The social environment of households and its impact on financial decision-making

Table of Contents

List of Abbreviations

List of Figures

List of Tables

1. Introduction

2. Social network structures as a “cushion” against various risks
2.1 The pivotal role of informal finance in developing markets
2.2 Risk-sharing agreements between households
2.3 Insurance within family networks
2.4 Cultural influences on decision-making under risk

3. The consequences of social support on the financial risk-taking of households
3.1 Social finance
3.2 The interplay of individual risk tolerance and stock market participation

4. Methodology
4.1 Hypotheses
4.1.1 Hypothesis 1
4.1.2 Hypothesis 2
4.2 The Panel on Household Finances (PHF)
4.3 Summary statistics

5. Empirical analysis
5.1 Regression results
5.1.1 Financial risk tolerance
5.1.2 Stock ownership
5.1.3 Stock percentage of financial assets
5.2 Robustness check
5.3 Limitations of the empirical research

6. Conclusion

References

Appendix

List of Abbreviations

Abbildung in dieser Leseprobe nicht enthalten

List of Figures

Figure I: Distribution of financial risk tolerance

List of Tables

Table I: Summary statistics

Table II: Investment behavior

Table III: Generalized ordered logit regression model

Table IV: Logit regression model

Table V: Fractional logit regression model

Table VI: Instrumental variable regression

Table AI: Variable description

Table AII: Brant test

1. Introduction

Understanding when households are disposed to make risky decisions is a question of great interest for economists, sociologists, and behavioral scientists. As daily life becomes increasingly complicated, the need to make choices and decisions arises more frequently. Whether buying a new apartment, making a financial investment or accepting a lucrative job offer – individuals confront significant decisions almost every day, in various forms. One thing that all decisions share is an association with a certain amount of risk and a potential impact on individuals’ personal circumstances.

People often carefully weigh whether risks should be taken and what consequences they might have. As a result, a single person thinks differently than the head of a family, who has a responsibility as a household provider.¹ To safeguard against risks, many households seek a form of protection against possible losses. One way to mitigate substantial risks is by concluding insurance policies or setting up savings accounts as a buffer against shocks.² Not all households have this option, however; others may trust another safety net in times of trouble, such as assistance from family and friends.³

In the developing world, in particular, social network structures play a major role because poorer sections of the population often lack access to financial institutions.⁴ For this reason, cohesion among families and individuals is generally more pronounced, and informal risk-sharing agreements are widespread.⁵ These obligations help households to cope with income fluctuations resulting from severe shocks, which could otherwise threaten families’ livelihoods. At the same time, it is perceptible that informal risk-sharing between families leads to changes in risk behavior. Research has shown that households with better risk-sharing agreements are more risk-taking.⁶ This exemplifies the extent to which network structures determine people’s everyday lives, including their decision- making behavior.

Nevertheless, these observations can also be seen in parts of the world where the family serves as a resilient safety net in the case of unforeseeable problems like unemployment or divorce.⁷ The family is often the first place of contact as well as a problem solver, offering both emotional and practical support. Consequently, it is not astonishing that in recent years, researchers have sought to examine the motives plus differential forms of intergenerational transfers and have sight to provide deeper insight into the field of family insurance networks. It has been found that, alongside financial transfers, parents prefer to support their children by allowing them to move back home.⁸ Particularly in times of global insecurity, this trend is becoming more prevalent. According to Kaplan (2012), children adjust their job search accordingly when the offer of co-residence with parents exists.⁹ Furthermore, it also becomes evident that children can take longer with the search for the ideal job or accept riskier job opportunities if they can rely on the support of their parents.

The aforementioned approach is supported by the seminal “cushion hypothesis”, which considers the consequences of a tightly-knit family network on the risk-taking of financial market participants.¹⁰ In a groundbreaking study analyzing the risk behavior of American and Chinese citizens, Hsee and Weber (1999) observed greater financial risk-taking among the Chinese, although the contrary is often assumed. The authors explain their findings by pointing to the social background of the participants; in concrete terms, they expound that the emphasis on cohesiveness and mutual dependency in collectivistic societies has a demonstrable influence on people’s perception of risk. In a collective society, individuals can generally seek support from the extended family or other in-group members in the case of financial constraints. Keeping this in mind, people tend to take greater risks, as the family functions as a social “cushion” in the event of a monetary loss. Social networks hence fulfill the role of an insurance channel that mitigates negative consequences. The financial market, in particular, is fraught with risks and many households hesitate to invest a large portion of their money in stocks.¹¹ Because consumers are usually not protected from loss, they are particularly cautious with their investments in financial assets.

