** 1 Desirable 0,4225** 0,5430** 1 Fear of Fail. -0,1887 -0,3135** -0,25** 1 Impdr. Dr. -0,4666** -0,2201** -0,2631** -0,0191 1 Necc. Dr. 0,5470** 0,3489** 0,3485** -0,0405 -0,6991** 1 Perc.Sk. 0,3807** 0,7233** 0,6349** -0,4310** -0,2217** 0,2722** 1 Perc. Opp. 0,2589** 0,5838** 0,4954** -0,2873** 0,0859* -0,0652 0,6394** 1 #disasters 0,0648* 0,1008** 0,0119 -0,1010** -0,093* 0,1724** -0,0251 -0,0465 #killed 0,0369 -0,0652 -0,0486 0,056 -0,0664 0,0593 -0,1305** -0.0911* #injured 0.039 0.1788** 0.042 -0.050 -0.010 0.059 0.065 0.041 #affected 0.0947** 0.0866* 0.057 -0.035 -0.1448** 0.1964** -0.068 -0.014 #homeless 0.0784** 0.027 0.035 -0.038 -0.049 0.1175** 0.025 0.050 Total Affected 0.0956** 0.0872** 0.057 -0.036 -0.1452** 0.1973** -0.068 -0.014 TA in % Pop. 0.1868** 0.1672** 0.1367** -0.050 -0,0836* 0.1891** 0.066 0.0915** Total Damage -0.0988** -0.008 -0.1379** -0.024 0.1018** -0.028 0.075* -0.0808* TD in % Pop. 0.051 0.057 0.059 0.031 0.039 0.017 0.027 0.058 #Deaths 0.055 0.1014** 0.068 -0.059 -0.080 0.1064** 0.054 0.017 #Wounded 0.1468** 0.1423** 0.0816* -0.053 -0.039 0.069 0.1705** 0.1249** D. Dev. 1.0000** 0.4944** 0.4225** -0.1887** -0.4666** 0.5470** 0.3807** 0.2589**
22 #disasters #killed #injured #affected #homeless Tot. Aff. TA % Pop. Tot. Dam.
#disasters 1 #killed 0.2387** 1 #injured 0.0968** 0.1783** 1 #affected 0.5021** 0.1982** 0.0940** 1 #homeless 0.3025** 0.4323** 0.1007** 0.1787** 1 Total Affected 0.5048** 0.2037** 0.0981** 0.9999** 0.1911** 1 TA in % Pop. 0.2528** 0.0902** 0.0866** 0.4245** 0.1113** 0.4250** 1 Total Damage 0.3823** 0.3279** 0.0797** 0.1572** 0.062 0.1579** 0.0872** 1 TD in % Pop. 0.0948** 0.1508** 0.024 0.044 0.1494** 0.046 0.2458** 0.3381** #Deaths 0.0804** -0.005 -0.009 0.001 -0.014 0.000 0.020 0.034 #Wounded 0.1141** 0.007 -0.007 0.004 0.021 0.005 0.1367** -0.005 D. Dev. 0.0648* 0.037 0.039 0.0947** 0.0784** 0.0956** 0.1868** -0.0988**
TD % Pop. #Deaths #Wounded D. Dev.
TD in % Pop. 1
#Deaths 0.1204** 1
#Wounded 0.036 0.2699** 1
D. Dev. 0.051 0.055 0.1468** 1
** indicates 5% of significance, * indicates 10 of significance Table 4 - Correlation matrix
Table 4 indicates that many of the variables describing entrepreneurial decisions have a significant correlation (at the 5% level) with the variable Developing. The correlation between TEA and Developing even appears to be one of the highest correlations of the table. This indicates that there is a difference in the perceptions of starting up a business (in this case concerning the motivation, the perceived opportunities and the perceived skills) in developing and developed countries. TEA also shows significant correlation with the variables describing disasters (Total Affected, Total Damage, Deaths and Wounded) indicating slightly higher entrepreneurial activities in disaster-struck regions. Besides, Developing shows significant correlation with the same variables, indicating that there is probably a similarity between the developing countries studied in the literary review and countries struck by disasters. The table also confirms a significant correlation between variables describing entrepreneurial motivations (Improvement Driven and Necessity Driven) and the variables describing natural disasters (Total affected and Total Damage), indicating that natural disasters have a slight negative effect on the entrepreneurship out of opportunity and have a slight positive effect on entrepreneurship out of necessity. The table also shows significant correlation with the variable describing the amount of disasters (#disasters) and the variables on deaths and damage from violence and natural disasters. Lastly, it is remarkable that the variables describing
23 violent disasters (Deaths and Wounded) have often less significant correlations compared to the variables describing natural disasters.
