Introduction

It is widely reported that women tend to be more likely to give to charity but reporting on the gender differences in the amounts given is more varied. While the former is verified in the current study, we find the distributions of the amounts to have subtle patterns, with the gender patterns varying substantially depending on marital status.

This article examines differences between men’s and women’s giving patterns. Since there are notable differences between single people and married/cohabiting people most of the analysis is conducted separately for both groups. (For brevity, we include cohabitees among “married” in the rest of the article.) We explore the impact of age, education, and other background factors on gender patterns in giving. Furthermore, examining both the probability of giving and the amount given, we use regression models to investigate whether gender differences remain once we control for background characteristics such as income, education, and age.

The current study adds to the established literature in several ways. First, we establish firmly that the common claim that women are more likely to give is true in Great Britain and that it holds for both single and married people, regardless of whether we control for background factors. Furthermore, we corroborate some recent studies in which, among single people, female donors were found to give larger amounts than male donors. We establish that this is not due to compositional differences between the two groups. We also give additional detail by examining the differences at distinct points of the distribution of donation amounts, using quantile regression. This approach also allows us to investigate the more nuanced gender differences among married people. In addition, we identify various gender differences in the support given to particular causes.

The article is set out as follows. The next section summarizes the literature on gender and charitable giving and orients the current study within this literature. This is followed by a description of the data on which the analyses are based. Then we present the results of our study. We begin this section with descriptive results on gender differences in the percentage of people giving each month and in the average amount given. We place a particular focus on distributional differences in the amounts donated and on gender preferences for different charitable causes. We then analyse giving patterns separately for single and married people. In the third part of this section, the impact of age and education on giving is explored in more detail. The final part identifies the extent to which gender differences in the donor percentages and amounts donated can be explained by background factors such as age and income. In the penultimate section, we discuss further aspects important for understanding the relationship between gender and giving. We also outline possible future research. The article is then concluded.

Literature Review

The literature on the relationship between gender and charitable giving is diverse but does not yet present a coherent picture. The research is presented either in peer-reviewed journals or in reports designed for direct practical use in the voluntary and community sector. We draw on both of these sources in our review. The UK-based sector reports are of particular value to the current study because of the dominance of US studies in the academic literature.

The growing body of research literature in this area bears witness to the increasingly important role that women play as major donors. Indeed there is an entire issue of New Directions for Philanthropic Fundraising devoted to this topic (Taylor and Shaw-Hardy 2005). While this is important from a fundraising perspective, large-scale philanthropy is outside the scope of this article. Instead, we focus on general charitable giving. In this context, a monthly gift of £100 is considered unusually high and the median gift among those who give in a specified month is £10 (CAF/NCVO 2006).

Core Figures in the United Kingdom

The CAF/NCVO UK Giving reports (CAF/NCVO 2006) present brief analyses of gender patterns in giving for the corresponding financial years. In 2005/06, 61% of women and 53% of men gave each month, while the mean amount given was £25 per donor for women, and £29 per donor for men.

A similar pattern is presented in Breeze and Thornton (2005, p. 12), which reports that: “Girls were more likely than boys to have given to charity the last time they were asked (87% vs. 74%).” This study also identified that 62% of children had seen their mother give to charity, while only 42% had seen their father give, although the report also mentions that this is likely to be in part due to the fact that children typically spend more time with their mothers. The mother was found to influence the child’s giving behaviour while the father had no impact. However, details on the statistical significance of these analyses are not presented.

Households/Marital Status

The focus of Andreoni et al. (2003) is the role of the household’s main decision-maker in the context of charitable giving. The study took place in Canada although we might expect the results to hold in a broader context. After identifying that preferences for giving are different for men and women in single-person households, the article asserts that in married households, the giving tends to follow the husband’s preferences. The study also finds that when the wife is the decision-maker, she tends to distribute the donations over more charities, giving less to each.

Wiepking and Bekkers (2005) identified that traditional norms can play a significant role in financial decision making and “couples with more traditional values are more likely to have the husband decide rather than the wife.” The article also identifies that “when religious affiliation is controlled for, it does not matter who decides.”

