1 Introduction

According to the theory of warm glow giving (Andreoni 1989, 1990), charitable giving, defined here as giving money, time, or blood to a nonprofit organization to support a social cause, leads to positive psychological benefits (i.e. a warm glow) for the donor (Bekkers and Wiepking 2011). The term “warm glow” refers to the pleasurable feeling a donor experiences, which is a direct reaction to the act of giving per se and also distinct from the benefit enjoyed by the recipient (Andreoni 1989). From this standpoint, the sole act of giving to a charitable cause creates a feeling of happiness within the giving individual. Prior nonprofit research has validated the positive effect of giving behavior on happiness (e.g. Dunn et al. 2008; Harbaugh et al. 2007). For example, Dunn et al. (2008) provide evidence that spending money on others creates greater happiness than spending money on oneself.

Another stream of research proposes the opposite relationship between happiness and giving (e.g. Isen 1987; Rosenhan et al. 1974; Wang and Graddy 2008). In accordance with the positive affect theory (Isen 1987), the rationale behind this perspective is that happy individuals are more emotionally capable of engaging in giving behavior than relatively unhappy individuals. Scholars agreeing with this perspective argue that the higher emotional capability of happy individuals directly correlates with additional factors, which in turn positively influence giving behavior. In this line, several empirical studies have shown that happier people have a stronger tendency to engage in altruistic behavior, which ultimately translates into various types of giving behavior (e.g. Rosenhan et al. 1974; Thoits and Hewitt 2001).

Surprisingly, and despite the valuable insights from research in the field of happiness and charitable giving, the question of causality has been predominantly neglected. Notable exceptions are the studies of Meier and Stutzer (2008) and Aknin et al. (2012). Meier and Stutzer (2008) provide a unique study by comparing happiness and time donations in Eastern Germany before and shortly after the collapse of the German Democratic Republic. They conclude that when individuals are deprived of their opportunities to donate time, their happiness decreases accordingly. In a more recent study, Aknin et al. (2012) hypothesize and test a model of a positive feedback loop between prosocial spending and happiness. The conducted experiment provides support for this hypothesis and suggests that happiness and giving run in a circular motion. However, their findings are only of little guidance for researchers that try to identify the right causal direction for their conceptual models. Consequently, with respect to the still ambiguous perception of the relationship between giving and happiness, research that takes a bi-directional approach remains long overdue (e.g. Aaker and Akutsu 2009; Liu and Aaker 2008). This study addresses the literature gap by assessing the path directionality of this relationship through the application of Cohen’s path analysis (Cohen et al. 1993). The aim of the article is to explore the dominance of the causal relationship between happiness and charitable giving. For this purpose we provide empirical evidence from the German Socio-Economic Panel survey.

Against this background, the study makes at least two contributions. First, whereas previous studies have focused on either directional relationship between happiness and giving, this study integrates both perspectives simultaneously. To the best of our knowledge, no previous study has empirically tested this bi-directionality. Given the ambiguity in the respective literature and the lack of empirical evidence, this article advances research particularly on happiness and giving. Second, we apply Cohen’s path analysis for the first time in a happiness context to identify which effect is prevalent (Cohen 1988; Cohen et al. 1993). Cohen’s path analysis is a statistical tool that enables the investigation of unclear causal relationship. Although it is a powerful method for determining the causal direction of the relationship between certain constructs, researchers have rarely applied it. Therefore, the current study intends to apply and empirically validate this method in a stepwise and thorough approach.

2 Literature Review

2.1 The Concept of Happiness Through Cultures and Time

Throughout history human behavior has been motivated by the individual pursuit of happiness. Nevertheless, the academic researcher’s pursuit of unravelling the concept of happiness, at least from a theoretical and empirical perspective, is much younger. One of the notable earlier empirical studies on happiness was conducted by Watson (1930), who surveyed U.S. graduate students of education. He concluded that intelligence, grades, the parent’s wealth and education, as well as age are not related to happiness, while health, work, social relationships, and love of nature are beneficial to happiness. In a first substantial review of this emerging field, Wilson (1967) draws partially deviating conclusions in that being young, well-educated, well-paid, optimistic with modest aspirations contributes a person’s happiness. At the same time Wilson (1967, p. 302) agrees that “the theory of happy life has remained at about the level where the Greek philosophers left it.”

