In the Frame or Out of the Picture? A Statistical Analysis of Public Involvement in the Arts (2008)
Authors Pete Lunn and Eilis Kelly, ESRI; Commissioned by NESF, 2008.
Authors Pete Lunn and Eilis Kelly, ESRI; Commissioned by NESF, 2008.
The Public and the Arts (2006), a population survey commissioned by the Arts Council, asked over 1,200 Irish adults about their involvement in and attitudes towards the arts. The data have so far been mainly analysed using ‘univariate’ statistical techniques (Hibernian Consulting, 2006). That is, indicators of involvement in the arts, such as attending arts events or participating in art clubs or classes, have been broken down by separate socio-demographic characteristics, such as gender, age, social class, educational attainment and income. The univariate picture is one of strong social gradients.
Multivariate analysis can help us to test the hypothesis that social gradients merely reflect different preferences or tastes. Once it is established that a set of characteristics such as age, gender, class and so on is related to the variable of interest – in statistical parlance, once we have a model – then it is possible to introduce additional variables, such as whether an individual is interested in the arts, to see if the relationships remain strong. In statistical analysis, this allows us to compare the ‘direct’ and ‘indirect’ effects. Consider, for instance, one of our results: the model for attending a play shows that people of lower income attend fewer plays. A direct effect of income might derive from the cost of admission: low income makes it harder to afford the ticket. However, an alternative is that low income people have less contact with people interested in theatre and so are less likely to develop an interest themselves. We can test these competing hypotheses, once we have built a basic multivariate model for attending a play, by adding variables that measure people’s interest (whether they say they are interested in the arts and whether they watch or listen to plays on the television or radio). Put simply, we can test whether a person of lower income who is interested is less likely to attend an event than a person of higher income who is similarly interested.
In all cases where it is significant, being female increases the likelihood of attendance. However, men are not significantly more likely to attend no arts event at all. This may seem like a contradiction, but can be explained by the fact that gender is not a significant factor for four of the five most popular arts events – women attend a greater variety of events. Higher educational attainment, higher social class, and higher income are associated with a greater chance of attendance at almost all events. The only exception is that higher social class is associated with a lower likelihood of attending a country and western event. In the model for attending no events at all, the direction of the relationships is reversed. Being from a non-white ethnic minority is always associated with a reduced chance of attendance.
The odds ratio for reading nothing deserves specific mention. Controlling for other variables including socio-economic circumstances and work status, the odds that a man read no kind of literature in the previous 12 months are more than double those that a woman did.
The odds that a woman is ‘interested’ are twice the odds that a man is, while those with higher educational attainment, social class and income are more likely to say they are interested in the arts. This model is consistent with
In comparison with the univariate breakdowns offered in previous reports (Hibernian Consulting, 2006; NESF, 2007), once a full multivariate analysis is conducted, which controls for other relevant factors such as gender, age, location, region and so on, the impact of socio-economic status, as measured by educational attainment, income and social class, is stronger than the univariate analysis reveals. Thus, the primary conclusion of the current exercise is that the association between socio-economic disadvantage and attendance at arts events is stronger than has been stated in previous reports.
Multivariate analysis can help us to test the hypothesis that social gradients merely reflect different preferences or tastes. Once it is established that a set of characteristics such as age, gender, class and so on is related to the variable of interest – in statistical parlance, once we have a model – then it is possible to introduce additional variables, such as whether an individual is interested in the arts, to see if the relationships remain strong. In statistical analysis, this allows us to compare the ‘direct’ and ‘indirect’ effects. Consider, for instance, one of our results: the model for attending a play shows that people of lower income attend fewer plays. A direct effect of income might derive from the cost of admission: low income makes it harder to afford the ticket. However, an alternative is that low income people have less contact with people interested in theatre and so are less likely to develop an interest themselves. We can test these competing hypotheses, once we have built a basic multivariate model for attending a play, by adding variables that measure people’s interest (whether they say they are interested in the arts and whether they watch or listen to plays on the television or radio). Put simply, we can test whether a person of lower income who is interested is less likely to attend an event than a person of higher income who is similarly interested.
In all cases where it is significant, being female increases the likelihood of attendance. However, men are not significantly more likely to attend no arts event at all. This may seem like a contradiction, but can be explained by the fact that gender is not a significant factor for four of the five most popular arts events – women attend a greater variety of events. Higher educational attainment, higher social class, and higher income are associated with a greater chance of attendance at almost all events. The only exception is that higher social class is associated with a lower likelihood of attending a country and western event. In the model for attending no events at all, the direction of the relationships is reversed. Being from a non-white ethnic minority is always associated with a reduced chance of attendance.
The odds ratio for reading nothing deserves specific mention. Controlling for other variables including socio-economic circumstances and work status, the odds that a man read no kind of literature in the previous 12 months are more than double those that a woman did.
The odds that a woman is ‘interested’ are twice the odds that a man is, while those with higher educational attainment, social class and income are more likely to say they are interested in the arts. This model is consistent with
In comparison with the univariate breakdowns offered in previous reports (Hibernian Consulting, 2006; NESF, 2007), once a full multivariate analysis is conducted, which controls for other relevant factors such as gender, age, location, region and so on, the impact of socio-economic status, as measured by educational attainment, income and social class, is stronger than the univariate analysis reveals. Thus, the primary conclusion of the current exercise is that the association between socio-economic disadvantage and attendance at arts events is stronger than has been stated in previous reports.