This emerges clearly when one considers the activity of German households in the stock market. Even though Germany is the largest economy in Europe, individual stock market participation is comparatively low with nearly 24%.¹²,¹³ Whereby the percentage distribution in the United States is more than twice as high. The average German household is thus characterized by a more skeptical attitude towards risky assets.¹⁴ The reasons for this hesitation are multifaceted and are presumably a general expression of a strong need for security, which is widely disseminated within the German population. Therefore, it makes sense to determine to what extent households adapt their risk behavior in the wake of support from family and friends. The question at hand is whether the social support produces an increased feeling of security and, by extension, an enhanced inclination to financial risk-taking. Concretely, I peruse the research questions if households that can rely on a form of social assistance (I) possess a higher financial risk tolerance and, in conclusion, (II) invest their money more strongly in the equity market. The thesis extends the literature regarding the determinants of stock holding and provides one of the first investigations exploring in greater detail the role of the households’ social environment and its repercussions on risky investment behavior.

Certainly, it is challenging to study household finance since an accurate measurement of household financial behavior is often difficult.¹⁵ To test the hypotheses, the research data is drawn from the novel Panel on Household Finance (PHF), a representative survey by the German Central Bank. Here, randomly selected households were asked about various aspects of their finances in two waves, conducted in the years 2010/11 and 2014.¹⁶ Specifically, the panel represents a rich source of information on the income, investment behavior as well as demographic data of a huge number of German households. Noteworthy is that households were, moreover, surveyed about the financial backing in their social circle in case of an emergency. The dataset is unique in this respect and provides new insights into social support nets of households and is consequently ideally suited for the thesis.

The contribution of this paper is threefold. The first is that households are indeed affected by the social backing and display a considerably higher tolerance for financial risks. To put it in more precise terms, it can be observed that socially embedded households are with 7.8 percentage points (pp) less likely to be financially risk-averse in comparison to their “unprotected” counterparts. In addition, it is apparent from the empirical investigation that the fallback option is associated with a significant increase in the financial risk tolerance levels and is therefore consistent with the first hypothesis. In the second contribution, I elaborate statistical evidence for the claim that those households are also more inclined to bear risks on the actual financial market. By using a wide variety of regression models, it is possible to establish a clear connection between the main independent variable and stock ownership. Third, it can be noted that this effect of the social environment is, moreover, reflected in the composition of financial assets, notwithstanding that the protected households only invest slightly more capital in high- risk assets.

The thesis is thereby structured as follows: The first two chapters provide a comprehensive review of existing literature in this field. First, I present a closer examination of social network structures and their role as a risk buffer. This lays the foundation for the next section of the paper, which further outlines the consequences of social support on household stock market participation. Subsequently, I undertake the data evaluation and describe both the data basis and the respective variables in detail, with a special focus on the investment pattern of German households. The fifth chapter deals with the empirical interpretation of the PHF data with the aid of multiple regression analyses in order to examine both hypotheses. The validity of these results is reviewed by a robustness check with due regard to endogeneity concerns applying the innovative Lewbel (2012) method.¹⁷ A discussion of the results together with a critical evaluation then concludes the final chapter of the paper.

