The source of data on the socio-economic background of new entrants to higher education was a special postal survey that sought information directly from students on their parents’ principal economic status, social class and socio-economic group as well as their highest level of educational attainment. The survey covered new entrants to higher education institutions admitted through the Central Applications Office (CAO), which processes the majority of Irish applications on behalf of the institutions. The survey covered all of the Universities, the Institutes of Technology, the Colleges of Education and eleven Other Colleges (see Figure 1.2 for a complete list of institutions of higher education whose entrants are included in the survey).
The population for the survey was confined to students from the Republic of Ireland who enrolled as first-time undergraduates in the first year of a higher education course in Autumn 2004. The survey was drawn from the CAO database, which records a wide range of data on student applicants as part of the admissions process. This includes information on second level education, Leaving Certificate examination results, school type, region, age and gender, as well as the outcomes of the higher education application process. The CAO files on the population of higher education entrants in 2004 contained a total of 34,682 records. Three mail-shots were carried out between November 2004 and January 2005. A total of 14,571 questionnaires were completed and returned to the CAO in pre-paid envelopes. This represents a response rate of 42%. The CAO subsequently delivered the completed questionnaires to the ESRI for coding, data entry and analysis.
Methodological Challenges
It should be acknowledged that all surveys, including that described here, face a number of challenges and, consequently, are subject to some potential inaccuracies. First, all surveys, be they samples or censuses, are subject to non-response, which can be a source of error. We discuss the problem of non- response in the new entrant survey in detail below. Second, surveys depend upon accuracy in item response. In the new entrant survey students were asked about parent’s principal economic status as well as their occupations. In respect of Principal Economic Status students were asked to distinguish between parents who were employed, unemployed, on home duties, retired, deceased and ‘other’. In order to measure parents’ social class and socio-economic group respondents were asked to provide precise details of current or most recent occupation and the questionnaire also sought to distinguish between employees, and self-employed with or without paid employees. Those whose parents or guardians were farmers were asked about farm size. These data were then coded; a process that entails interpretation in some cases, and then computerized. At each stage of any survey process, therefore, there is some potential for inaccuracy which researchers attempt to counteract by drawing a random sample and by checking, where possible, the accuracy of data collected. We discuss the various strategies adopted in the new entrant survey below.
A central element of the analysis of the data on socio-economic background is the comparison of the distribution of new entrants by socio economic group with the distribution of a similar age group in the population, drawn from the Census of Population. However, as we discuss below, a substantial proportion of the cases in the Census are allocated to the “Gainfully Occupied but Unknown” category, substantially more than in the survey of new entrants, rendering an accurate comparison between the survey results and the underlying population difficult.
Previous studies of new entrants to higher education have incorporated a review of long-term trends in social background (see, for example, Clancy 2001, which examines the 1980-1998 period). However, such a long-term analysis was not possible in the present study. This is because significant changes in the measurement and classification systems for both social class and socio-economic group were introduced in the Census data from 1996. Following this classification change the analysis of trends over time before and after 1996 has become progressively more difficult. Clancy (2001) accomplished his analysis of long-term trends by converting his socio-economic group data relating to 1998 entrants, and the corresponding 1996 Census data, from the new to the old classification system. The conversion was based on a special cross-tabulation provided by the CSO showing the relationship between the old and the new classification systems for the 1991 Census data. Given the extent of change in the structure of Irish society between 1991 and 2002 it would not be appropriate to use 1991 data as a basis for adjusting the 2002 Census and 2004 new entrant data to the old classification system, so it is not possible to undertake an analysis of long-term changes in social background of new entrants. Accordingly the present study focuses on changes between 1998 and 2004.