Much has happened since this very sobering review on the state of research on happiness. One major advancement in this field lies in the definition and measurement of happiness, which enables a clearer distinction and analysis of the components of happiness (Diener 1984). Happiness, or subjective well-being as denoted in most scientific literature, is a broad phenomenon that includes multiple aspects of an individual’s evaluation of life (Diener et al. 1999). It is composed of both an affective (e.g. positive or negative mood) and a cognitive (e.g. life satisfaction) component (Diener et al. 2003). However, although these components are distinguishable, they often correlate substantially (Diener et al. 1999). Thus, we use the term happiness as a synonym to subjective well-being and will measure it later on through life satisfaction, which is in line with previous research in the field (e.g. Aknin et al. 2013a; Leung et al. 2011).

Another major advancement is related to the origin of representative international studies such as the Gallup World Poll and the World Values Survey, which enable the comparison of happiness across countries and cultures. Whereas earlier studies were conducted in wealthier and westernized countries, these large samples make it possible to explore for universal predictors of happiness and differences between cultures and nations. For example, feelings of social support, trust and power predict happiness across the world, while other factors are strongly affected by societal circumstances, such as cultural norms and economic development (Diener 2012; Tov and Diener 2007). Most of the differences between nations can be attributed to societal characteristics, i.e. people are happier in rich nations that are “characterized by rule of law, freedom, good citizenship, cultural pluriformity and modernity” (Veenhoven 2010a, p. 615). In this line, Diener et al. (2010) found in their study that the relation between income and happiness was slightly stronger in wealthy than in economically underdeveloped nations. This finding is counter to the idea that money is more important for individuals with lower income, but in accordance with the idea that a higher income enables a higher fulfillment of material aspirations, which in turn is higher related to happiness in wealthier nations (Diener et al. 2010).

However, some differences in happiness between nations, which is also to an extent a difference between cultures (e.g. Hermans and Kempen 1998), can be attributed to culture-person congruence (Diener 2012). This means that when an individual has the characteristics that are valued in the respective culture, he or she tends to be happier (Diener 2012). For instance, self-serving biases, such as self-enhancement, occur less frequently among East Asians when compared to North Americans (Diener et al. 2003). Another example is the relationship of self-esteem and happiness, which is much stronger in individualistic cultures than in collectivistic cultures (Diener and Diener 1995).

In Germany with its rather individualistic society happiness is perceived as revolving around freedom, autonomy and hedonic pleasures (Pflug 2009). Being part of the Northern European region Germany is comparable with the Anglo societies, which are both most successful in their economic development (Diener et al. 2010). Thereby, the German citizens enjoy a fairly high standard of living. As a whole the German society is ranked 29 in a global comparison of 149 nations with an averaged happiness score from 2000 to 2009 of 7.1/10 (Veenhoven 2010b).

2.2 The Meaning of Donations

The charitable giving behavior of people can have several forms. The most prominent acts of giving are donations of money and time (e.g. volunteering). While these two types of giving are interrelated to a certain extent they have different characteristics (e.g. Steinberg 1990; Wang and Graddy 2008). The monetary giving process tends to be more transactional and bureaucratic, while volunteering on the other hand is usually perceived to be a more personal act of giving (Boenigk and Helmig 2013). However, notwithstanding the different means and costs of giving, donors in general tend to report greater happiness (Aaker and Akutsu 2009).

In a cross-national study on the effects of monetary donations and people’s happiness, Aknin et al. (2013a) investigated if this relationship represents a possible psychological universal, i.e. a core mental attribute shared by all human beings. In fact, they find first evidence that monetary donations are associated with greater happiness above and beyond country or culture specific traits. Another interesting finding revolves around the role of wealth in this context, which might constitute a cause for both monetary donations and happiness (e.g. Ball and Chernova 2008; Wang and Graddy 2008). When controlling for income and food inadequacy, Aknin et al. (2013a) find almost no change in the relationship between monetary donations and happiness. However, due to the available panel data we focus solely on monetary donations and not on others types of donations, such as time (see Meier and Stutzer 2008 for volunteering and happiness).

2.3 Literature Overview

In the scientific literature, the relationship between charitable giving and happiness is represented by two alternative perspectives. Figure 1 provides an overview of the articles addressing each alternative and their focus with regard to conceptual or empirical contributions. The selected articles in this comprehensive overview predominantly follow the rationale of one of these research streams. On the one hand, causality is considered to run from charitable giving to happiness, thus assuming that the act of giving creates happiness (Alternative 1). On the other hand, some authors argue that happiness causes charitable giving (Alternative 2). In the following two sections, we provide an overview of both alternatives and refer to the respective theory and empirical findings.