2. Social network structures as a “cushion” against various risks

2.1 The pivotal role of informal finance in developing markets

To assess the role that social networks play in people’s everyday lives, it is worthwhile looking at the numerous studies that address informal finance mechanisms. Although the existing studies are primarily based on data from developing countries, the research still provides vital insights into the functioning of social groups¹⁸ under extreme conditions and helps us to understand how financial risks get mitigated. This becomes evident when one considers the circumstances that confront individuals in the poorer regions of the world. Individuals not only face the challenge of extremely low income, but they are also confronted with the problem of lack of access to banks and insurance companies, who may refuse their services to low-income individuals.¹⁹

In some cases, no options remain except mutual financial support among community members, which functions as a kind of safety net. This is referred to as “informal finance” because the financial support within the network replaces the services of formal institutions. Thus, the informal financial sector may be understood as a market response to the economic environment of low-income individuals without access to formal banking services. Although more prevalent in deprived regions, the size and growth of this informal financial sector are by no means smaller than the formal financial sector.²⁰ It is expected that the demand for informal credit and saving groups will continue to increase tremendously over the next years. As the gap between the poor and rich contains to widen, the victims of that development are forced to depend on each other more heavily. The phenomenon is visible worldwide and underscores the importance of both social networks and the informal financial sector.

Although the term “informal finance” is used frequently in the extant literature, a precise definition has proven elusive.²¹ In contrast to the formal sector, the structure of the informal financial sector is undoubtedly more complex and multifaceted. Depending on the regional conditions there are different channels that can be drawn on; those range from loans between related parties to organized cooperatives, which are distinguished by the fact that they do not comply to the banking rules and are neither regulated nor controlled by the government. Even though there are no official restrictions, the groups are very organized and often keen to set up rules for all members.

While banks require collateral from their customers, this is not necessarily the situation in the informal sector.²² Beyond that, the borrowing process often implies paying zero interest, with the rule of thumb being the smaller the amount, the higher the probability that this is the case.²³ Abraham and Platteau (1987) claim that so-called quasi-credit transactions are a quick and cheap method to get credit in rural communities.²⁴ These combine the advantages of interest-free credit with access based on individual need. Contingent loan repayments are also frequent, where borrowers can postpone their repayment or pay the debt in labor, allowing them to avoid falling deeper in debt.²⁵

Having said this, the prerequisite for credit-sharing between villagers is mutual trust amongst the participants. Whereas banks have algorithms which categorize borrowers according to their creditworthiness, no such option exists in the informal setting. Nevertheless, the contractual relationship is occasionally characterized by the fact that the partners know each other personally and, therefore, can build upon more detailed knowledge.²⁶ According to Banerjee et al. (1994), this can be considered an advantage because the lenders are more efficient in the monitoring process.²⁷

Defined as “peer monitoring view”, the process stands out for its lower monitoring costs. This result is congruent with the work of Stiglitz, who demonstrates that certain associated persons, such as neighbors, have better information about the personal circumstances of borrowers, which reduces moral hazard and adverse selection problems.²⁸ The contracts are, thereby, more state-contingent and less risky for both parties.²⁹ A more intimate position allows them to promptly react to potential financial difficulties and monitor the financial behavior of borrowers. An example might be a lowering of the default rate because of the productive use of the funds; likewise, in the opposite case, an increase is conceivable.³⁰

Constant monitoring is also one of the factors why people avoid going into debt in their immediate social environment. Empirical evidence confirms this suspicion and notes that groups with stronger ethnic ties are more willing to rely on formal finance procedures, while groups with weaker interconnections favor the informal approach.³¹ So, it is postulated that the punishment of recalcitrant group members may be regarded as unpleasant for participants. Since it turns out to be difficult to enforce contracts by official courts, other enforcement methods are used in rural communities.³² In the event of serious infringements, people might be excluded by kin, whereby both the punished person and remaining group members suffer.³³

The fact that financial help from kin gets associated with several shadow costs is also asserted in a qualitative analysis from Guérin et al. (2012).³⁴ Namely, members perceive informal finance as a source of mental stress due to the monitoring of expenditures and the social obligation to act reciprocally. This discomfort is mirrored in the circumstance that most of the households only rely in an emergency on informal capital and, hence, see the family more as insurance against consumption shocks, rather than as an investor. In addition, individuals distinguish between different loan sources: While loans from the social network are a popular method to cover health costs, for example, banks are used for significant expenditures such as house purchases.³⁵ Consequently, households think twice before they ask other in-group members for financial backing. Kinship networks, in particular, attach great emphasis to binding commitments since it is considered a privilege to be a part of the group.³⁶ This aspect marks a decided difference to friendship networks, where help is offered on a chiefly voluntary basis. It even goes so far that the kinship exerts pressure on wealthy individuals to share their income with the community.³⁷ This pressure acts as a kind of informal taxation which reduces a household’s incentive to save or invest money. Jakiela and Ozier (2016) confirm this in a real-world experiment in which participants proved willing to pay a fee in order to conceal their income.³⁸ A recent study even detected that people pretend to be impoverished to escape the pressure and signal the social circle that they are not able to react to financial demands.³⁹

