8823

Comparison of deterrents to adult education participation in Britain and the United States

Gordon G. Darkenwald, Rutgers University

Abstract

An earlier investigation of deterrents to participation conducted in the U.S. was replicated in Britain. The replication addressed the cross-national comparability of the magnitude of individual deterrent variables, deterrent factor structures, and relationships between sample characteristics and factor scores. The British and American findings were generally comparable.

Introduction

Although a large body of factor-analytic research has accumulated over the years on what impels participation behaviour, only recently have studies of comparable sophistication examined what deters it. This gap in our knowledge base is particularly serious in that the construct of deterrent or ‘barrier’ occupies a central place in theories of adult education participation. Although motivational orientation studies have been cross-culturally replicated many times (thus establishing the extent to which U.S. findings can be generalised), the present inquiry was the first to examine deterrents to participation from a comparative perspective.

The investigator undertook the replication in the spring of 1987 during a yearlong stay in Britain as a Fulbright Senior Scholar. The USA study, titled ‘Factor Structure of Deterrents to Public Participation in Adult Education’, had, as noted, already been conducted; it was published in 1985[1]. Since the U.S. report is readily available, the present paper focuses on the process of replicating the research in Britain and, of course, on a comparison of the key findings.

The goals of the research, as stated in the Abstract, imply its overarching purpose, namely, to establish the extent to which the USA findings are generalisable to Britain and to suggest interpretations of significant discrepancies.

Methodology

The investigator made every effort to adhere to the canons of comparative quantitative inquiry, most notably to the criteria of cross-cultural equivalence in concepts and measures as explicated by Hui and Triandis[2]. However, replications, due to their ex post facto nature, preclude strict compliance with these criteria. Instruments need to be re-worded and otherwise modified, samples can be only roughly comparable, and various other controls are simply impossible. Even Hui and Triandis concede that the ideal is problematic: ‘Precision and meaningfulness of comparison are two basic desiderata that, very often, cannot be maximised at the same time in cross-cultural research’[3]. Making a virtue of necessity, the present research sacrificed precision in the interest of meaningfulness.

Instrumentation

The Deterrents to Participation Scale - Form G (for general public) was used to collect the American data. A full description of the development and psychometric properties of the DPS-G can be found in Darkenwald and Valentine. Briefly, the instrument is a valid and reliable (a = .86) 34-item Likert-type scale designed for self-administration. Before it could be used in Britain, however, modifications were deemed necessary by the investigator and a group of collaborating UK scholars. The most important changes were the deletion of five original items (on logical and/or psychometric grounds), the addition of four new items, and the re-wording of seven items. For both versions the respondents are provided with a broad definition of adult education (which includes training), followed by the statement (British version): ‘Although not true of everyone, many adults find it hard to participate in adult education activities even though they want to. Take a minute or two and try to think of something - anythingat all - that you wanted to learn during the past year, but never did. Then look at the reasons below and decide how important each one was in your decision not to enrol in a class or other educational activity . . .’. The items are then listed (by random assignment) with four response categories ranging from ‘Not at all Important’ to ‘Very Important’. All items were written with the root ‘because’. For example, ‘Because the course was scheduled at an inconvenient time’. The alpha reliability of the British version was .87.

