Study 1
In Study 1, we checked if increased Expectancy predicts declarations regarding readiness to help even when the desire to help is low. We thus ran an online study in which we presented participants with a helping scenario and recorded their declared intentions to help. Their desire to help (Want) and the Expectancy that the help would be impactful were measured as well. We investigated whether the interaction between the two factors predicts the declared magnitude of the help.
Participants
A priori power analysis with the use of G*Power 3.1 (Faul et al. 2009) showed that a sample of at least 153 would be necessary to obtain the assumed power. Since the study was conducted online, we increased the sample size by 30% to take into account possible attritions and non-attentive responding. Accordingly, we aimed to recruit at least 204 participants for this study.
Two hundred and nine registered users of the Research Online platform took part in the online experiment. There were 126 women and 83 men, aged between 18 and 84 (M = 33.94, SD = 11.35). Sixty-eight participants had a high school education, and 5 had occupational education; 21 were students, and 115 had completed their higher education at the time of the study. Participants were paid for their participation in line with the rules of the platform (approx. 2 EUR). All participants gave informed consent before participation in the study, and they were debriefed on completion of the study.
Fourteen participants failed our attention check (see below). Therefore, the final sample consisted of N = 195 participants (116 women, 79 men; with a mean of age M = 33.93, SD = 10.85).
Materials and procedure
In the study, we presented participants with a story of a child, or children, in need (it transpired that the version of the story was unimportant, see Supplementary Materials). Then, the participants’ desire to help, and expectancy in relation to the satisfaction of this desire, were measured. Want was measured using one item: To what extent is it important for you to help [the victim]? (1–7 scale anchored with not at all important and definitely important). We also used an additional item: Do you want to help in this situation? But we decided not to include it to the analysis as it is linked too close to measure of intention. We however re-analyzed data using both items and the results were the same. Alongside this, ExpectancyFootnote 2 (Cronbach’s α = 0.81, M = 5.08, SD = 0.93) was measured with four items: (1) In your opinion, by donating, will you effectively help [the victim]? (2) In your opinion, will the necessary amount be collected? (3) In your opinion, if the necessary amount is collected, will it be possible to save…? (4) In your opinion, are actions like this effective? (1–7 Scale anchored with definitely not and absolutely). Indices of Want and Expectancy were obtained by averaging responses to the respective items. The order of Want and Expectancy items was counterbalanced.
Finally, participants were asked about their donation declarations with the following question: What amount would you be willing to donate to this cause? (Participants were asked to mark an amount on a slider). The minimum amount was 0 PLN, and the maximum was 100 PLN, with responses possible in 1 PLN increments. We decided to set an upper limit to make sure that all participants operated within a given range and that their donation decisions would not be influenced by other factors (e.g., their socio-economic status).
In addition, we decided to control for the participants’ gender and attitude towards money as these variables may influence the declarations to donate or actual donation. Indeed, some research show that women are likely to give and give more than man in similar situations (Mesch et al. 2011; Cox and Deck 2006; Anderoni and Vesterlund 2001). Besides, it is already known that people differ in their attitudes towards money, that money attitudes are mostly independent from income, and money perceptions are an additional important factor in the understanding of charitable behavior (Wiepking and Breeze 2011). Thus, even though participants did not donate actual money, their decisions could be influenced by their attitude towards spending. Therefore, we asked participants to rate to what extent they agreed with the following statements: (1) I attach a lot of importance to money. (2) I appreciate the value of money. (3) I carefully think about each zloty spent. (4) Money is important in my life (1–7 scale from definitely disagree to definitely agree). Responses to all items were averaged and the resulting score was entered into the analyses as a statistical control (reported in the Supplementary Materials).
Since the study was run online, we also included an attention check. Specifically, we added a question in which we asked participants to mark response number 3.
Results and discussion
Means and intercorrelations between variables are presented in Table 1. As Want and Expectancy are correlated, we explored the variance inflation factor (VIF) to be sure that performing interaction is accurate (Freund et al. 2003). As VIF equals 1.37, we performed interaction analysis. To test the interactional hypothesis, we regressed Want, Expectancy, and the product of the two on the donation (log transformed prior to the analyses due to the skewness in the data, DV’s skewness = 1.28). The analyses were performed with the use of Process macro for SPSS version 2.13 (Hayes 2013). In all analyses, bias corrected bootstrap confidence intervals are reported; 10,000 bootstrap samples were used in all analyses. All variables were standardized prior to analyses.