In a similar vein, findings from Lee and Persson (2016) support the notion that informal finance – outside of developing countries – is perceived as a burden.⁴⁰ The authors demonstrate using a theoretical approach that entrepreneurs prefer assistance from regular financial institutions over those from family and friends, although the latter is considerably cheaper. Apart from the reasons mentioned above, discomfort stems from a lack of limited liability which borrowers are confronted with.⁴¹ A failure of an envisaged project strains the relationship with the backers to a different extent than financing via bank would have.

Interestingly, the authors demonstrated that borrowers adjust their risk behavior if informal capital was received before. Noticeable is the conservative approach to money handling, whereby risky investment opportunities are scaled back during the process. Therefore, if the primary objective is to grow, formal financing remains the method of choice. In conclusion, the authors propose that social tensions, which inevitably arise in the informal lending relationship, can be circumvented with the impersonal transaction. Neutral third parties or intermediaries can help to combine both social connections and financial transactions, but without the associated drawbacks.⁴² Taken together these findings suggest that many people take a skeptical view of informal financing, but are nevertheless obliged to participate in it, for lack of options.⁴³ It is made easy for them since social networks almost seamlessly replace the services of formal institutions. However, it is unclear whether this foundation can provide protection even in the event of unforeseen circumstances.

2.2 Risk-sharing agreements between households

No one can foresee what risks will affect them, or to what extent; but everyone has their own way of dealing with potential perils. In large parts of the world, the social network takes the role of an insurance channel against external events. For example, labor pooling is a common practice to insure households against severe health risks.⁴⁴ In a labor-pooling arrangement, neighbors or the community assist with workforce when one is absent due to illness and when there is a heavy reliance on the completion of time-sensitive tasks. Furthermore, the extended family helps with unemployment shocks by providing information about new job opportunities.⁴⁵ Another form of insurance was witnessed by Dercon et al. (2006) in Africa, where funeral societies foster children in the wake of a sudden parental death or offer a way to deal with expensive funeral costs.⁴⁶ Above all, it is in poorer households where social network support is essential.⁴⁷ As indicated in the previous chapter, households are not able to enjoy a steady income flow, because most of them make their living from agriculture, where fluctuations in weather or commodity prices directly translate into temporary income shocks.⁴⁸

Given this setting, it is astonishing that even though household income varies greatly, consumption is noticeably smooth.⁴⁹ Thus, a virtue was made out of necessity, and households came up with a myriad of mechanisms to counteract income shocks. These mechanisms can be diversified into two broader categories, which also function as substitutes for each other.⁵⁰ The first is known as income smoothing and describes the anticipation of adverse income shocks.⁵¹ This is achieved by, among other things, a more conservative approach to production or employment, as well as by diversifying economic activities. People are thus distinguished by a risk-averse approach, which is strongly reflected in their economic activities. The second technique to cope with risks is consumption smoothing, which allows households to obtain a more stable living standard. Concretely, households are dependent on external support and seek gifts, transfers, and credits from a network of family and friends or village residents as a buffer against income variability.⁵²