Sampling and data collection

The studies required fairly large, heterogeneous samples of the adult public (that is, persons 18 or older who are not full time students). In both Britain and the United States, market research firms were engaged to draw random samples from defined populations. The population for the USA sample was an affluent New Jersey county. For Britain, the sample was randomly selected from the electoral polls of England. Wales and Scotland were excluded due to their cultural and linguistic distinctiveness. The DPS-G questionnaires were mailed with cover letters and stamped, pre-addressed return envelopes. Follow-up was not possible in the USA. In England, two follow-ups were conducted: a postcard reminder and a second questionnaire. For both studies, as anticipated, response rates were low. In Britain, a sample of 600 yielded 178 unable questionnaires. (nb. this sample was not that employed for my other conference paper). The postal service returned 127 mailings as ‘undeliverable’! The unadjusted response rate (including undeliverables, refusals, deaths, hospitalisations, etc.) was therefore 30%. The US response rate was even lower (215 of 2,000). Bias analysis indicated that in both countries, but less so in Britain, the non-respondents tended to have less formal education and lower incomes than the respondents. Nonetheless, the final samples were reasonably comparable to the general populations. For example, the sex-adjusted average gross annual income for England was £8,268 in April, 1985. Wage inflation was projected at 8% per annum, or £9,580 in April 1987[4] (Department of Employment, 1986). Of the survey respondents, 29.3% reported annual household incomes between £5,000 and £10,000, and 37.2% between £11,000 and £20,000. In addition, 44.4% of the respondents reported that they held no formal educational qualifications - very close to the national average.

Since the purpose of the present research was not to estimate population parameters, but to explore and compare relationships among certain variables, the low response rates are of little import. Both samples were sufficiently large and heterogeneous for statistical analysis.

Description of samples

Only a cursory description of the two samples can be given here. The fundamental demographics for England and New Jersey are these. Sex: NJ, 62% female; Eng., 51% female. Age: NJ, mean=43 years; Eng., mean=44 years. Employment status: NJ, 61% employed full time; Eng., 48.5% employed full time. Educational attainment: NJ, 4.7% no credential; Eng., 44.4% no credential; NJ, 32.2% secondary school credential; Eng., 37.4% CSE to GCE A-level secondary school credential; NJ, 9.8% two-year post-secondary credential; Eng., 10.5% two or three year post-secondary credential; NJ, 28.5% first college/university degree; Eng., 5.8% first university degree; NJ, 24.8% postgraduate study or degree; Eng., 2% postgraduate study or degree (nb. at secondary level and below, a simple conclusion would be grossly misleading since British school completers who do not sit for exams receive no credential). Gross family income: due to exchange fluctuations, comparisons are of little utility; however, 39.3% of the NJ sample reported incomes of $45,000 or more (£22,000 at current exchange rates), but only 7.9% of the English sample reported comparable gross income, namely £21,000 or more. The income and educational attainment statistics may strike some as shocking. Nevertheless they reflect the realities: Americans, not only in affluent areas, are much bettereducated than Britons (nationally, more that a fourth of U.S. adults have completed a first degree) and earn considerably more money (especially after taxes).

Data analysis

The British and USA data were subjected to the same data analysis procedures. Descriptive statistics, such as means and frequencies, were calculated for all variables. Relationships among non-factorial variables were explored using the Pearson and Spearman (rho) correlation formulas. For both the British and US versions of the DPS-G, principal components factor analysis was employed to extract the initial factors. The Kaiser criterion determined the number of factors retained for rotation. To generate uncorrelated factors with the most interpretable structures, orthogonal rotation (Varimax) was employed to reach a terminal solution. Only those DPS items that had loadings greater than .35 were used to define the final factors. The complete estimation method was utilised to compute factor scores, which were employed in correlational analyses to explore relationships between respondent characteristics and the deterrent factors. In all cases, the criterion for statistical significance was set at p< .05.

Findings

Space limitations preclude lengthy tables or a detailed narrative description of the findings. Consequently, only the more important result are reported, with emphasis on the factor analyses. The following discussion addresses in turn the three research questions stated in the Abstract.

Magnitude of individual deterrents

Means were computed for each deterrent item to provide a ranking of their perceived importance. As expected from prior studies employing versions of the DPS[5], the majority of item means were low, ranging between ‘Not at All Important’ to ‘Fairly Important’ (that is, between scale values of 1 and 2). This established pattern suggests that a decision not to participate in an organised adult education activity is due either to the combined or synergistic effects of multiple deterrents or the salience of just a very few for particular individuals.