Giving by Gender and Marital Status

Derived from a more general study based in the United States, Mesch, Rooney, Steinberg, and Denton (2006) present findings on the effects of race, gender, and marital status on both giving and volunteering. This article provides interesting results, which we build upon in the current study: single women give more than single men after controlling for other demographic variables and married people give substantially more than single people.

Other Areas of Research

While there is some consensus on the core figures—particular the gender patterns in the donor percentages—in other areas, there is some controversy. We do not examine these questions in detail in the current study but discuss them here to provide a broader context for the research.

Andreoni and Vesterlund (2001) suggest men respond more strongly to the price of giving (with men being more altruistic when giving is cheaper), whereas Meier (2005) suggests they do not. This question becomes particularly important in the context of tax incentives for giving (such as Gift Aid in the UK) and in studies of the influence of income on giving.

The findings presented in Bolton and Katok (1994) are based on a dictator game experiment and suggest that there are actually no gender differences in generosity.Footnote 1 Reporting on another dictator game experiment, Ben-Ner et al. (2003) find that women are less generous when giving to other women and that they may give less overall.

The Current Study

In all of this, there is the sense of a growing theory, but one which needs a substantial effort to bring together into a coherent whole. This study builds on but also significantly extends research by Mesch et al. (2006) which examined gender differences in the US context controlling for income, education, ethnicity, and marital status. The data for Great Britain available offer the additional benefit of including information on donations to different causes, which enables us to examine gender preferences for charitable causes. In addition, the large sample size allows us to run models separately by gender and marital status in order to examine gender differences conditional on a greater range of background characteristics. We also enhance results by focusing on the distribution of amounts given separately by gender and marital status. By using quantile regression for our analysis of donation amounts, we examine gender differences at different parts of the amounts distribution conditional on background characteristics. We also briefly discuss issues such as household giving and the price of giving.

Data: The CAF/NCVO Individual Giving Survey

Survey Methodology

The research in this article is based on the CAF/NCVO Individual Giving Survey (IGS). This survey is run three times each year as a module in the Omnibus survey carried out by the Office for National Statistics (ONS). The ONS Omnibus Survey is a face-to-face survey using Computer Assisted Personal Interviewing. Respondents are asked about their donations to charity over the previous 4 weeks. The pattern of the questions is to ask about giving to a cause, then to ask the amount given by different methods to that cause. The ONS Omnibus Survey also includes a broad range of demographic questions, which are provided along with responses to modules. We discuss some key methodological issues below.Footnote 2

Sample

The ONS Omnibus uses random probability sampling stratified by region, car ownership, socio-economic status, and age. The sample is drawn from the Royal Mail’s Postcode Address File (PAF). The Omnibus sample taken from the PAF covers Great Britain but not Northern Ireland.Footnote 3 Only one person per household is interviewed. For the following analysis we use weights correcting for the higher probability of people in small households being selected.Footnote 4

We have merged data from nine Omnibus modules that were conducted during the years 2004 to 2007. The total sample size is 12,679. The response rate in each survey round is between 62% and 66%.

Other Methodological Issues

There are a number of other issues that should be borne in mind when reading this study. There are outlined below.

Seasonality

Although the months of the surveys are spread evenly across the year, and in supposedly “typical” months, there may be problems with the seasonality of giving. In particular, major campaigns such as Comic Relief and the increased levels of giving around Christmas might reduce the accuracy of the survey. However, since our focus here is on gender differences, this is unlikely to be a major concern.

Household Giving

Another problem is that the survey is based on individual giving, whereas some reported donations may be derived from the household. This issue is discussed in more detail in section “Household Giving and the Price of Giving.”

Definition of Charitable Giving

The survey asks the donor about charitable giving but does not refer to the legal definition of “charity.” For example, while organizations such as Greenpeace and Amnesty International are not official charities, we would expect people to report fees and donations to these organizations in the survey.

Definition of Causes

The survey expects the respondent to allocate their donations to particular categories of causes. However, an organization such as Red Cross could be classified as either “health” or “overseas.” For this study, we have to assume that there is no gender bias in the way a survey respondent classifies such organizations.