In this context, Dercon and Christiaensen (2011) speak of “risk induced poverty traps” and demonstrate that participation in risk-sharing agreements allows households to avoid these traps and develop out of poverty.⁵³ In contrast, individuals who have no such agreements not only seem to be imprisoned in an allegedly hopeless situation but also are highly vulnerable to fluctuations. The view is backed up by Kinnan and Townsend (2012), who prove that the presence of kin in a village significantly mitigates the fluctuation for investment if income shocks can be observed.⁵⁴ Financial support from kin or family members allows households to invest their money, even though they suffer from a shock. Current studies appear to support the notion that networks of blood and kin are the most central risk-sharing channel for households.⁵⁵ Individuals prefer to supply members of the same ethnic group with credit, in order to remain faithful to the kin group’s roots.⁵⁶ Ambrus et al. (2014) claim in their theoretical model that informal risk-pooling is more pronounced under socially closer agents and persons with similar risk attitudes.⁵⁷ The same conclusion is reached by Attanasio et al. (2012), who prove in a dyadic regression analysis that people are more likely to form risk-sharing groups with family and relatives since the risk preferences of each member are familiar.⁵⁸ These results indicate that members group assortatively to their risk attitudes. Likewise, Fafchamps and Lund (2003) argue that market transactions vary with social proximity. Noticeable is that gifts mainly take place within social networks of friends, while informal loans are more likely to be provided by distant acquaintances.⁵⁹ On top of that, people with a bigger social circle find it easier to borrow money to alleviate the effects of income shocks.

Townsend (1994) was one of the first to explore whether these risks are fully shared.⁶⁰ He assumed that if risks are pooled efficiently, household consumption levels should not be affected since villagers share an aggregate constraint to their budget. The village setting is ideally suited for the investigation since the employment of questionnaires and sampling procedures allows a clear identification of social clusters.⁶¹ Additionally, it is rather difficult for people to freeride due to social monitoring mechanisms.⁶² Also, non- pecuniary punishments in the community ensure that households refuse to terminate their contracts, whereby no enforcement problems arise. On that basis, it should be feasible for individuals to insure themselves against idiosyncratic risks. The data provided, nevertheless, convincing evidence against the hypothesis of a fully Pareto-efficient risk- pooling and showed that a substantial number of households were not insured.⁶³

Subsequently, numerous scholars turned their attention to hypothesis tests of full risk- sharing. In discussing borrowing and lending networks, Udry (1994) pointed out that people see it as a personal obligation to help their local social circle.⁶⁴ Consistent with the paper of Townsend (1994), he raised doubts regarding a fully efficient risk-pooling equilibrium in rural Nigeria. Most other real-world tests found a small but significant response of consumption to idiosyncratic fluctuations in income and rejected the null hypothesis of full risk-pooling.⁶⁵ Even outside the village setting, studies reached very similar results. In discussing the effectiveness of different informal insurance channels across the United States, Cochrane (1991) and Mace (1991) were also unable to show that all risks are shared efficiently.⁶⁶ The same is the case with risk-sharing across different countries⁶⁷ and, taken together, this has led to a rethinking of the empirical risk-sharing model. To date, ample researchers are still divided on which motives persuade households to share risks.

Coate and Ravallion (1993) explained the failure of full risk-sharing due to the limited commitment among individuals and assumed mutual insurance arrangements to be reciprocal. Thus, people first and foremost act in a selfish manner and only participate under the assumption of quid pro quo.⁶⁸ The authors modeled a two-sided game, in which risk-averse households were facing varying, uncorrelated income streams. Because income pooling is not feasible under this setting, both agents show a keen interest in self- enforcing risk-sharing agreements with the stipulation that monetary transfers are dependent on the realized income.⁶⁹ Although this model performs significantly better, it also encounters its limits and cannot explain the consumption distribution across households.⁷⁰

Another strand of literature argues that individuals are more typified by an intrinsic demeanor and not able to suppress their emotions in mutual insurance agreements. Accordingly, it can be inferred that people are altruistic and receive personal gratification when they share risk with others.⁷¹ Or, they have a feeling of guilt because they could not keep a promise and thence want to return the favor.⁷² Thereby, a kind of social collateral is present in informal arrangements, by which self-enforcement constraints get mitigated as the individual agents have no personal interest to break the contract.⁷³ Cox and Fafchamps (2007) suggest that risk-pooling is most efficient when groups are small and do not act out of self-interest.⁷⁴ This guarantees that information asymmetries are kept relatively small and freeriding gets eliminated. Following this assumption, risk-sharing contracts under altruism are more sustainable and minimize potential problems between the informal contractual partners. Since altruistic behavior is considerably stronger among genetically related persons⁷⁵, one might conclude that risks are more efficiently shared in the family.