Table 1 Descriptive statistics and intercorrelations The results showed that both Want and Expectancy significantly predicted declarations to donate, β = 0.48, t = 7.26, p < 0.001, 95% CI [0.35, 0.61] for Want, and β = 0.21, t = 3.23, p = 0.001, 95% CI [0.08, 0.34], for Expectancy. Importantly, however, there was a significant interaction, β = − 0.10, t = − 2.10, p = 0.04, 95% CI [− 0.19, 0.00]. Conditional effects analysis showed that, whereas there were no significant effects of Expectancy at high (+ 1 SD) values of Want, β = 0.12, t = 1.45, p = 0.15, 95% CI [− 0.04, 0.28], there was a significant effect of Expectancy at low (− 1 SD) values of Want, β = 0.27, t = 3.94, p < 0.001, 95% CI [0.14, 0.41]. The interaction is graphically presented in Fig. 1.
We re-ran the analyses while controlling for gender and attitudes towards money, and they yielded similar results (see Table 1 in Supplementary Materials).
The results thus demonstrated that, although a desire to help (Want) is a crucial factor influencing on declarations to donate, if it is low, Expectancy that the donation is effective matters as well. Specifically, when the desire to help was high, Expectancy did not seem to affect the donation. However, when the desire to help was low, Expectancy became a significant predictor, with high Expectancy levels significantly increasing the declared amount.
Study 2
The aim of Study 2 was to replicate the results described above in a laboratory setting. In addition, we also focused on actual helping behavior, not merely on declarations to donate as in Study 1. It is important extension of Study 1 as declarations made by people about how they behave might not always be consistent with their overt behavior (see Doliński 2018). Thus, it is crucial to demonstrate that our predictions hold also when actual helping behavior is taken into account. Therefore, we set up a laboratory experiment connected with a charity fundraiser in which participants could donate money earned in the lab to a person in need. As in the previous study, Want and Expectancy were measured, and we tested whether they interactively predicted participation in the fundraiser (i.e., in the form of offering help).
Participants
We ran an a priori power analysis for logistic regression with a power of 0.90 and odds ratio of 2. The analysis showed that a sample size of at least 106 participants would be necessary to obtain the assumed power. We intended to recruit more participants in case there were any data losses; we therefore endeavored to enroll at minimum 130 participants in the study.
We recruited 134 participants to take part in a laboratory study on solving cognitive tasks. The sample was comprised of young adults who answered an announcement at university websites and local internet portals. There were 111 women and 21 men (2 participants did not provide their demographics) and they were aged between 18 and 35 (M = 23.66, SD = 3.54). 63 Participants had received a high school education; 2 had occupational education, 2 basic education, and 65 higher education. 50 Participants were students, 20 were employed, 44 both studied and worked, and 18 were doing neither at the time of the experiment.
Participants were paid 20 PLN (around 5 Euro) for participation in the study. All subjects gave informed consent before participation in the study and were debriefed on its completion.
Materials and procedure
Participants were recruited to take part in a study on task solving. In this study, which took about an hour to complete, they solved a set of computer tasks of low to medium difficulty. Participants worked in sessions of up to six participants at the same time, each in a separate cubicle. After completing the task, they were paid the amount of 20 PLN (in one zloty coins; app. 4.5 Euro) and thanked for participation in the study. However, they were also informed that the study was run in cooperation with a local university charity organization which was currently raising money to help children in need. Each participant was presented with a description of the cause in writing, i.e., they read about a child suffering from a kidney disease (in one of the three versions, as presented in the Supplementary Materials; the version did not matter).
Participants were informed that if they decided to support the cause, their money would be given to the charity in full. They were also told that the fund-raiser was not part of the study and that donations were entirely voluntary so their decision to take part was entirely up to them. Nevertheless, all participants were asked to answer several questions, allegedly for evaluation purposes, even if they decided not to donate any money. The questionnaire aimed to measure the desire to help (Want), the expectancy that this desire would be satisfied (Expectancy), and to assess their attitude towards money. Since we wanted the participants to believe this was not a part of a study, we kept the questionnaire short and asked one question per each variable: To what extent is it important for you to help the child?—To measure Want; In your opinion, are campaigns like this effective?—To measure Expectancy; and I carefully think about each zloty spent—to measure their attitude towards money (each responded on a 1–7 scale).