Reference Period

The survey only asks about giving in the previous four weeks. There are two main issues with this. First, “four weeks” acts as a proxy for “one month” and is in line with government studies such as the Citizenship Survey. Throughout this article, we refer to the reference period as 1 month. Second, both “non-donors” and “high-level donors” are to some extent artificial categories based on the 4-week reference period. In fact, there may be considerable overlap between the two groups, with some people preferring to give a large donation once each year rather than smaller more regular donations. The high skew in the observed amounts given may well not occur in a study that measured annual donations.

Data Cleaning

The raw data was cleaned before analysis was carried out to remove obvious reporting/recording errors including money that had been fundraised through events being reported as individual gifts, and other anomalies. The cleaning procedure included the deletion of high value gifts that did not seem to be appropriate given the method of giving used. For example, an event gift of £2,000 is likely to reflect giving from fundraising and not an individual gift and should therefore not be in the data set.Footnote 5 However, since mean amount given is around £30 per donor the deletion of extremely high gifts has an impact on mean amounts reported. We therefore predominantly discuss median amounts given that those are not sensitive to high value gifts.

Findings

General Results

We begin with a descriptive analysis of the patterns of charitable giving for men and women. In Table 1, along with the data on the percentage of people giving each month, and the mean and median amount given, we present the values of the 75th and 90th percentiles on the amount given by donors rounded to the whole number amounts in pounds.

Table 1 Core data on giving behaviour, by gender
Full size table

As expected, we find a significantly higher percentage of donors for women than for men (1% level): approximately 60% of women give, compared with 50% of men. When we look at the means per donor in Table 1, male donors appear to give more than female donors. The distribution of amounts is extremely skewed with about 60% of the total amount donated being given by the top 10% of donors (Micklewright and Schnepf, forthcoming). As such, the median amounts and other percentile values are more instructive if we want to examine gender differences between the average donors.

The median amount given among donors (which is not sensitive to the high values at the top of the distribution) is just over a third of the value of the mean, at £10 for both men and women. The 75th percentile is also the same for both genders, at £25. The 90th percentile is higher for men than for women, and it is the high values of these donors that lead to the higher mean amount among male donors.

Figure 1 further clarifies the gender differences in the distributions of amounts. This diagram presents the percentile range for a selection of round number amounts given for men and women separately.Footnote 6

Fig. 1
figure 1

Percentile range for a selection of round number amounts given by donors, by gender

Full size image

Although men and women both have a median amount given of £10, the £10 block for men is slightly to the right of the corresponding block for women. This same pattern occurs for almost all of the other blocks in the diagram, only reversing at the level of £30, above the 75th percentile in both groups. This demonstrates that in fact, for the majority of people, women tend to give more. We will see in section “Gender Differences in Choice and Number of Causes” (in Figs. 2 and 3) that this pattern arises largely from differences in the giving behaviour of single people.

Fig. 2
figure 2

Percentile range for a selection of round number amounts given by single donors, by gender

Full size image
Fig. 3
figure 3

Percentile range for a selection of round number amounts given by married donors, by gender

Full size image

Gender Differences in Choice and Number of Causes

The results up to now refer only to giving as a whole. We can also ask whether there are distinct gender patterns in the levels of support for different causes. It is important to note that, although we may discuss a choice of cause, this is not necessarily how it appears to a donor. Rather, the cause a person supports arises from a combination of personal motivations, marketing activity, and chance events.

Table 2 presents descriptive data on the differences in the causes supported by men and women. We find that the percentage of female donors is significantly higher than that of male donors for almost all causes.Footnote 7 Gender differences in the percentage giving are highest for animal welfare: the percentage of donors for that cause is about 70% higher for women than for men. In absolute terms, 4% more women than men give to that cause. Also for education, the elderly, hospitals, and children the percentage of women’s donors compared to men’s is at least 40% higher. Gender differences however are very small for causes related to the environment, religious organizations, and mental health.

Table 2 Percent donors and average amount given to different charitable causes, by gender
Full size table

Those causes for which men appear to be more likely to give (Arts and Sports) attract relatively few donors overall so even if a statistically reliable conclusion could be drawn, it would have little impact on the overall patterns of giving. Nevertheless, within these specific areas, the possible greater involvement of men is noteworthy and further carefully targeted research on the demographics and motivations of these donors would be rewarding.