Fig. 1
figure 1

Rival perspectives on the causality between charitable giving and happiness

Full size image

3 Hypotheses Development and Conceptual Model

3.1 Alternative 1: Charitable Giving Causes Happiness

According to Andreoni (1989), the experience donors derive from their donation can be attributed to either purely altruistic motives (i.e. benefiting from the payoffs of another individual) or impure altruistic motives (i.e. the warm glow realized through the act of giving per se). Drawing from Becker’s (1974) work, Andreoni (1989, 1990) developed the concept of impure altruism due to the lack of predictive power by the pure altruism assumption. He argued that if the pure altruism assumption were true, government expenditures would completely crowd out voluntary donations to a point at which average donations would approach zero—none of which is commonly observable (e.g. Andreoni 1993; Harbaugh 1998). Consequently, the happiness donors experience through charitable giving is also induced by a private psychological value, such as the warm glow feeling (e.g. Andreoni 1993; Rose-Ackerman 1996). In warm glow models, a giving individual realizes a certain warm glow payoff due to acting in accordance with his or her ethically driven aspirations rather than other less costly alternatives (Cherepanov et al. 2013). Even when individuals perceive their contributions to a charitable cause as costly and nearly inconsequential, they will still choose to contribute if the warm glow payoff is sufficiently large (Cherepanov et al. 2013). In this vein, the term warm glow payoff could be perceived as more transactional and short-term, but we argue that these acts of giving stay in the memory of donors and therefore also contribute to a more long-term happiness.

Substantial additional evidence also lends support to the theory of warm glow (e.g. Aknin et al. 2013a; Dunn et al. 2008; Helliwell et al. 2014). For example, in their multi-national analysis of happiness Aknin et al. (2013a) find that citizens from countries with greater prosocial spending are happier than otherwise equivalent citizens. Dunn et al. (2008) find evidence that people who spend money on others are significantly happier than people who spend money on themselves. As indicated previously, these findings are also true for other types of donations, such as time and blood. Helliwell et al. (2014) show in their analysis of U.S. metropolitan areas that communities with greater prosocial engagement are happier than comparatively lesser engaged communities. In turn, Reid and Wood (2008) provide evidence for the theory of warm glow in their analysis of blood donors. In summary, the majority of research in this area demonstrates that charitable giving has a positive impact on happiness. Thus, we propose the following:

H 1 :

Charitable giving is positively related to happiness

3.2 Alternative 2: Happiness Causes Charitable Giving

The theory on positive affect (Isen 1987) suggests that individuals in positive affective states are substantially influenced in both their social behavior and cognitive processes. The positive affective state, which is induced by mild, everyday positive events, comes with a tendency to be more helpful, generous, and socially responsible (Isen 2008). Simultaneously, this positive spirit leads to improved proactivity (Bindl et al. 2012) and problem-solving behavior (Ashby et al. 1999). Increased attention to problems of others combined with enhanced problem-solving capabilities makes individuals in this condition more willing to engage in charitable giving. In support of this theory, Wang and Graddy (2008) reason that happier individuals show more emotional capability of helping others than people who are unhappy with their lives. They also argue that the overall more optimistic personality of happy people enhances additional characteristics that positively affect giving behavior.

In line with this rationale, early experiments manipulated the mood state of individuals, for example, through experiencing positive events (winning a game) or other mood-inducing stimuli (reminiscing about certain life events), and analyzed the effect on giving behavior (e.g. Isen et al. 1973; Rosenhan et al. 1974). In their experiments on happiness, Isen et al. (1973) induced positive and negative moods of young children by manipulating their success and failure in a certain game, which led to higher generosity of children in a positive mood. In a similar study, young children were asked to reminisce and talk about mood-appropriate memories and were then allowed to reward themselves with either candy or money (Rosenhan et al. 1974). While both happy and unhappy children rewarded themselves with more candy than the control group, only happy children gave more money to others (i.e. were more generous). Taken together, substantial theoretical and empirical evidence shows that happy people are more capable and willing to engage in giving behavior overall. Therefore, we propose the following:

H 2 :

Happiness is positively related to charitable giving

3.3 Determining the Dominant Alternative

Although research in the field of happiness and charitable giving has provided valuable insights, the question of causal direction has been predominantly neglected. A notable exception is the dictator experiment conducted by Konow and Earley (2008). Surprisingly, their initial analysis provides only weak evidence of a direct relationship between these constructs, leading to the possibility that a third factor causes both. In contrast with these findings, Aknin et al. (2012) provide empirical evidence for a direct relationship, while proposing a positive feedback loop between prosocial spending and happiness. In their experiment, participants who thought about an act of prosocial spending in the past reported higher levels of happiness and in turn were more likely to engage in future acts of prosocial spending. Consequently, it is important to investigate which causal path between the focal constructs is prevalent and thereby to identify the dominant causal direction. Following this rationale, we question whether the path from charitable giving to happiness (Alternative 1) dominates the path from happiness to charitable giving (Alternative 2) or vice versa. Unlike Aknin et al. (2012), who confirm a sequential positive feedback loop while assuming the path originating from prosocial spending to happiness, we explore which alternative causal direction is the dominant one.