This question was addressed by Altonji et al. (1992) in their formative paper, which sheds light on whether extended family networks can be viewed as one altruistically linked group that shares resources and risks altogether. To test this relationship, the authors studied if the consumption distribution between parents and children is independent of their income distribution.⁷⁶ Contrary to expectations, the altruism model was strongly rejected after reviewing the statistical data.⁷⁷ In an ensuing paper, the authors specifically examined inter-family risk-sharing in American households. But they, too, could not find any evidence for full insurance; although it does not mean that there is no form of protection within family networks at all.⁷⁸ Attanasio et al. (2015) went even further and improved the empirical model in order to measure whether the extended family provides full insurance against idiosyncratic shocks. What is striking about the results is that over 60 percent of these shocks are potentially insurable through the social network.⁷⁹ This statement is, however, relativized by the fact that no proof of a viable family insurance is found in the data and consumption smoothing is not discernible.⁸⁰ The authors attribute these outcomes to imperfect information, transaction costs or moral hazard problems⁸¹, but cannot answer the question of underlying causes with absolute certainty.

On that front, Mazzocco (2007) argues that household members are identified by individual preferences and are therefore not able to commit to future resources allocations.⁸² He was not the only one who perceived that the unitary household model reaches its limitations in explaining the insurance level. Macroeconomists abandoned the idea that preferences are shared among all members and that the maximization of the pooled family utility is of utmost importance.⁸³ More realistically the intra-family interaction was modeled as a dynamic non-cooperative game, where members have heterogeneous preferences and act strategically.⁸⁴ In a quantitative model, Laitner (1988) found that a form of moral hazard arises as certain risk-averse households show a more risk-loving behavior when they can count on intergenerational transfers.⁸⁵ In an overlapping generations (OLG) model, Barczyk and Kredler (2014) endorse this finding and document that the family demonstrably affects the risk-taking behavior of individual members. In this way, the family consists of two imperfect altruistic agents, where the poor agent acts regardless and undertakes riskier portfolio decisions on the assumption that the downside risk is limited.⁸⁶ Thus, the more affluent agent of the two would carry the losses and support the aggrieved person with transfers, while potential gains are enjoyed alone.⁸⁷

2.3 Insurance within family networks

The family often is the first port of call if things go differently than initially envisaged. Extensive literature indicates that parental altruism expresses itself in the form of intergenerational monetary and in-kind transfers, such as co-residence or transfers of time.⁸⁸,⁸⁹ In contrast to bequest transfers, where parents endeavor to distribute equally irrespective of income differences, inter-vivos transfers are preferably made to poorer children.⁹⁰ These support options allow children to save, consume and invest their capital even when they face severe budget constraints and are thus an important, but often overlooked economic factor.⁹¹ Current figures indicate that this market is vast. McGarry (2016) estimates the yearly flow of monetary assistance between parents and their non- resident children at $65 billion in 2010 dollars.⁹² Using the Panel Study of Income Dynamics (PSID) data from 2013, Attanasio et al. (2015) showed that over a third of adult children received monetary or time transfers from their parents.⁹³ In the opposite direction, the figures are significantly lower but do not detract from the fact that parents are also reliant on assistance from their children. This is usually the case when aging parents become dependent on their children for care. Therefore, children take on a substantial role, and quasi insure their parents against exorbitant health or caregiving expenses.⁹⁴

A clear pattern can also be seen in inter-vivos transfers. Predominantly, these are made during certain life stages of the children. In that regard, it can be noticed that the support is correlated with adverse income shocks, in order to smooth consumption.⁹⁵ McGarry (2016) distinguishes between several events in her empirical analysis and notes that a recent job loss or a divorce are among the strongest motives for co-residence or alternatively for a cash transfer. The intra-family support enables children to partly replace their regular insurance as the achieved level of protection is comparable.⁹⁶ Additionally, there are indications that the family operates as a kind of unemployment insurance, where an unemployment spell increases the likelihood of financial assistance from the family by nearly 50 percent.⁹⁷ This explains why areas with high unemployment rates have a greater probability of co-residence.⁹⁸