Precise statistical comparisons of mean item ratings were not possible because the original format of the DPS-G was altered for administration in Britain (from a five to a four-point scale). However, a rough comparison is possible if we consider the lowest and the highest rated deterrents included in both versions of the DPS-G. As the data indicate, the differences were pretty much negligible.

For the US sample, the five highest-ranked items were: ‘Because the course was scheduled at an inconvenient time’; ‘Because I didn’t have time for the studying required’; ‘Because I didn’t think I could attend regularly’; ‘Because the course was offered at an inconvenient location’ (ranked nine in Britain); and ‘Because taking a course would subtract from time with my family’ (ranked six in Britain). British respondents ranked the first three deterrents among the five highest. The other two were ‘Because of the amount of time needed to finish the course'’ (ranked six in the USA) and ‘Because I couldn’t afford extra expenses like transport, books, etc.’ (ranked 20 in the USA). Clearly, only one discrepancy is meaningful - the high British ranking and low US ranking of the ‘extra expense’ item. The explanation is almost certainly a function of the large income gap between the British and American respondents.

For Britain and the USA, the deterrent items ranked lowest in importance exhibited few discrepancies. The low-rated items concerned health problems, lack of encouragement from friends and family, and competition from younger students.

Factor analytic findings

Despite significant modifications made to the version of the DPS-G used in Britain, the factor analytic findings were remarkably similar for the two countries. In brief, the British analysis yielded an additional factor (seven vs. six) and ‘broke down’ the US factor labelled ‘Lack of Confidence’ into two more refined factor structures. The seventh factor for the British sample was also in essence a ‘break down’ of a US factor dubbed ‘Personal Problems’, which included the family problem items. One US factor, ‘Low Personal Priority’ was not duplicated in Britain. However, the ‘Priority’ items did load on appropriate British factors.

In neither country did DPS-G items load on more than one factor and the majority of loadings were high (60 or more). Thus the final solutions, with factorial complexities of zero, met the most rigorous criteria for simple structure. All factors in the two analyses were conceptually meaningful and readily interpretable. Because the ‘content’ of the deterrent factors is inherently negative, negative labels were assigned.

Although the factor-analytic findings cannot be presented here, the factor labels provide an indication of the nature of the factor structures and the difference between the British and American findings. The following factors comprised essentially the same items and were given the same labels for the two sets of findings: ‘Lack of Course Relevance’, ‘Time Constraints’, ‘Cost’, and ‘Personal/Family Problems’. The US ‘Lack of Confidence’ factor split in the British analysis into two factors labelled ‘Low Confidence - General’ and ‘Low Confidence - Age’. The latter consisted of such items as ‘Because I felt I was too old to take the course’ and ‘Because I didn't have the qualifications for the course’.

Relationship of factor scores to demographics

Data on sex, age, educational attainment, income and employment status were secured for both samples. To compare the relationships of respondent characteristics to factor scores, only the factors common to both analyses can be utilised: Course relevance, Low Confidence (British ‘Low General’), Time Constraints, Cost, and Family Problems (US ‘Personal Problems’). The pattern of correlations was comparable across the two samples and largely what one would expect. For example, as was the case in both Britain and the US, one would expect Low Confidence to be significantly and negatively correlated with education, Cost to be negatively correlated with income and education, full time employment to be positively correlated with Time Constraints, and Family Problems to be negatively associated with sex (female) and educational attainment.

Conclusion

The factors generated from both the US and British data were comparable and represented clearly defined, conceptually meaningful components of the deterrent construct. Despite departures from methodological purity, the findings provide strong support for the generalisability of the USA to Britain. This conclusion has considerable theoretical significance, as well as implications for program planning and marketing. A priority for future research is replication in other Western nations; additional suggestions can be found in Darkenwald and Valentine[6].

[1] Darkenwald, G. and Valentine, T. (1985) Factor structure of deterrents to public participation in adult education. In Adult Educat