Compelling arguments exist for both causal directions, and the dominance of a given path also depends on environmental conditions (Sun and Zhang 2006). That being said, we argue that Alternative 1 features the dominant causal direction. In other words, the causal direction originating from charitable giving to happiness will outweigh the alternative causal direction from happiness to charitable giving. Meier and Stutzer (2008, p. 53) critically assess the causal direction between volunteering and happiness and conclude that “volunteering does increase happiness.” Volunteering, as a form of prosocial behavior, rewards people in terms of greater happiness, while people who lose their opportunity to do volunteer work likewise experience a decline in happiness (Meier and Stutzer 2008). Thus, charitable giving, as a similar form of prosocial behavior, should function comparably.

Another argument for the prevalence of Alternative 1 is attributed to fundraising managers, who in their marketing efforts create awareness of the benefits of giving in the minds of donors. Prospective donors promised the psychological benefits of their donations tend to increase their generosity (Bekkers and Wiepking 2011). The dominance of these arguments in both economic and psychological domains offers further reason to believe that Alternative 1 represents the dominant causal direction. Accordingly, we assume that Alternative 1 dominates Alternative 2. Thus, we propose the following:

H 3 :

The path from charitable giving toward happiness dominates its alternating path (from happiness towards charitable giving)

3.4 Conceptual Model

Figure 2 represents the underlying conceptual model of this study. In the center of this model is the focal relationship between charitable giving and happiness. It includes both alternative path directions between the two constructs. To correctly apply Cohen’s path analysis, the focal constructs must be embedded in a nomological net with further well-defined causal relationships (Sun and Zhang 2006). We conceptualize eight relevant determinants to both charitable giving and happiness, though we keep the discussion of these additional and well-defined relationships brief because they are not the focus of the study.

Fig. 2
figure 2

Conceptual model representing determinants of both charitable giving and happiness

Full size image

On the one hand, we include four determinants that are predominantly discussed in the literature on charitable giving but are also possible determinants of happiness. First, pure altruism as a key driver of prosocial behavior naturally constitutes a strong predictor of charitable donations (e.g. Smith et al. 1995). The second determinant is income, which affects the financial flexibility required to make donations (Aknin et al. 2013a). Third, the level of an individual’s education also has positive ramifications on donations (Webb et al. 2000). Fourth, an individual’s religiosity can lead to helping behavior (Ranganathan and Henley 2008).

On the other hand, the conceptual model comprises four determinants predominantly discussed in the happiness literature. First, an individual’s health directly affects his or her mood and evaluation of life in general (Borghesi and Vercelli 2012). Second, the family situation and the social relations it entails represent a major part of people’s lives and are also likely to influence happiness (Borghesi and Vercelli 2012). Third, current job status affects an individual’s happiness (Booth and van Ours 2008). If an individual does not have a job or is unsatisfied with conditions at work, he or she will inevitably experience a decline in happiness overall. Fourth, an optimistic outlook on the future can similarly lead to happiness (Lopes et al. 2011).

As mentioned previously, the identified determinants are predominantly related to each of the focal constructs; however, because they are also potentially related to the respective other and account for possible spurious correlations, as indicated by Konow and Earley (2008), we allow for multiple paths. In this line, there are possible relationships among the predictor variables that have to be considered. For instance, the educational level could be related to job and income. To reduce model complexity we did not include all possible relationships between these predictors and instead test for potential collinearity issues between the predictors. The findings are reported and discussed in the results section.

Finally, we also test several donor characteristics to control for possible effects on the giving–happiness relationship. From both the literature on happiness and charitable giving we can derive that age, gender, and marital status might affect this relationship. First, given that happiness and donations are expected to increase with the age of donor’s, the focal relationship between giving and happiness is potentially stronger for older donors (Houston 2006; Lacey et al. 2006). Second, the giving–happiness relationship might also diverge between genders as for example women give differently than men (Andreoni et al. 2003). Last, the marital status could affect the strength of the giving–happiness relationship since married couples might experience greater happiness and giving behavior (Stack and Eshleman 1998).