Participants answered the questions in their cubicles so that they could have some privacy and not feel pressured in any way. After answering the questions, they were asked to put the questionnaire, along with the amount (if any) they were willing to donate, in the envelope and leave it on the desk (as participants were paid in single zloty coins, they could leave any amount ranging from 1 to 20 PLN). After completing the study, they were informed that they would soon receive detailed information regarding the purpose of the study and confirmation of their financial donation to the charity organization via email.
In the follow-up email, the purpose of the study was explained to them, and they were provided with more information about the charity organization to which the money was donated (the organization was the Association for Aid to Children with Kidney Diseases associated under the auspice of Professor Marta Uszycka-Karcz, www.nefrologiadziecieca.pl). A receipt for the money transfer was also attached.
Results and discussion
Means and intercorrelations between variables measured in this study are presented in Tables 1 and 2. As Want and Expectancy are correlated, we explored VIF index to be sure that performing interaction is accurate. As VIF equals 1.22, we performed interaction analysis.
Table 2 Mean differences for variables tested in Study 2 depending on decision to donate (N = 134) Since it has been argued that observed behavior, unlike declarations, is often binary, and that it is the decision to donate or not that is of crucial importance (Doliński 2018), we analyzed the effects of Want and Expectancy on the decision to help. To that end, we ran logistic regression on the decision to help (0 when a participant did not donate anything, 1 if they donated something).
The results showed that there was a significant effect of Want, b = 0.77, Z = 2.96, p = 0.003, 95% CI [0.27, 1.28], whereas Expectancy was non-significant, b = 0.20, Z = 0.86, p = 0.390, 95% CI [− 0.25, 0.65]. Importantly, however, there was a non-significant Want × Expectancy interaction, b = − 0.53, Z = − 1.86, p = 0.063, 95% CI [− 1.08, 0.03]. Conditional effects analysis showed that, whereas there were no significant effects of Expectancy at high (+ 1 SD) values of Want, b = − 0.33, Z = − 0.91, p = 0.364, 95% CI [− 1.05, 0.39], there was a significant effect of Expectancy at low values of the moderator, b = 0.72, Z = 1.99, p = 0.047, 95% CI [0.01, 1.44]. The interaction is graphically presented in Fig. 2. Similar results were obtained when controlling for gender and attitudes towards money (see Table 11 in Supplementary Materials).
To test if Want and Expectancy predicts the sum of donation (log transformed prior to the analyses due to the skewness in the data, skewness = 1.58) we ran similar analyses as we did in Study 1. The results showed that there was no interactional effect of Want and Expectancy on donation, β = − 0.02, t = − 0.32, p = 0.748, 95% CI [− 0.16, 0.12]. Want was a significant predictor of donation, β = 0.35, t = 3.93, p < 0.001, 95% CI [0.17, 0.53], whereas Expectancy was not, β = 0.10, t = 1.08, p = 0.280, 95% CI [− 0.08, 0.28].
Thus, we demonstrated that, although it was the desire to help that mainly predicted helping behavior, when it was low, high Expectancy predicted donations as well. The study adds to the previous one by showing that the interactional effect of Want and Expectancy not only predicts declarations, but also the actual behavior subsequently displayed. Unlike in the previous study, we didn’t find significant effects on the amount of money donated, which might stem from the fact that this time we asked participants for real donations not just declarations. It is in line with the results of other studies investigating actual helping behaviors (vs. declarations), showing that the decision whether to donate differs from the decision on how much to donate (Parsons 2007; Karlan and Wood 2017; Bergh and Reinstein 2020).
Study 3
In Study 3, to more directly test if increased Expectancy may predict donations if desire to help is low, we manipulated the level of Expectancy by providing participants with information about the chances of success of collecting the money for the given victim.
Participants
We calculated power for this study in a similar manner to that carried out for Study 1 but the interaction was calculated for the high and low Expectancy conditions rather than for high and low values of the moderator. The analysis yielded similar results and we aimed to recruit 204 participants for this study.
The sample was comprised of 205 participants. There were 167 women and 38 men ranging in age from 18 to 44 (M = 23.19, SD = 3.91). All participants were enrolled in a lottery in which they could win three prizes of 50 PLN each. All subjects gave informed consent before participation in the study and were debriefed upon completion of the study.