3.2 Methodology analysis
It took quite extensive explorative statistical research to come to a proper methodology in order to reach a set of figures ready to be analyzed. This is mainly due to various factors that are at play, of which little statistical literature or guidance is available. First of all the direct and indirect effect described in the literature section both influence entrepreneurs. The indirect effect ‘drags’ on for several years and needs to be measured using a lag. Second, there is a difference in perception (and effect in statistical analysis) between absolute values and relative values (for example; analyzing damage in dollars or damage as a percentage of GDP), as studies indicated that absolute values have a higher impact on people’s perception of effects of disasters (Messner & Meyer, 2005). Third, the variables perceived opportunities, perceived skills, fear of failure and desirable are correlated to the disaster-indicating variables as the correlation matrix shows, and therefore they might play a role in the regression. However, their impact and relevance was still unclear at the start of this study. Before the figures were actually used for analysis all exogenous variables have been analysed in panel regressions by both the absolute figures (concerning the people affected and the damage in USD) and by relative figures (people affected as a percentage of the total population and damage as a percentage of the countries’ GDP), as a robustness check on measurement methods. Contradicting literature the absolute figures appeared to have more significant and higher coefficients and
therefore to avoid confusion solely these results will be treated in the analysis section (relative results are available in the appendix). Also, in the indirect effect regression the variables Desired and Fear of Failure have been added in one extra regression and in another the variables Perceived Opportunities and Perceived Capabilities have also been added, as a robustness check. The analyses using all four robustness check variables appeared most significant and showed the largest
coefficients, these will therefore be used in the analysis-section. Last, robustness checks for the lags are tested. For convenience lags from zero to five years have been tested but only those of three lags will be treated (the rest is available in the appendix) as these showed the highest significance with the largest coefficients.
3.2.1 Direct effect
The variables have first been analysed in the ‘direct’ approach in order to form an opinion on the direct effect of disasters. The direct effect is measured using the regular Stata-command for panel analysis (.xtset id year and .xtreg) and lags have been imposed by creating a new variable
24 (command: xgen L1var = variable[_n-1] for one lag). The dependent variables (TEA, Imprdr, Neccdr) have been analysed using the disaster-related variables (Total Damage, Total Affected, Deaths, Wounded). In this case panel regressions are carried out in the following manner: y is the
dependent variable, x1i,t is the exogenous variable, t-j is the chosen lag. The equation is performed for five lags. The equation looks as follows:
n = {1, …4}; i = {1, …, 65}; t = {1, …, 13} (1)
n = {1, …4}; i = {1, …, 65}; t = {1, …, 13}; j = {1, …, 5} (2) The difference between equation (1) and (2) is that in the latter also the disaster-related control
variables are included as a robustness check. Analysis with and without these variables have been carried out in order to determine their statistical significance and therefore relevance. As will be explained, including these variables improved the quality of the regressions (higher coefficients with higher significance values).
3.2.3. The cumulative effect
The direct effect analysis showed few significant coefficients for years beyond one used lag. This is an indication that further lags (possibly incorporating different effects) might have to be analysed using a different methodology. Therefore, lags after one year have been analysed using a method measuring the cumulative effect; this method incorporates lags, but also takes into account the period between the lag and the base year (also called a rolling window), as also effects from these years might still affect the entrepreneur (in the literature section referred to as the indirect effect, for example macro-economic events sparked by the disaster but accounted for a few years after). The analysis for the indirect effect will be performed using the dynamic panel data estimation command in Stata (.xtdpd-command) which is a method derived from the Arellano / Bover (.xtdpsys) and Arellano / Bond (.xtabond) models. Using the dynamic panel data estimation it is possible to work with low-order moving-average correlations in the idiosyncratic errors and predetermined variables. Besides, the .xtdpd-command allows analysing panel data with a lagged effect of the descriptive statistics. All data is analysed according to the following expression:
∑
{ } { }
Where Y are the variables explained by the model (TEA, Neccdr, Impdr), xit is a vector of exogenous variables (Total Affected, Total Damage, Deaths and Wounded), B1 is an estimated coefficient, wit is a
25 vector of control variables (Desired, Fear of Failure, Perceived Opportunities and Perceived Skills), B2 is an estimated coefficient of the control variables, vi are the panel-level effects and eit come from a low-order moving-average process. A lag of t years is included. In total this equation is used 108 times as for every one of the explained variables (TEA, Neccdr, Impdr) the regression is performed using absolute and relative figures for 0 to 5 lags, in three different manners (one without control variables, one including Desired and Fear of Failure and one that also includes Perceived Opportunities and Perceived Skills). Stata interprets the formula as follows:
(4) .xtpd is the command for a linear dynamic panel-data estimation. In this case the command concerns the dynamic panel data regression on the list of variables specified. The model has been elaborated with the following aspects: .hascons is a check for collinearity, .dgmmiv is a GMM-type instrument for the difference equation, which is in this case restricted by the Lagrange function. The Lagrange restricts the lags to the first lag used and the last; there is thus a cumulative effect of the lags used (as in real-life the effect of an event does not disappear after one year). .artests2 is a robustness check for the maximum order of AR-tests; concerning autocorrelation.