4 Methodology

4.1 Data and Sampling Procedure

The data for this study came from the German Socio-Economic Panel (GSOEP). The GSOEP is a longitudinal study originating in 1984 that surveys German citizens about their socioeconomic status and various other aspects of life, such as their living situation, income level, and happiness (Wagner et al. 2007). Approximately 20,000 individuals respond to the GSOEP questionnaire on an annual basis, and the sample is representative of the German population. While a wide range of questions appear in the survey every year, others vary over time. Questions on happiness are collected on an annual basis, while questions on donation behavior accompany various items that were only available for the year 2009. As a result, we used the cross-sectional data of the GSOEP for the year 2009 to test the conceptual model. The initial data set from the GSOEP 2009 featured 17,928 respondents. To conduct the analysis, we derived a subsample from the original data set, which only included respondents who actually donated money in 2009. The exclusion of nondonors yielded a subsample of 6906 donors. Table 1 displays the characteristics of the donor subsample.

Table 1 Sample characteristics of the donor subsample (N = 6906)
Full size table

4.2 Measures

We consistently measured the presented conceptual model with single-item measures. Although there are valid concerns with the predictive validity of single-item measures (e.g. Diamantopoulos et al. 2012), some researchers argue that they are not necessarily inferior to multiple-item measures and highlight their practical advantages, such as minimization of respondent refusal and misperceptions (Bergkvist and Rossiter 2007), which is particularly true for panel data, such as the GSOEP. Table 2 provides the complete set of questions.

Table 2 Complete list of measures from the GSOEP 2009, including means and standard deviations
Full size table

We measured charitable giving with the total amount of monetary donations in the year 2009. The item on charitable giving is the only item derived from the GSOEP 2010. Respondents were asked the following: “How high was the total amount of money you have donated last year (2009)?”. Beforehand respondents were informed that donations was defined as the total amount of monetary donations to any kind of social, religious, cultural, or charitable cause in that year (excluding membership fees). Since respondents that take part in the GSOEP can be identified through their unique identification number, the respondents of the GSOEP 2010 can be directly linked to their responses in 2009. As mentioned earlier, due to the panel data, additional types of charitable giving were not subject to this study.

We assessed happiness through the respondents’ subjective evaluation of life (Aknin et al. 2013a). Specifically, respondents were asked “all in all, how satisfied are you currently with your life?”. As Diener et al. (2003) note, the study of happiness is always concerned with an individual’s perception of his or her life. In this vein, subjective measurement is necessary because people react differently to identical objective conditions depending on their personal expectation, values, and previous experiences (Diener et al. 1999). Thus, research in this field commonly relies on the subjective evaluation of the happiness phenomenon, given that it is primarily defined from the perspective of the individual (e.g. Frey and Stutzer 2002; Lyubomirsky et al. 2005).

During the time of the data collection, in year 2009, the German economy was still recovering from the subprime crisis that has stiffened the global economy. However, due to fundamentally good shape of the German economy, the initial fiscal stimulus and the strong foreign demand for German goods, the German economy recovered more quickly than other European economies (Funk 2012). As far as the happiness of German citizens is concerned, the financial crisis did not seem to have a lasting negative impact, instead happiness even started to slightly increase from 2009 onwards (Veenhoven 2010b).

For control purposes, respondent’s age and gender were simply collected through a question on the year of birth and gender, respectively. For family status, respondents chose among the following options: “Married, living together”; “Married, not living together”; “single”; “divorced”; and “widowed.” These categories were merged into “married” and “unmarried” accordingly (e.g. Andreoni et al. 2003).

4.3 Descriptive Results

The initial descriptive results for happiness reveal that donors are overall satisfied with their life’s with a mean of 7.2 (SD = 1.61). The size of the monetary donation differed substantially among donors, with donations ranging from €1 to €25,000. However, the average donation was approximately €246.6. Figure 3 illustrates the average donation per level of happiness in an overview.

Fig. 3
figure 3

Average donation in EUR per happiness level (N = 6906)

Full size image

These results show that comparatively happy people are, on average, more generous when it comes to monetary donations. Donors with a happiness level of 10 donated on average €516.9 in 2009 to charitable causes, while unhappy respondents (a happiness level of 0) donated €42.8 on average. Surprisingly, comparatively unhappy respondents with a happiness level of 1 still donated a relatively high amount of money (i.e. €514.4). This circumstance is attributed to the small number of observations and three cases that donated between €1000 and €2500 despite their unhappiness. These three cases also suffered from bad health (M = 3.67), which might explain their unusual charitable giving behavior similar to motives to leave a bequest (Sargeant et al. 2006). If we were to drop these three potential outliers, the average donation of respondents with a happiness level of 1 would drop to €142.8, as Fig. 3 indicates. However, we decided to keep these cases and conclude that the overall picture lends support to the idea that with increasing levels of happiness, people become more generous.