Looking at the mean amounts given per donor, we find that for almost all causes women appear to give lower or similar amounts to those of men per donor. However, these differences are not significant, except for religious organizations, for which male donors give significantly more than female donors (£46 per month compared with £28). In addition to the differences in giving proportions and amounts, we also find that men and women have a different profile when we look at the number of causes supported (see Table 3).

Table 3 Percent donors giving to specified numbers of causes, by gender
Full size table

Women appear on average to give to more causes: 55% of male donors support only one cause each month, compared with 46% of female donors. This contributes to the higher percentage of female donors giving to particular causes. In addition, this means that women tend to give smaller individual gifts since they distribute the total amount given among more causes. (Note that, although this seems likely, it is not certain: the IGS records the total amount given by each method to each cause but does not record the exact number of gifts.)

Marital Status

Marital status is one of the most important background factors in a gender analysis of charitable giving. We will see that marital status has a pronounced effect on individuals’ giving behaviour, and more importantly for this analysis, that the difference is quite distinct for men and for women. In this section we present the core differences between men and women independently for single and married people. A second important reason for analysing by marital status is that this also gives us an insight into the possible extent of household giving. In the discussion at the end of this article, we address the question of whether individual charitable giving makes sense in the context of a couple with shared finances.

Before presenting this analysis, it is important to be clear that a comparison of married and single people is complicated by the fact that other aspects of life stage (such as income, wealth, or house ownership) might also be expected to influence giving behaviour. As well as matters of income, there are natural ways in which a person’s personal circumstances might influence his or her choice of charitable cause. For example, we might expect the pattern of giving to schools or children’s charities to change once a person has children. Similarly, an older person might have a greater awareness of health issues and this might lead to an increase in giving to medical research or healthcare organizations.

Gender Differences in Giving Proportions, by Marital Status

Table 4 presents the percent of donors by charitable cause, marital status, and gender. The pattern of women being more likely to give can be seen for both, single and married people. In addition, the ratios of donor proportions are very similar between singles and married people for most causes. This indicates that gender differences in the level of support appear to be largely independent of marital status.

Table 4 Percent donors of different charitable causes, by marital status, and gender
Full size table

For animal welfare and religious organizations, the gender differences in support vary substantially between single and married people. In interpreting this, it should be kept in mind that there are substantial differences in the composition of these two groups and between genders within groups. Using data for all 3 years of the survey, we see that more than a quarter of single women are above 64 years old, while only 15% of single men are in this age group. In addition, more than 50% of these women are living alone, compared with 30% of the men. Thus, gender differences may be due to compositional differences between genders.

It is particularly striking that while married men and women show the same level of support for religious organizations, among single people, women are nearly twice as likely as men to give to them. If we control for different characteristics (age, income, living alone, region, education, and proxies for wealth), this gender difference decreases, suggesting that compositional differences between single men and women can explain some part of this gender difference.Footnote 8 However, the gender difference remains significant. One interesting finding is that while giving to religious organizations increases with age for women, the situation is more complex for men, as will be seen below.

For animal causes the gender difference is again smaller for married people (with a ratio of 1.6) than for single people (with a ratio of 1.9). In this case, we find that even if women’s and men’s background characteristics were equal, the ratio of 1.9 found for singles would hardly change.Footnote 9

Gender Differences in Giving Amounts, by Marital Status

While marital status has a fairly marginal effect on the ratios of giving proportions, the effect on the distributions of giving amounts is substantial. Table 5 presents differences in the mean and percentile amounts by marital status.

Table 5 Amounts (in £s) given in different percentiles of the distribution, by gender and marital status
Full size table

Not surprisingly, we find that amounts given by married people are higher than that of single people (Mesch et al. 2006). As well as differences in incomes and household structures between the two groups, this may also reflect the fact that married people report household gifts as individual gifts (see below). Although the mean amounts for men and women are closer for single people than for married people, it is with single people that we find the strongest difference in the overall distributions of giving amounts. Up to the 90th percentile of the respective distributions, single women give more than men. It is only if we compare the highest-level givers, above the 90th percentile, that men appear more generous. Among married people, the turning point of gender differences arises much earlier, at around the 75th percentile.

Figure 2 gives more detail on the respective distributions of the amounts given by single men and women. We see that while 50% of single female donors gave less than £10, nearly 60% of single male donors did so.