Numerous scholars claim that co-residence is the most preferred approach by the family to cope with redundancy and liquidity constraints – for the simple reason that residence sharing is much cheaper than equivalent monetary support.⁹⁹ The practice has already been used for a long time, which Matsudaira (2016) concludes after he examined the household data of American households and signified a coincidence of economic downturns and the rise of parental co-residence among young workers.¹⁰⁰ Beyond that, a look at the current figures of young home-living adults in the United States, which rose by nearly 15 percent between the years 2005 and 2014, illustrates that the demand is still high.¹⁰¹ Dettling and Hsu (2018) assert that the increase in parental co-residence and length of stay can be partially explained by the rise in student debt.¹⁰² The financial burden is so high that children do not want to slip further into the debt trap and prefer to save their rental payments for a faster repayment of obligations.

[...]

---

¹ Slovic (1987), pp. 280–281.

² Barasinska et al. (2012), pp. 1–3.

³ Ben-Porath (1980), pp. 8–12.

⁴ Banerjee et al. (2013), pp. 364–366.

⁵ Besley (1995), pp. 115–118.

⁶ Ambrus et al. (2014), pp. 162–167.

⁷ Edwards (2015), pp. 18–20.; McGarry (2016), pp. 10–12.

⁸ Wiemers (2014), pp. 2156–2160.

⁹ Kaplan (2012), pp. 467–469.

¹⁰ Hsee and Weber (1999), pp. 169–172.

¹¹ Campbell (2006), pp. 1558–1560.

¹² Badarinza et al. (2016), p. 116.; Bannier and Neubert (2016), p. 132.

¹³ The evaluated data in the thesis confirms the restraint on the equity market with 18.65% of households that directly hold stocks.

¹⁴ Statman and Klimek (2008), pp. 40–41.

¹⁵ Campbell (2006), pp. 1555–1561.

¹⁶ Deutsche Bundesbank (2014).

¹⁷ Lewbel (2012), pp. 67–70.

¹⁸ The mentioned social networks describe here village communities, as found in developing countries. Apart from that occupies the extended family or kinship an important place in the life for most of the people. The membership in these collective institutions is acquired by “bloodlines, marriage, or adoption” (di Falco and Bulte (2011), p. 1128.) and not seldom consists of several hundred members. Primarily they are spread in sub-Saharan Africa and a few Asian countries.

¹⁹ Collins et al. (2009), pp. 57–64.

²⁰ Montiel et al. (1993), p. 9.

²¹ Aryeetey (1998), p. 11.

²² Ayyagari et al. (2010), pp. 3048–3050.

²³ Collins et al. (2009), p. 49.

²⁴ Abraham and Platteau (1987), pp. 483–484.

²⁵ Fafchamps and Gubert (2007), pp. 653–657.

²⁶ Stiglitz (1990), p. 352.

²⁷ Banerjee et al. (1994), pp. 492–498.

²⁸ Stiglitz (1990), pp. 359–361.

²⁹ Guirkinger (2008), p. 1437.

³⁰ Stiglitz (1990), p. 352.

³¹ Anderson and Francois (2008), pp. 411–413.

³² Cox and Fafchamps (2007), pp. 3727–3728.

³³ Anderson and Francois (2008), p. 413.

³⁴ Guérin et al. (2012), p. 133.

³⁵ Ibid., p. 129.

³⁶ Coate and Ravallion (1993), pp. 7–13.; Collier and Garg (1999), pp. 133–135.

³⁷ Jakiela and Ozier (2016), p. 232.

³⁸ Ibid., pp. 250–256.

³⁹ Baland et al. (2011), pp. 8–15.

⁴⁰ Lee and Persson (2016), pp. 2362–2372.

⁴¹ Ibid., pp. 2375–2377.

⁴² Ibid., p. 2344.

⁴³ Collins et al. (2009), p. 16.

⁴⁴ Fafchamps (1992), p. 148.