4.4 Analytical Approach

We conducted the quantitative analysis by applying Cohen’s path analysis method. The basic principle of Cohen’s path analysis is that the estimated correlations between the latent constructs derived from path analysis should be as close as possible to their actual correlations (Callaghan et al. 2007; Cohen et al. 1993). The alternative model that is closest to these actual correlations represents the dominant causal path direction and thus is preferred to the alternative causal direction.

We conduct the method in multiple steps (Sun and Zhang 2006). In a first step, we estimate the conceptual framework for both Alternative 1 and Alternative 2, to obtain the actual and estimated latent variable correlations. For this task, we chose partial least squares (PLS) path modeling over other structural equation methods for the following reasons. First, PLS path modeling is particularly viable in theory development, for example when identifying key target constructs (Hair et al. 2014; Ringle et al. 2012). Since it is not our primary aim to identify the best model, but are instead more interested in investigating the causal direction of the happiness-giving relationship, this makes PLS more suitable for our study. Second, PLS path modeling makes fewer assumptions about the underlying data, which makes it possible to work with non-normally distributed data as well as single-item constructs without risking identification problems (Hair et al. 2014).

The model analysis through PLS path modeling usually follows a twofold approach, i.e. the evaluation of the measurement model and the structural model (Hair et al. 2014). In our case the evaluation of the measurement model cannot take place through the average variance extracted and composite reliability due to the single-item constructs. However, the GSOEP applies global items, which summarize the essence of the according constructs. Moreover, we test for possible collinearity issues by calculating the variance inflation factors (VIF). In the context of PLS path modeling, VIF values of below 5 are still considered as acceptable (Hair et al. 2014). The VIF values for all predictor variables are below 2, thus collinearity should is not an issue for our model and sample data. The following assessment of the measurement model takes place through the evaluation each path coefficient and the R2 values. The statistical software applied to empirically test both models is SmartPLS 2.0 (Ringle et al. 2005).

In a second step, we calculate the differences between the actual and estimated latent correlations for every valid path—namely, their squared errors. We then compute the sum of all squared errors for both alternative models. After obtaining the total squared errors per model, we determine the error change achieved by switching from one path direction to the other. The alternative model, with the lower total squared errors, features the preferable causal direction (Callaghan et al. 2007). In the third and final step, we calculate Cohen’s d value to determine the effect size of the dominant path over its alternative causal direction. In this way, we can derive how much one alternative model is favored over the other.

5 Results of the Cohen’s Path Analysis

5.1 Step 1

Following the proposed method, we conducted the path analysis for both alternative models. To do so and to obtain viable results, we transformed the data on donations. Although PLS path modeling does not require normal distributed data, extreme deviations from normal may prove problematic in the assessment of the parameter’s significance (Hair et al. 2014). Therefore, we log-transformed the data on donations, which suffered from a substantially positive skew (14.14) and kurtosis (353.32), to achieve a more normal distribution (Manning 1998). The resulting log-transformed donation data highly improved (skew = .19; kurtosis = .21), thus becoming more viable for further statistical analysis. Figure 4 depicts the results of the PLS path analysis for both the original model (Alternative 1: giving → happiness) and the alternative model (Alternative 2: happiness → giving).

Fig. 4
figure 4

Results of the rival models

Full size image

For Alternative 1, the results reveal that all determinants except health had a significant impact on charitable giving. Income, with a path coefficient of .237, and education, with .187, had the strongest impacts. Regarding the explained variance of the endogenous constructs (R2), charitable giving had a rather low R2 value of .16. For happiness, all determinants except altruism and education had a significant impact. Health (.310) and family (.307) had the strongest impact, contributing to the moderately high R2 value of .42. The focal path from charitable giving to happiness had a weak but significant impact (.082); thus, H1 is supported by the data.

For Alternative 2, all determinants except family and optimism had a significant impact on charitable giving. Income and family still had the strongest impacts, with .232 and .184, respectively. The R2 value of charitable giving was .17. The results for happiness show that all determinants had a significant impact. Again, health (.310) and family (.310) had the strongest impact on happiness, resulting in an R2 value .41. The focal path from charitable giving to happiness was significant at .117; thus, H2 confirmed.

5.2 Step 2

After obtaining the path coefficients for both alternative models, we calculated the estimated path correlations for all antecedents. For this purpose, Cohen et al. (1993) provide a set of rules to determine all relevant paths: (1) there must be a path from every variable to the dependent variable; (2) a path may not go through a node twice; and (3) after a node has been entered through an arrowhead, it may not leave through an arrowhead. In a subsequent step, we calculated the differences between these estimated path correlations and the actual correlations, which we derived from the correlation matrix. This step was performed for every relevant path of Alternative 1 and Alternative 2 accordingly.