The 2004 survey of New Entrants achieved a response rate of 42%. This response rate is substantially lower than the 67% response rate achieved in the previous survey of new entrants to higher education conducted by Patrick Clancy in 1998. In the 1998 survey, the CAO sent questionnaires to all applicants from the Republic of Ireland who accepted an offer in September 1998. Students were asked to return the completed form directly to the CAO office. So the CAO, having allocated successful applicants to their higher education courses, and thus having established substantial contact and rapport with the applicants, then asked them to complete a questionnaire about their social background. The 2004 survey adopted a similar procedure, with the CAO sending out questionnaires, cover letters and pre-paid envelopes to be addressed to the CAO. The 2004 survey added a number of questions relating to choice of college and course and to parents’ educational attainment. There were good reasons for the additional questions,6 but this had the effect of increasing the length of the questionnaire, from one to two pages,
and this may have depressed the response rate.
Notwithstanding these specific considerations, response rates to postal surveys are frequently quite low and the response rate achieved is about average for postal surveys. Moreover, in recent years encouraging
respondents to participate in postal surveys has become challenging. Rigorous re-weighting of the achieved sample ensures that the data are representative of the population of interest. In the current study we were in a position to analyse the entire population – those who accepted offers through the Central Applications Office – so we could measure the key population parameters (e.g. gender, age, HEI type, school type and region) with precision and use these parameters to re-weight the data.
In the present study, as in the case of the report by Clancy (2001), extensive tests were undertaken to ensure that the resulting sample is representative of what is known about the population of interest. With regard to the social background of new entrants Appendix Tables A3.1 to A3.4 compare the distribution of the achieved sample with the population of applicants who accepted offers of higher education places in autumn 2004. This is drawn from CAO records across a range of dimensions: second- level school type, higher education sector, county and Dublin postal code and receipt of registration grant. While there are some small differences between the achieved sample and the population of CAO new entrants, these differences are not sufficiently large to lead to significant bias in reporting in relation to the social background of new entrants.
Weighting
In line with best practice in surveys of this nature the data was statistically adjusted or re-weighted prior to analysis. This ensures that the completed sample for analysis is wholly representative of the structure of the population from which it has been selected and, accordingly, addresses any potential issues of bias arising from differential non-response within subgroups of the population.7
The weighting procedure was implemented using a minimum distance algorithm which adjusts an initial weight so that the distribution of characteristics in the completed sample matches a set of externally derived control totals. The population parameters (control totals) were derived from data provided by the HEA on the characteristics of new entrants.
The completed sample of 14,559 respondents was weighted and grossed to the population of 34,678. The control totals used in the adjustment procedure were as follows:
• gender by age group (12 categories) • gender by region (16 categories) • gender by school type (14 categories)
• gender by higher education sector (10 categories) • gender by course level (6 categories)
7. Although all 34,678 relevant students were contacted a successfully completed questionnaire was secured from just under 15,000. Hence although an attempt was made to implement a census of all entrants an effective sample was achieved. Accordingly, the weights do not have to address any issue of design bias - only potential bias arising from differential non-response within subgroups of the population. As shown in Table A3.5 it is clear that there was no evidence of any such bias. Nonetheless, in line with best practice, statistically adjusted
Appendix Table A3.5 outlines the distribution of controls in both the population and unweighted sample. Comparison of columns D and B (shown in Column E) shows that there is, in fact, very limited difference in the structure of the completed sample compared to the population. Columns F, G and H illustrate that the limited structural differences which exist are fully addressed by the reweighting procedure.
Given the methodological challenges outlined above, it should be acknowledged that the results presented in this report are estimates of the true population parameters, as in the case of any survey based data. The results represent our best estimates. In reporting on distributions of new entrants by socio-economic group and by class, we present confidence intervals indicating the degree of precision of the estimates. We can also asses the reliability of our results by comparing them with other sources. In this respect, we find that our results from the 2004 survey are consistent with the results of the small sample survey conducted in relation to 2003 entrants. We will also see later in this chapter that the trends in access to higher education suggested by our analysis of the School Leaver’s Surveys, a different data set covering a different population, are broadly consistent with the changes between 1998 and 2002 over time revealed by the surveys of new entrants.