From this cumulative effect analysis an interesting result appeared. For some variables analysed the observed effect drastically changed after one year. This is an indication obviously of the direct and indirect effect that have different effects on the entrepreneur. However, as can be seen in the figure below when using the cumulative effect methodology in year 2+t one also accounts for the (contradicting) effects in year 0-1, thereby weakening the results of the indirect effect:
-0.03 -0.02 -0.01 0 0.01 0.02 0.03 0 1 2 3 4 5 Co e ff ic ie n t LAG
Coefficients People Affected and Damage (in USD) on TEA
People Affected Damage (In USD)
26
3.2.4. Combined effect
To overcome this weakening result, a third analysis has been carried out. The combined effect analysis is a simple panel analysis that incorporates the regular variables without a lag used in the direct effect analysis and the same variables with a three-year lag. This concerns equation (2), including a lagged effect of the disaster-related variables. By using this method, both the direct effect and the indirect effect should be distinguished from each other.
The combined effect is included in the analysis-section in a table that also shows the results of the direct effect analysis and the cumulative effect analysis. The direct effect is shown without a lag, the cumulative effect at a lag of three years (using the rolling window, so also incorporating years 0, 1 and 2) and the combined effect with the variables at a lag of three and at a lag of zero. The variables are shown at the absolute effect and including all control variables (Perceived Opportunities,
Perceived Skills, Fear of Failure, and Desirable). The other regressions – at all other lags from years 0 to 5, in both relative and absolute figures and with and without robustness check variables - are included in the appendix. As an indication of the available data the calculated coefficients on the absolute values of TEA is included:
TEA Direct Effect Cumulative Effect Combined Effect
#Affected (in Mil.) 0.0123 0.0176 0.0112
(0.0120)** (0.0147) (0.0808)
Dam. (in Mil $) -0.0295 0.00891 0.0078
(0.0143)* -(0.0078) (0.0104)
#Deaths 0.0006 -0.0006 0.0010
(0.0007) (0.0007) (0.0007)
#Wounded -0.0006 -0.0018 -0.0002
(0.0007)** (0.0008)** (0.0007)
#Affected (in Mil.) – L1 0.0175
(0.0147)
Dam. (in Mil $) – L1 0.0088
-(0.0079)
#Deaths – L1 -0.0006
(0.0008)
#Wounded – L1 -0.0018
(0.0008)**
#Affected (in Mil.) – L2 0.0171
(0.0147)
Dam. (in Mil $) – L2 0.0089
(0.0078)
27
(0.0008)
#Wounded – L2 -0.0018
(0.0008)**
#Affected (in Mil.) – L3 0.0250 0.0087
(0.012* (0.0110)
Dam. (in Mil $) – L3 0.0090 0.0300
(0.0060) (0.0099)**
#Deaths – L3 0.0000 0.0008
(0.0000) (0.0005)
#Wounded – L3 0.0010 0.0007
(0.001)* (0.0006)
#Affected (in Mil.) – L4 0.0187
(0.0148)
Dam. (in Mil $) – L4 0.0010
(0.0007)
#Deaths – L4 -0.0007
(0.0007)
#Wounded – L4 -0.0018
(0.0008)**
#Affected (in Mil.) – L5 0.0192
(0.0148)
Dam. (in Mil $) – L5 0.0009
(0.0008)
#Deaths – L5 -0.0005
(0.0007)
#Wounded – L5 -0.0001
(0.0008)**
** indicates significance at the 5% level, * indicates significance at the 10% level. Robust standard errors are shown in parentheses
28
4. Results & Analysis
I will treat the effects according to the order of the hypotheses. The variables TEA, Imprdr and Neccdr will be analysed through a table showing direct effects without a lag, and cumulative and combined effect at a lag of three years. A lag of three years is chosen as this is for the variable TEA in general the lag with the most explanatory power (in the size of coefficients and their significance), and this variable is the most important for this study. For convenience and comparability the other variables are also shown at a lag of three years. The other variables show slightly more significant effects at lags other than three years, but as their qualitative effects (positive or negative impact) do not change this does not influence the conclusion. The analyses at other lags are visible in the appendix. For simplicity the coefficients of control variables have not been included in the table. Perceived opportunities, Perceived Skills, Desirable and Fear of Failure have been included in the regression. All the graphs shown in this section are created with coefficients of the cumulative effect analyses.
Hypothesis 1: A natural disaster increases the probability one decides to become an entrepreneur. The variable TEA will be examined to provide an answer to Hypothesis 1. The coefficients of natural disasters are shortly after the event low, and for the variable Total Damage the effect even starts negative. The coefficients have a peak after three to four years. The coefficients of violent disasters are high shortly after an event has struck and decrease over time:
-0.03 -0.02 -0.01 0 0.01 0.02 0.03 0 1 2 3 4 5 C o e ff ic ie n t Lag
Effect natural disasters on TEA
#affected damage (in $)
-0.002 -0.0015 -0.001 -0.0005 0 0.0005 0.001 0 1 2 3 4 5 C o e ff ic ie n t Lag Effect violence on TEA
#deaths fatalities#wounded