We squared these differences between estimated and actual correlations (i.e. error terms). Then, we calculated the sum of all squared errors [i.e. the total squared errors (TSE)] for both alternatives. Table 3 depicts the process and results. The total squared error value is .1262 for Alternative 1 and .0166 for Alternative 2. This means that by switching from Alternative 1 to Alternative 2, an error change of −86.8 % (.1262–.0166/.1262) can be achieved. In other words, by switching the causal direction between the two focal constructs of charitable giving and happiness, the total squared error is substantially reduced.

Table 3 The results of Cohen’s path analysis
Full size table

5.3 Step 3

In this final step we determined the effect size of choosing one alternative over the other. Cohen (1988) introduced the Cohen’s d formula: \({\text{d}} = \frac{{{\text{TSE}}_{2} - {\text{TSE}}_{1} }}{\sigma }\), where σ is the pooled standard deviation of both total squared error values (in our case σ = .035). A positive (negative) d-value indicates that Alternative 1 (Alternative 2) is preferred to the other, while a size of ±0.2 is perceived as small, ±0.5 as medium, and ±0.8 as large (Cohen 1988). In our study, the Cohen’s d value estimated for the difference between the two alternatives is −3.13; thus, the data do not provide support for H3. The conducted Cohen’s path analysis reveals that Alternative 2 (happiness → giving) strongly dominates Alternative 1 (giving → happiness); this result thus provides support for the perspective that happiness causes charitable giving.

To validate the robustness of these results we conducted a subsample analysis. To perform the analysis we divided the sample by gender and thereby achieved two comparably large subsamples of women (n = 3740) and men (n = 3166). We found that both the Cohen’s d value for the female-subsample of −3.16 and for the male-subsample of −3.10, were consistent when compared to the original value of −3.13. In sum, these subsamples support the overall picture, i.e. the path from happiness towards giving being the more dominant one.

5.4 Controls

Furthermore, we controlled for certain donor characteristics by means of multigroup comparison analysis, because these variables cannot be directly tested within the PLS path model (Sarstedt et al. 2011). As Henseler (2007) recommends, we applied a bootstrapping approach to compare the happiness-giving relationship across two subsamples, the results are provided in Table 4. For age we distinguish between younger donors (<40 years) and middle-/older-aged donors (≥40 ears), which is in line with previous research (Lindenberger et al. 2000). The significant P Henseler value indicates that the effect of happiness on charitable giving varied significantly between the respective groups. The happiness-giving relationship was substantially stronger for older donors (β = .128) when compared to younger donors (β = .024). The non-significant P Henseler for gender and marital status show that no significant group differences emerged.

Table 4 Results of the multigroup comparison analysis for the control variables
Full size table

6 Discussion

6.1 Research Issues

Taken together, the results show that (1) happiness and giving are positively related, and (2) support the perspective that the causal path from happiness towards giving is dominant. With this initial evidence, we contribute to the academic literature on happiness and charitable giving at least in three ways. The first contribution pertains to the lack of consensus on the path directionality between happiness and charitable giving. As Fig. 1 shows, two alternative perspectives are present in this stream of research. In contrast to the predominant perspective in the literature, this study provides empirical evidence from a sizable sample of German donors that the opposite perspective is more dominant—that is, happiness causing giving behavior. The sizeable Cohen’s d value indicates that this causal direction is strongly preferred over the opposing direction. Since only a minority of scholars in this field follow this perspective, this piece of evidence is likely to stimulate and contribute to a much needed discussion on path directionality. In particular, when we consider that the bi-directional relationship such as the one between happiness and giving can be easily overlooked by confirmatory methods. Methods like structural equation modeling are not sensitive to bi-directional relationships and thus mutually reinforcing relationships might stay undetected. Therefore we reinforce the existing belief that much more attention has to be directed to “why” or “when” instead of “what” relationships (e.g. Lee et al. 1997).

In this context, note that the results of this study contradict one of the few studies on causation between time donations and happiness; that is, Meier and Stutzer (2008) who concluded that the decline in happiness of East German citizens was partially attributed to the decline in opportunities to volunteer. However, this divergence may be due to the different type of giving (monetary vs. time donations) analyzed. Individuals who donate their time to a nonprofit organization are potentially more involved in the cause and organization. They may find different forms of benefits, such as belonging to a community or a career boost (e.g. Clary et al. 1998), than money donors. Overall, it is important to acknowledge that the relationship between happiness and giving is complex, and despite the popularity of the warm glow theory, causality could possibly go the other way. Nonetheless, this does not exclude the possibility that this relationship may form some sort of positive feedback loop in which these constructs reinforce each other, as Aknin et al. (2012) posit.

