Sociology. Anthony Giddens
avoided.
Interpreting and reporting the findings
Once the material has been gathered together for analysis, the researcher’s troubles are not over. Working out the implications of the data and relating these back to the research problem are rarely easy. While it may be possible to reach a clear answer to the initial questions, many investigations are, in the end, less than fully conclusive. The research findings, usually published in a report, journal article or book, provide an account of the nature of the research and seek to justify whatever conclusions are drawn. This is a final stage only in terms of the individual project. Most reports also indicate questions that remain unanswered and suggest further research that might profitably be done in the future. All individual investigations are part of the continuing process of research which takes place within the international sociological community.
The preceding sequence of steps is a simplified version of what happens in actual research projects (see figure 2.1). In real-world research, these stages rarely succeed each other so neatly and there is almost always a certain amount of ‘muddling through’. The difference is a bit like that between following a recipe in a cookbook and the actual process of cooking a meal. People who are experienced cooks often do not work from recipes at all, yet their food may be better than that cooked by those who do. As Feyerabend saw, following a rigid set of stages can be unduly restrictive, and many outstanding pieces of sociological research have not followed this strict sequence. Still, most of the steps discussed above would be in there somewhere.
Understanding cause and effect
One of the main problems to be tackled in research methods is the analysis of cause and effect, especially in quantitative research which is based on statistical testing. A causal relationship between two events or situations is an association in which one event or situation produces another. If the handbrake is released in a car that is parked on a hill, the car will roll down the incline, gathering speed progressively as it does so. Taking the brake off was the immediate cause of this event, and the reasons for it can readily be understood by reference to the physical principles involved. Like natural science, sociology depends on the assumption that all events have causes. Social life is not a random array of occurrences. One of the main tasks of sociological research and theorizing is to identify causes and effects.
Causation cannot be directly inferred from correlation. Correlation means the existence of a regular relationship between two sets of occurrences or variables. A variable is any dimension along which individuals or groups vary. Age, gender, ethnicity, income and social class position are among the many variables that sociologists study. It might seem, when two variables are found to be closely linked, or correlated, that one must be the cause of the other. Yet this is very often not the case. Many correlations exist without any corresponding causal relationship between the variables involved. For example, over the period since the Second World War, a strong correlation can be found between the decline in pipe-smoking and the decrease in the number of people who regularly go to the cinema. Clearly one change does not cause the other, and we would find it difficult to discover even a remote causal connection between them. There are other instances in which it is not quite so obvious that an observed correlation does not imply a causal relationship. Such correlations are traps for the unwary and easily lead to questionable or false conclusions.
In his classic work of 1897, Suicide (discussed in chapter 1), Emile Durkheim found a correlation between rates of suicide and the seasons of the year. Levels of suicide increased progressively from January to around June or July and then declined over the remainder of the year. It might be supposed that temperature or climatic change is causally related to the propensity of individuals to commit suicide. We might surmise that, as temperatures increase, people become more impulsive and hot-headed, leading to higher suicide rates. However, the causal relationship here has nothing to do directly with temperature or climate at all. In spring and summer, most people engage in a more intensive social life than they do in the winter months. Those who are isolated or unhappy tend to experience an intensification of these feelings as the activity level of other people around them rises. Hence they are likely to experience acute suicidal tendencies more in the spring and summer than they do in autumn and winter, when the pace of social activity slackens. We always have to be on our guard, both in assessing whether correlation involves causation and in deciding in which direction causal relations run.
Causal mechanisms
Working out the causal connections involved in identified correlations is often difficult. For instance, there is a strong correlation in modern societies between level of educational achievement and occupational success. The better the grades an individual gets in school, the better paid the job they are likely to get when they leave. What explains this correlation? Research tends to show that it is not formal schooling in itself; levels of educational attainment are influenced much more by the type of home from which the person comes. Children from better-off homes, whose parents take a strong interest in their learning, where books are abundant and a place to study exists, are more likely to do well than those from lowerincome groups where these aspects may not be available. The causal mechanisms here are the facilities that parents are able to provide for their children to study.
Causal connections in sociology should not be understood in too mechanical a way. The attitudes people have and their subjective reasons for acting as they do are causal factors in relationships between variables in social life, and qualitative research is required if we are to gain the kind of in-depth understanding of how individuals interpret their world. Max Weber (1979 [1925]: 13) was clear that sociological work should be explicable at this individual and interactional level, which is where the meaningfulness of social life is produced.
A discussion of some recent ‘critical realist’ approaches which focus on establishing causal mechanisms in social life can be found in chapter 5, ‘The Environment’.
Controls
In quantitative research, assessing the cause or causes that explain a correlation usually involves distinguishing independent from dependent variables. An independent variable is one that produces an effect on another variable. The variable affected is called the dependent variable. In the example above, academic achievement is the independent variable and occupational income the dependent variable. The distinction refers to the direction of the causal relation. However, the same factor may be an independent variable in one study and a dependent variable in another. It depends on what causal processes are being analysed. If we were looking at the effects of differences in occupational income on people’s lifestyles, then occupational income would be the independent rather than the dependent variable.
To find out whether a correlation between variables is a causal connection, we can use controls, which means we hold some variables constant in order to look at the effects of others. By doing this, we are able to judge between explanations of observed correlations, separating causal from non-causal relationships. For example, medical researchers studying smoking behaviour may suggest that vaping via e-cigarettes leads to a reduction in conventional tobacco smoking in young adults aged eighteen to twenty-four – that there is a causal connection between e-cigarette use and smoking cessation. To find out, we could gather a sample of conventional smokers in this age group and assign them randomly into two groups: an experimental group that is given e-cigarettes and a control group that is not. After the study period, we would measure