![Sample our Health and Social Care journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5938/TOC-Subject-Sample-2021-Health-Social-Care.jpg"})
Abstract
Aim: This article explores the link between income inequality and adolescent cannabis use at the national level, in the context of other relevant social conditions, in developed countries. Methods and data: Fuzzy set qualitative comparative analysis is applied to two data sets that contain information on the national prevalence of past year cannabis use among 15 and 16 year olds, taken from the ESPAD and HBSC surveys, with supplementary data from the MtF and ASSAD surveys for the USA and Australia (n = 97 for the ESPAD and n = 72 for the HBSC data set). The data sets also include data on national rates of income inequality (Gini coefficient), wealth (GDP per head), welfare support (average benefit replacement rates), urbanization and labour market conditions (youth unemployment). Findings: The combination of high inequality and high urbanization forms part of configurations that are consistent with being usually sufficient to cause high-adolescent cannabis use, alongside high GDP per head in the ESPAD data set, and low welfare support in the HBSC data set. Conclusion: Social conditions, and particularly the combination of income inequality and urbanization, should be considered when studying the causation of high levels of adolescent cannabis use at the national level in developed countries.
Acknowledgements
I would like to thank colleagues in the European Society of Criminology and the International Society for the Study of Drug Policy – particularly Alison Ritter and Beau Kilmer – and two anonymous reviewers for their helpful comments on previous versions of this article.
Declaration of interest
The author declares no conflict of interests and did not receive specific funding for this research.
Notes
1When upper case is used for AND or OR, this indicates that they are being used as Boolean logical operators.
2The ASSAD survey does not report drug use by the school year in which the respondent is, as ESPAD, HBSC and MtF do, but separately by age. Therefore, the mean of 15 and 16 year olds was used for the ESPAD data set, while the value for 15 year olds was used in the HBSC data set (as this survey has a younger average age). In addition, there is a slight mismatch in some of the survey years, as shown in table one. Data on other conditions was also taken from these years for Australia.
3 The CWED also contains an index of welfare generosity which is based on more information than the replacement rates. However, this is available for a smaller number of countries. In those countries that have it, it is highly correlated (Pearson’s r > 0.8) with the calculated average welfare replacement rate.
4In both data sets, there were configurations that were consistent with being necessary for high adolescent cannabis use. But they were contradictory in that they were also consistent with being necessary for the negation of that outcome. So they cannot be considered as necessary causes of either outcome. Such paradoxical evidence on necessity can occur where a large proportion of fuzzy set scores for the potentially causal configuration are >0.5 or higher, as this makes it more likely that the scores for this configuration will be higher than the scores for both the outcome condition and its negation. So Thiem and Duşa (Citation2013) provide routines for the identification and exclusion of contradictory necessary configurations.
5 Technically, zero represents the presence of the negation of the calibrated set for that condition. Due to the calibrations thresholds chosen for this analysis, this can be taken to indicate a low level of each condition.
6This consistency value is calculated by dividing the sum of cases’ scores for that configuration that are consistent with sufficiency (i.e. the total of those parts of their fuzzy set score for the configuration that are equal to or less than their score for the outcome) by the total of their scores for that configuration (Thiem & Duşa, Citation2013; Ragin, Citation2008).
7In order to produce a more parsimonious solution, configurations that have no actual cases can be dealt with as if they may affect the outcome in either direction. These logical remainder configurations are treated “as instances of the outcome if doing so results in a logically simpler solution” (Ragin, Citation2008, p. 156). The low welfare support condition – w – cannot be excluded from this element of the parsimonious solution because the configuration of W*I*GDP*U*YU is present in the sample (in the row 6 of the HBSC truth table; the configuration of Spain in 2001) and is not considered as an instance of the outcome. This leaves open the possibility that w forms part of the causal configuration.
8It is also possible to produce an “intermediate” solution in QCA by using information on which direction it is expected that each condition will influence an outcome to prevent the assumption that it may affect the outcome in the opposite direction from being used in the process of minimisation. As the theoretical and empirical link between the conditions and the outcome in these analyses are not firmly established, it was considered that there were no such “easy counter-factuals” in the analyses, and so no intermediate solutions were generated for these truth tables.
9Coverage gives an indication of the empirical relevance of each consistent configuration. It represents the proportion of the total of all the cases’ scores for the fuzzy set of the outcome that is covered by the consistent part of cases’ scores for the configuration (Thiem & Duşa, Citation2013; Ragin Citation2008).
10It is described as consistent with “usually” being sufficient as its value for consistency is less than perfect (i.e. <1).
11 García-Castro and Arino (Citation2013) also recommend the calculation of the Euclidian distances between two vectors – one for the consistencies calculate between cases for each time period and one for consistency within each cases across time periods – and the vector of the pooled consistency value. For the analyses of high adolescent cannabis use, this distance was well below the suggested threshold for identifying significant effects of time. As would be expected from the nature of the sample, which was created in order to examine variation across countries, there was a non-random pattern of different consistencies within countries across years.
12 fsQCA analysis was performed to test whether adolescent cannabis use formed part of configurations that were consistent with being necessary or sufficient for high income inequality. This found no configurations of social conditions with high or low adolescent cannabis use that were consistent with being necessity or sufficient to cause high income inequality at the consistent thresholds suggested by Thiem & Duşa (Citation2013).