As a consequence, we find that single women are not only more likely to give to charitable causes but that they are predominantly also more generous in terms of their amounts given. This corroborates the findings in Mesch et al. (2006) and Brennan and Saxton (2007) and will be examined in greater detail below. The distributions for married men and women do not show the same marked differences that we saw for single people. This may be due to household giving or to a convergence of social norms of giving (see Fig. 3).

Gender Differences in Giving Conditional on Age and Education

Gender variations in giving are closely linked not only with marital status but also with factors such as age and education. We now investigate more closely the possible impact of these background factors. Figure 4 demonstrates how likelihood to give changes with age, by gender and marital status for different causes. The y-axis scales presenting the percent of donors vary between the different graphs in order that the overall pattern is clear for each. Nevertheless, care should be taken in interpreting the graphs. In particular, the much broader scale for the “All causes” graph means that quite substantial changes (such as the peak for married men in the 36–45 age group) can be less apparent than on the cause-specific graphs. We outline below some of the more substantial findings.

Fig. 4
figure 4

Percent donors for different charitable causes, by age group, marital status and gender. Note: The x-axis refers to different age groups; the y-axis denotes the percent of donors

Full size image

With the exception of married women giving to medical research and married men giving to children charities, the percentage of people giving is lower for 16–25 year-olds than for 26–35 year olds. In almost all cases, the donor proportion for women is higher for both single and married people. The pattern for children charities and religious organizations are quite different however: for children’s charities, donor proportions tend to fall with increasing age (after an initial increase), while for religious organizations, it generally increases with age. For all groups except married men and for all causes except religious organizations, the likelihood to give falls when we move from the 55–64 age band to the 65+ band. For married men, the donor proportion increases when we move from the 55–64 band to the 65+ band, overall and for all causes except children’s charities.

For overseas and religious organizations, there are noticeable peaks in the donor proportion for married men in the 36–45 age band. This may well be two measures of the same phenomenon since many overseas organizations have a religious affiliation. If this were only about rising incomes followed by increased responsibilities in the 46–54 group, we would expect to see the same pattern across other causes. A quantitative study might be useful to help assess this pattern. Lastly, we find that married men’s giving to overseas and animal charities falls dramatically in the 46–55, perhaps due to changing personal priorities.

Although the likelihood to give increases with education (Fig. 5), it is interesting to note that the gender variation persists across all levels of education. Note that this relationship may in part be due to the correlation of educational attainment with income.Footnote 10

Fig. 5
figure 5

Percent donors by education, gender and marital status

Full size image

We also see that at all levels of education; there is a much greater difference in donor percentages between men and women among single people. It is not clear whether the difference falls because of household giving or because of social influences causing spouses’ giving habits to converge. For both single and married people, the difference is particularly strong for those who have completed a degree. Although the difference is even greater for people with education recorded as “other,” this is not easily interpreted since it includes foreign qualifications of varying levels.

Gender Differences after Controlling for Background Characteristics

As discussed above, the male and female populations differ in their composition, according to demographic factors such as age and income. These compositional differences might explain gender differences found in the amounts given and the probability of giving. In the analyses above, we have seen that gender variations persist even when we take into account marital status together with age or education. To examine this more rigorously, we perform regression analyses on both the probability of giving and the amount given, drawing on a broad range of background factors.

Gender Differences in the Probability of Giving, Conditional on Background Characteristics

To model the probability of giving, we use logistic regression. The first model uses gender as the only independent variable. In successive models, we then add additional variables describing: age, household structure (whether a dependent child lives at the same address and whether the person is living alone), and region (England, Scotland, or Wales); then proxies for wealth (property ownership and the number of cars per adult in the household) and income (a dummy for whether individual income is below or above the top quintile of the income distribution); and finally education.Footnote 11 This allows us to assess the extent to which gender may be acting as a proxy for these other factors in its apparent relationship with giving.

