2.2.3 Middle-range theory Robert Merton (1968) referred to the archetypal positivist approach as 'middle-range theorising' to distinguish the approach from grand theoretical speculation on the one hand and microscopic analysis of specific activities on the other.
In essence, the positivist approach tends to be portrayed as methodical, with a clear sequence of activities. The sequence is 'circular' in that it starts with theory and ends with theory. However, Merton argues that it is more akin to a spiral than a circle, as each time round the theory gets more refined or becomes stronger by not being refuted. This process, as we have seen, is also closely linked to falsificationism: setting up an hypothesis (a conjecture) and seeing whether it is refuted. Using multivariate analysis this amounts to testing to see whether an observed relationship can be shown to be spurious when other control variables are introduced.
In practice, as we have suggested, the positivistic approach does not operate in quite such clear-cut stages as the model suggest. Sometimes hypotheses are constructed or 'reconstructed' after the data has been collected. The operationalisation process that should precede data collection, is often redefined once the range of available data is known. It is rare that a sample is really representative and, in practice, available data has to be used. Thus a justification has to be made that the data does not seem to be unrepresentative, otherwise no generalisations can be made and the theory is not advanced. In short, the neat models and prescriptions elaborated in text books about how to do research can virtually never be reproduced in the 'real world'. Indeed, it is rarely desirable that they are, as they would take away the flexibility of the researcher to respond imaginatively to changing circumstances and new insights.
126.96.36.199.1 Testing for spuriousness: the incomes of art and design graduates A survey of nearly 2000 art and design graduates (Blackwell and Harvey, 1999) showed, amongst other things, that females earned significantly less than males even up to four years after graduation. Art and design is an area in which about two-thirds of graduates are female. However, taking the average income of all male and all female art and design graduates showed that females earned significantly less than males. That is, the difference was not likely to be accounted for by sampling error.
Could this difference be accounted for by other things? Several other factors, apart from gender, appeared to account for differences in income. First, the longer it is since graduation the more the graduate is likely to earn. The study covered graduates from 1993 through to 1996. Those who graduated in 1993, on average, had higher incomes than those who graduated in 1996. However, when the year of graduation was used as a control variable, female graduates still had lower overall incomes than male graduates.
Similarly, art and design covers a wide range of subject areas and there are significant differences in income within each area. Fine art graduates, for example, earn less than fashion and textile graduates who, themselves, earn less than product design graduates. Fashion and textiles tends to be a predominately female area and industrial design a predominately male area. So perhaps this could account for the difference in incomes. However, when the discipline was used as a control, the differences in male and female income did not disappear.
Other variables, such as age, social class, ethnicity, degree classification were also evaluated to see if they had any impact on income. Although age had some minor effect, the key variables were year of graduation and subject discipline. When these two variables were both used as controls simultaneously the relationship between income and gender remained (Table 188.8.131.52.1:1). Thus the researchers concluded that there is a clear gender bias in graduate income in favour of males in art and design.
Activity 2.2.4 Is using average income an appropriate way of showing that females have lower incomes than males? In which areas in the example above (Table 184.108.40.206.1:1) do female graduates have higher incomes on average than males? Does this information falsify the hypothesis that female art and design graduates have the same incomes than male graduates? Or does it 'specify' the relationship and, if so, how? Time 20–40 minutes.
220.127.116.11.2 Evaluating the evidence: race and sentencing Roger Hood (1992) undertook a study to explore whether or not the larger proportion of ethnic minority respondents in custody was a result of biased sentencing or proportionately more involvement by ethnic minorities in activity that would lead to a prison sentence.
Ideally, Hood argued the appropriate method to explore whether there was any racial bias in sentencing would be to follow a large cohort of cases from arrest through to final outcome (known as the 'disposition') in the Crown Court. However, there were insufficient resources for this so he decided to 'work backwards' from the Crown Court records. He compared the charges and dispositions of all ethnic minority defendants in a sample of West Midlands courts with a large sample of white respondents at the same courts.
To be a meaningful analysis, a range of other variables, that have a bearing on sentencing, had to be used as control variables. These independent control variables included: the seriousness of the crime; age; employment status; gender; whether the defendant had pleaded guilty; and 'above all, the court centre to which they had been committed for trial and the judge before whom they appeared for sentence' (Hood, 1992, p. 184). In all, 15 variables were considered relevant and included in the analysis.
Hood concluded that taking all the male cases across the West Midlands, that 'a black offender had a probability of receiving a custodial sentence about 5 to 8 per cent higher than a white offender. Asians, on the other hand, had about a 4 percent lower probability'. He goes on to comment that, given the number of cases 'these differences were sufficiently large to be to the disadvantage of a considerable number of black defendants'. However, he asked 'Does this amount to evidence of discrimination?' (Hood, 1992, p. 184).
Hood admitted that no study can control for all the possible variables that might have an impact on sentencing in courts. However, the results were consistent, the relationship between 'race' and 'sentencing' was not spurious. The results showed a small but significant correlation between race and sentence. Nonetheless, Hood reminds us that 'it is, of course, always hazardous to move from correlation to explanation'. A correlation just shows that two sets of data are related, as one changes, in this case ethnicity, the other changes, in this case the percentage of offenders getting a prison sentence. However, that in itself does not mean a causal link. Any causal relationship can only be determined by the researcher, who, on theoretical grounds, uses the association between the data to infer a causal relationship.