Fifteen participants failed our attention check (see below). Therefore, the final sample comprised N = 190 participants (156 women, 34 men; with the mean age M = 23.06, SD = 3.68).
Materials and procedure
Participants read a story about a child (or children) in need. Additionally, we manipulated the Expectancy of the fund-raiser’s success by adding information about the amount of money that had been collected so far. One group of participants was told that: The money collection is ending soon and, so far, only 50,000 PLN has been collected; this was the low Expectancy condition. Another group of participants was told: The money collection is ending soon and, so far, 450,000 PLN has already been collected; this was the high Expectancy condition. All participants were given the information that 500,000 PLN needed to be raised. We assumed that by informing participants about the progress of the fund-raiser, we would influence their expectations of success regarding the whole campaign. The manipulations used in each condition are presented in the Supplementary Materials.
We used the same measures of Want, Expectancy (Cronbach’s α = 0.75, M = 5.13, SD = 1.13), and donations as in Study 1. The participants’ attitudes toward money were also measured. An attention check question was also included (participants were asked to mark answer number 4).
Results and discussion
Descriptive statistics and intercorrelations between the variables are presented in Table 1. As Want and Expectancy are correlated, we explored VIF index to be sure that performing interaction is accurate. VIF equals 1.74, thus we performed analysis.
As predicted, the manipulation significantly affected Expectancy, F(1, 188) = 11.52, p = 0.001, partial η2 = 0.06 (M = 4.84 in the low and M = 5.38 in the high Expectancy condition). There was no effect on Want, F(1, 188) = 0.80, p = 0.372.
Next, we tested whether our manipulation significantly predicted the magnitude of the donation (log transformed prior to the analyses due to the skewness in the data, skewness = 1.16) via Expectancy at different levels of Want. We thus tested a moderated mediation model (Hayes 2013, model 14), with the manipulation as the IV, Expectancy as the mediator, Want as the moderator, and the donation as the DV (the model is pictured in Fig. 3). As in Study 1, the variables were log transformed prior to the analyses due to the skewness in the data. The analyses were performed with the use of Process macro for SPSS version 2.13 (Hayes 2013). In all analyses, bias corrected bootstrap confidence intervals are reported; 10,000 bootstrap samples were used in all analyses. All variables were standardized prior to analyses.
The results showed a significant moderated mediation effect, IMM = − 0.06, SE = 0.03, 95% CI [− 0.14, − 0.01], with the manipulation significantly predicting donations at low levels of Want, IE = 0.11, SE = 0.06, 95% CI [0.02, 0.27]. The effect was not significant at high values of Want, IE = − 0.004, SE = 0.06, 95% CI [− 0.12, 0.12], suggesting that when Want was high, it was of no consequence what our participants were told about the likelihood of collecting the necessary amount. The Want × Expectancy interaction is presented in Fig. 4.
Thus, this study replicated our previous findings and provided further support for the interactive influence of Want and Expectancy on the declaration of donating. As in previous studies, Want was the main predictor of declarations to help, and when it was high, the level of Expectancy was immaterial. However, when Want was low, the amount of declared donations significantly increased with increasing Expectancy. Moreover, we showed that experimentally manipulating Expectancy translated into an increased willingness to help among participants who were not otherwise inclined to do so (i.e., those feeling low levels of Want).
Summary of the results
To present the results obtained in a more illustrative manner, we statistically summarized the results of Studies 1, 2, and 3. The summary of the study results was performed in Comprehensive Meta-Analysis Software (CMA; Borenstein et al. 2013). We used Pearson’s r as a measure of the effect size and the sample size to calculate sampling variance. We fitted a fixed-effects model with the level of Want as a moderator. The heterogeneity among the effects was significant [I2 = 60.86%; Q(5) = 12.76, p = 0.003] and expected, given the differences in studies designs.
The results showed that the overall effect for the three studies followed the same pattern as the individual studies. To be specific, Expectancy was not associated with helping behaviors when Want levels were high (r = − 0.02, p = 0.703, 95% CI [− 0.10, 0.07]). However, it was significant and positive at lower levels of Want (r = 0.17, p < 0.001, 95% CI [0.09, 0.25]). These results are summarized in Fig. 5.