Table 6 presents the logistic regression coefficients for gender (taking 0 for male, 1 for female) for models run separately for single and married people. The eight coefficients reflect eight different logistic regression models and are all significant at the 1% level.Footnote 12

Table 6 Gender parameter estimate of different logistic regression models
Full size table

Model 1 was a void model, with the dependent variable of being a donor and the independent variable of being female. As can be seen in Table 6, for Model 1, the coefficient for the gender variable for single people is 0.476. We can calculate that the unconditional probability of giving increases for women by 0.476/4 = 0.119, and hence that the proportion of women who give each month is approximately 12% higher than that of men, a fact corroborated in the descriptive analyses earlier in this article. (Table 4 shows that gender difference in giving is 11.8%).Footnote 13

Results for Models 2–4 show that the conditional gender coefficients are significant in all six models controlling for background factors. In fact, controlling for background factors leads the coefficients to increase slightly for both single and married people. Thus, we can also see that once we take other background factors into account (for example, comparing men and women with the same income, education, etc.), for single people, the difference in donor proportions rises from 12% to nearly 14% (as 0.545/4 = 0.136).

As a consequence, we can conclude that our unconditional results do not simply mirror some compositional differences for women and men. Women seem to be genuinely different from men in terms of the probability of giving. Although we have aimed to control for the main factors of interest, we have not, of course, included every possible factor. Other factors which could be examined in future analyses might include: spouse’s income; employment status (whether full-time, part-time, or no paid employment); or whether the respondent’s home is in an urban, suburban, or rural environment.

Gender Differences in the Amount Given, Conditional on Background Characteristics

As described earlier in this article, the amounts given are heavily skewed towards lower amounts, even when we exclude non-donors. As such it is not possible to carry out a straightforward least squares regression on the amount. We present results of quantile regression models, which allow us to explore separate parts of the giving distribution independently.Footnote 14 As with the logistic regression on likelihood to give, we run four models with successively more background factors, doing this separately for single and married people. These models allow us to give a value in pounds and pence to assess the extent of the gender difference at different parts of the distribution.Footnote 15

We analyzed the impact of gender at the 25th percentile (Q25), the median (Q50), the 75th percentile (Q75) and the 90th percentile (Q90) of the distribution of the amounts given by donors. Table 7 presents the coefficients for the gender variable (again coded 0 for men and 1 for women) of these different regression models.

Table 7 Gender parameter estimate of different quantile regression models
Full size table

The models presented in Table 7 display several interesting results. We begin by reviewing the Q50 models (which analyse the distribution around the median). As with all of the models, the results for single people are quite different from those for married people. In Model 1 (in which gender is the only independent variable), we find that single women give £3 more than single men, whereas we find no significant gender differences for married people. Note that this corresponds to the unconditional results presented in Table 5. Once we introduce control variables into the model, the gender difference decreases for single people. Conditional on household structure, age, region, wealth, income, and education single women give just £1 more than men (Model 4).

The change in the gender coefficient from £3 to £0.93 (in Model 2) for single women is significant. Compared to single men, single women are more likely to live alone and to be older, which are both correlated with higher giving. Still it is notable that controlled for wealth and income (both higher for men) the gender coefficient becomes significant again (Model 3). Including in addition a control for education decreases the coefficient slightly. However, gender does still matter (Model 4). For married people, we find that once we control for factors proxying wealth and income, the gender difference becomes significant (at the 5% level), with women again giving about £1 more than men.

We turn now to the other models and hence to the other parts of the distributions of amounts. For single people, the results are very similar to those found at Q50, except that the lower coefficients for the Q25 models and the higher coefficients for Q75 and Q90 models reflect the fact that we are looking at lower or higher parts of the distribution, where the amounts given are, of course, correspondingly higher or lower. For married people, a very different pattern in the higher percentiles (Q75 and Q90) appears. We find a gender difference in giving amounts in only one of these models, Model 2 for Q90. This coefficient is negative indicating that men give significantly more than women.

It is notable that throughout all 32 regressions run at different percentiles of the distributions and by marital status, 17 gender coefficients are significantly positive indicating that female donors give more compared with just one negative gender coefficient showing a higher generosity of male donors.

Household Giving and the Price of Giving

There are various other aspects of the link between gender and charitable giving that we have not discussed in detail. We briefly consider the way in which gender patterns in charitable giving are influenced by household finances and the price of giving, and give some suggestions on how these might be investigated in further research.