⁴⁵ Munshi (2003), p. 562.

⁴⁶ Dercon et al. (2006), pp. 685–688.

⁴⁷ Banerjee and Duflo (2007), pp. 150–151.

⁴⁸ Morduch (1995), p. 105.

⁴⁹ Fafchamps and Lund (2003), pp. 261–262.; Townsend (1994), pp. 584–588.

⁵⁰ Morduch (1995), p. 104.

⁵¹ Ibid., pp. 104–111.

⁵² Townsend (1994), p. 540.; Rosenzweig (1988), pp. 1158–1160.

⁵³ Dercon and Christiaensen (2011), p. 160.

⁵⁴ Kinnan and Townsend (2012), pp. 292–293.

⁵⁵ Cox and Fafchamps (2007), p. 3717.

⁵⁶ Fisman (2001), pp. 315–321.

⁵⁷ Ambrus et al. (2014), pp. 170–178.

⁵⁸ Attanasio et al. (2012), pp. 156–158.

⁵⁹ Fafchamps and Lund (2003), pp. 274–285.

⁶⁰ Townsend (1994), pp. 561–570.

⁶¹ De Weerdt and Dercon (2006), p. 338.

⁶² Maaskant (2015), p. 3.

⁶³ Townsend (1994), pp. 561–584.

⁶⁴ Udry (1994), pp. 509–516.

⁶⁵ Grimard (1997), pp. 410–420.; Deaton (1992), pp. 265–273.

⁶⁶ Cochrane (1991), pp. 967–973.; Mace (1991), pp. 939–951.

⁶⁷ Obstfeld (1994), pp. 1320–1326.

⁶⁸ Posner (1980), pp. 23–28.

⁶⁹ Coate and Ravallion (1993), pp. 4–13.

⁷⁰ Ligon et al. (2002), pp. 239–242.; Bold (2009), pp. 579–585.

⁷¹ Andreoni and Payne (2003), pp. 796–800.

⁷² Platteau (1994), pp. 777–791.

⁷³ Karlan et al. (2009), pp. 1307–1310.; Foster and Rosenzweig (2001), pp. 390–396.

⁷⁴ Cox and Fafchamps (2007), pp. 3754–3775.

⁷⁵ Ibid., pp. 3727–3729.

⁷⁶ Altonji et al. (1992), p. 1178.

⁷⁷ Ibid., pp. 1188–1192.

⁷⁸ Altonji et al. (1996), pp. 280–288.

⁷⁹ Attanasio et al. (2015), pp. 30–31.

⁸⁰ Ibid., pp. 32–35.

⁸¹ Ibid., p. 4.

⁸² Mazzocco (2007), pp. 857–865.

⁸³ Heathcote et al. (2009), p. 339.

⁸⁴ Nishiyama (2002), pp. 897–899.; Mommaerts (2016), pp. 13–19.

⁸⁵ Laitner (1988), pp. 285–292.

⁸⁶ Barczyk and Kredler (2014), pp. 731–735.

⁸⁷ Ibid., pp. 705–707.

⁸⁸ Ben-Porath (1980), pp. 1–8.

⁸⁹ For a comprehensive review of the different intergenerational relationship types, see Bianchi et al. (2006).

⁹⁰ McGarry and Schoeni (1995), pp. 185–186.; Dunn and Phillips (1997), pp. 135–136.

⁹¹ Cox (1990), pp. 187–189.

⁹² McGarry (2016), p. 1.

⁹³ Attanasio et al. (2015), p. 7.

⁹⁴ Mommaerts (2016), pp. 43–47.

⁹⁵ McGarry (2016), pp. 10–13.

⁹⁶ Fogli (2004), pp. 12–25.

⁹⁷ Edwards (2015), pp. 8–13.

⁹⁸ Wiemers (2014), pp. 2155–2159.; Matsudaira (2016), pp. 182–191.

⁹⁹ Rosenzweig and Wolpin (1993), p. 85.

¹⁰⁰ Matsudaira (2016), pp. 176–186.

¹⁰¹ Dettling and Hsu (2018), pp. 225–227.

¹⁰² Ibid., pp. 229–235.