Hood discussed his data in some detail, specifying the relationship (that is, showing how the relationship varies in different circumstances). For example:
There was strong evidence to suggest that factors which would have been regarded as mitigating the seriousness of the case if the defendant was white were not given the same weight if the defendant was black in the cases dealt with at Dudley courts. Yet they were given a similar weight for black offenders dealt with at Birmingham. (Hood, 1992, p. 186)
He tested whether this might be a spurious relationship by asking whether black offenders sentenced at Dudley courts could have been less well served by the pleas of mitigation than those at Birmingham. Unfortunately, there is no way of measuring the performance of barristers and probation officers but for a number of reasons, which he listed, Hood argued that this seems an implausible explanation. He did not directly test whether this relationship is spurious or not but rather undermined the plausibility of any theory that the barristers performed worse in Dudley than Birmingham.
Taking the above, and much more evidence of a similar nature into account, Hood concluded that:
When one contrasts the overall treatment meted out to black Afro-Caribbean males one is left wondering whether it is not a result of different racial stereotypes operating on the perceptions of some judges. The greater involvement of black offenders in street crime and in the trade in cannabis, their higher rate of unemployment, their greater resistance to pressures to plead guilty, and possibly a perception of different, less deferential, demeanour in court may all appear more threatening. And, if not threatening, less worthy of mitigation of punishment. It was significant that being unemployed increased the risk of a black male getting a custodial sentence, but not, in general, for a white or Asian offender.
Hood, in this conclusion, is 'theorising' beyond what the data directly 'tells him'. Indeed, he admitted that 'this for the moment must remain speculation' but suggested a clear case has been made for more research to explore the issue of race and sentencing in more detail. What he has done, in his conclusion, is to go beyond the analysis that shows different treatment for black offenders, to propose that this is unfair treatment and that it appears to be the result of racial discrimination. He speculates that this discrimination is based on the perception of judges. He does not just make this last bit up. The evidence shows that the variation in sentencing is different in different courts. Hence the reference to 'some judges' in his conclusion. Furthermore, Hood draws on other research that has suggested judges take demeanour and attitude into account when sentencing.
18.104.22.168.3 Retaining a falsified theory: the case of a Neighbourhood Watch scheme A study was commissioned by the Home Office (Bennett, 1988) to evaluate neighbourhood watch schemes in London. The hypothesis of this study was that crime would be reduced as a result of neighbourhood watch schemes. For each area in which a neighbourhood watch scheme was to be implemented, an area that was similar in socio-demographic terms, and which had no neighbourhood watch scheme, was selected for the purposes of comparison. The areas in which no scheme was to be implemented thus served the purpose of 'control groups', as in a scientific experiment. The crime rate was the dependent variable and the neighbourhood watch scheme was an independent variable that may have had an affect on the crime rate.
There are other possible independent variables, such as, the level of policing, the proportion of young people, and the environment. However, by matching a neighbourhood watch area with a similar non-neighbourhood watch area it was hoped that the impact of these other variables would have been removed.
The study examined the rates of crime and residents' fear of crime in both areas one- to-two months before the implementation of the scheme in the experimental areas. In this way, an attempt was made to ensure that any changes that occurred in the experimental area could be shown to have taken place as a result of the implementation of the neighbourhood watch schemes as opposed to any other independent variable (although the matching process might have missed out some other important independent variables). Another survey was carried out twelve months after the implementation of the scheme.
The results of the study showed that crime had actually increased in the experimental areas (the areas with the neighbourhood watch schemes), although fear of crime had been reduced. In the 'control' areas (without the neighbourhood watch schemes), crime had either stayed the same or reduced. The hypothesis that crime rates would fall as a result of neighbourhood watch schemes was therefore falsified. However, this did not lead Bennett to conclude that neighbourhood watch schemes were ineffective or even counterproductive. As Williams and May (1996) pointed out, Bennett did not question the theory behind the implementation of neighbourhood watch schemes, nor did he think that the research design was poor. Instead Bennett argued that in the cases under study the schemes were poorly implemented '... the design of Neighbourhood Watch as expressed in the Metropolitan (London) police guidelines was not a good example of Neighbourhood Watch in general' (Williams and May, 1996, p. 147). In this way the theory is saved from the apparent falsification.
Activity 2.2.5 Bennett's (1998) research (cited in 22.214.171.124.3) attempts to follow scientific standards and that it appears to be value-neutral regarding the implementation of Neighbourhood Watch. How far do you agree that this is the case? What do you think are the weaknesses associated with this type of research? Are there, for example, problems of validity in this study? Time 20–40 minutes.
126.96.36.199.4 Using a non-typical sample to falsify an hypothesis: the affluent worker Researchers who are following the falsificationist (see 188.8.131.52) method do not always choose a representative sample. Sometimes they search for untypical cases. One of the more famous examples of a falsificationist study using a non-representative sample is Goldthorpe and Lockwood's (1968)Affluent Worker study in which they attempted to falsify the theory that the affluent working class were adopting the lifestyles of the middle classes. This was called the embourgeoisment thesis.
They chose a sample of the most affluent workers because they would have been the most likely to have adopted middle-class lifestyles. They chose car workers in Luton as an example of well-paid workers who potentially would adopt non-working class lifestyles.
Goldthorpe and Lockwood did not find that the affluent workers in their 'critical case study' had adopted the lifestyle of the middle classes so on this basis they were happy to reject the embourgeoisment thesis. Of course, other theorists are free to test the hypothesis in different circumstances despite the falsification.