Household Giving

The Individual Giving Survey aims at measuring giving from the point of view of the individual. In particular, individuals are asked how much they gave to specific causes assuming that there is for all people a concept of individual giving. This is not necessarily true. The apparent convergence of giving behaviour among married people suggests that the household does have an interesting impact on giving behaviour. Couples sharing a home may decide together about their household spending, possibly including giving to charitable causes, especially for large gifts or for regular Direct Debit donations. The extent to which couples share resources has been investigated, although the details are still unclear. For example, while Lundberg et al. (1997) found that couples do not pool resources, Hotchkiss (2003), in reanalysing the same data, found the opposite. Both of these articles are based on a comparison of changes in child benefit legislation with clothing expenditure.

Since it may be that decision-making over household finances is shared, in the context of charitable giving, both the decision to give and the amount given may derive from the household and not individual. An important question is whether people report household gifts which they would not have made as an individual (meaning the person might not have been a donor in the first place). Similarly we can ask whether the amount reported includes household giving, potentially increasing the apparent level of giving.

Above, we observed a strong distinction between gender differences for single people and those for married people, with the former being much more pronounced. The apparent convergence in giving behaviour may well arise from giving becoming a household decision. On the other hand, it may be that within a marriage, giving behaviour of the couple converges as each partner is exposed to new social norms. It is worth noting that spouses are likely to have similar background characteristics, and once these are controlled for in the regression model, married women are found to give more than married men at the lower end of the distribution of amounts. Further research in this area would be helpful.

A detailed analysis of preferred causes may also give useful evidence of household giving. To consider the idea that women may be more likely to report household giving, we note that, according to Andreoni et al. (2003), in negotiating household giving, “married couples tend to resolve these conflicts in favour of the husband’s preferences.”Footnote 16 This implies that if women are indeed more likely to report household giving as their own, the preferences of causes for household giving would differ for married and single women, as giving is diverted towards the husband’s preferences. However, in Table 4, we observed that the ratios of giving for the various causes are very similar between married people and single people, indicating that it is not household giving that leads to a higher percentage of female donors. The lack of convergence in preferred causes among married people also suggests that household giving is not a straightforward replacement of individual giving. Here again, further research is needed to elaborate on this question.

Response to the Price of Giving

Various experiments suggest that men are more responsive to the price of giving, as outlined in the literature review. We might expect this to be reflected in data that compares those who use Gift Aid with those who do not. Indeed, in the data from the IGS, for both men and women, donors using Gift Aid tend to give more than those who do not. However, it is difficult to separate the influence of the price of giving from other effects. For example, a person using Gift Aid may be generally more engaged with charities and would perhaps feel inclined to give more regardless of tax incentives.

The relationship between income and giving behaviour is another area in which we might expect to observe the price of giving. Again, however, real-world complexities make it difficult to analyse this question from the survey data available. For example, it may be more reasonable to look at the level of disposable income as a measure of the price of giving, rather than simply the total individual income.

A specific study focused on the price of giving may be helpful in clarifying how the theoretical findings of Andreoni and Vesterlund (2001) and Meier (2005) influence behaviour in a real-world context. This might be of value to policymakers (for example in the area of tax efficient giving) and to fundraisers (in deciding on the level of suggested donations to appear in targeted appeals).

Conclusions

A number of points have arisen from the analyses presented in this article. Looking at the overall picture, women are significantly more likely than men to give to charities and this is not due to different background factors such as age and income. The same pattern holds when we look separately at single and married people.

Much of the research examining gender differences in amounts given focus on the mean amount. Our focus on distributional differences by gender shows that the mean amount is misleading. This is due not only to the high skew of the amounts data but also to the very distinct patterns of the distributions of donation amounts for men and women. For single people, results presented show that about 90% of female donors give more than male donors. For married people, although gender is not significant in the unconditional regression models of amounts given, when controlled for background factors, gender becomes significant at the lower end of the distribution. Thus, at and below the median, married women tend to give more than married men with similar background characteristics. Once background factors are controlled for, women’s higher probability to give increases, whereas the amount they give decreases (compared to the unconditional model). This indicates that the decision to donate and the decision about how much to donate are impacted differently on by background characteristics.

Finally, men and women appear to have different preferences for causes, with a notably higher support among women for animal welfare, education, and the elderly. For men, the support for religious organizations is also related to marital status, with married men nearly twice as likely as single men to support such organizations. Women also tend to support more causes.