Second, we conduct Cohen’s path analysis in the research field of happiness. Cohen’s path analysis is one of multiple techniques to quantitatively assist in solving path directions. With its rare application in the past, this study contributes to the literature by providing a stepwise and thorough approach to this technique. Although such techniques may never “confirm” a given path direction, they may provide additional evidence and complement logical arguments. Sophisticated techniques, such as the Cohen’s path analysis, aid researchers in solving the question of path directionality. This method in combination with logical reasoning has the potential to determine causation of other ambiguous relationships.

The third contribution involves the determinants of an individual’s happiness. In accordance with prior literature, we confirmed substantial and highly significant effects of job status, family, and health. The only exception was the effect of income on happiness. Despite the significant result, the path (.042) was comparatively weak, particularly when compared with the effects of family (.310) and health (.310) on donors’ level of happiness. One explanation for the relatively weak effect of income is that we measured it on an absolute basis rather than as a relative measure. As Ball and Chernova (2008, p. 498) state, “money does buy at least some happiness, but having more than others around you matters more to happiness than simply having more.” Overall, it is not surprising that the domains of family and health greatly affected respondents’ happiness. Another interesting finding originates from the multigroup comparison. While gender and marital status did not significantly differ between the subsamples, age proved to have a strong and significant impact on this relationship. For older donors the relationship between happiness and giving was substantially stronger, which might be attributed to their increasing well-being and propensity to make monetary donations.

6.2 Limitations

This study provides a first attempt to determine path directionality and therefore is not without several limitations. First, the measurement of happiness (and other constructs) by single-item measures might not suffice to capture all facets of this broad construct. Although the global measurement of happiness through the evaluation of an individual’s satisfaction with life has validity, further research should combine multiple measures to cover all aspects of happiness (Diener 2000). At the same time, this does not necessarily mean that the predictive validity of the applied single-item measures is inferior to multiple-item measures (e.g. Bergkvist and Rossiter 2007).

Second, we assessed charitable giving through the amount of monetary donation in a given year. While nonprofit organizations heavily rely on monetary donations to sustain their survival, they also heavily depend on time donations of their volunteers (Moore 2000). Therefore, further research could explore whether the favored path direction herein can be transferred to additional types of donations, such as the donation of time or blood to a good cause. Third, from the comparatively small R2 of donations we can derive that not all antecedents were captured by the panel data. For example, from a motivational perspective, individuals engage in charitable giving, due to both altruistic and egoistic behaviors (e.g. Batson 2011). This raises the interesting question if happier people donate more money to feel better themselves (egoistic motive) or for the welfare of another (altruistic motive)? It could be possible that particularly happy individuals engage more in charitable giving behavior, because the resulting warm-glow payoff is higher than purchasing consumer goods for the equivalent amount of money. Such possibilities are very interesting and we believe there is still much to be gained by further investigating the happiness-giving relationship.

6.3 Managerial Implications

Facing the competitive pressures of today’s globalized fundraising market, nonprofit organizations are forced to focus even more on their performance and effectiveness (e.g. Hwang and Powell 2009; Willems et al. 2014). In this context, it is vital for nonprofit organizations to gain deeper knowledge of what drives the giving behavior of their donors. If they misinterpret the happiness-giving relationship wrongly, nonprofit organizations risk developing strategies and allocating resources not in the most effective way. According to our results, fundraising managers should consider to evaluate and use happiness as an additional instrument to develop promising donor segments. Moreover, happiness-oriented marketing measures could also contribute to fundraising success. For instance, infusing a positive mood of potential donors through minor gifts or mild gestures before approaching them might prove beneficial. Consequently, all services offered by nonprofit organizations should be evaluated for their contribution to a donor’s happiness and adjusted accordingly. In addition to the positive effect of happiness on charitable giving, nonprofit organizations benefit from happy donors in additional ways. Organizations that achieve the complex goal of enhancing the happiness of their donors create a competitive advantage while also increasing loyalty (Rosenbaum et al. 2011).

However, this does not mean that nonprofit managers should refrain from advertising the positive effects of charitable giving on the happiness of donors. Although this study found that the path originating from happiness to charitable giving was dominant, it still revealed a significant effect from charitable giving to happiness. Thus, nonprofit managers may still find it fruitful to advertise these benefits and thus complement their efforts in enhancing the happiness of their donors. This approach seems promising because charitable giving and happiness potentially run in a circular fashion, as Aknin et al. (2012) suggest. In summary, the focus of nonprofit organizations should shift to happiness-inducing services, while their communication efforts should highlight the positive effect of donations on the donor.