A comparison of A‐level performance in economics and business studies: How much more difficult is economics?

Abstract

This paper uses A‐Level Information System data to compare academic performance in two subjects often viewed as relatively close substitutes for one another at A‐level. The important role of GCSE achievement is confirmed for both subjects. There is evidence of strong gender effects and variation in outcomes across Examination Boards. A counterfactual exercise suggests that if the sample of Business Studies candidates had studied Economics nearly 40% of those who obtained a grade C or better in the former subject would not have done so in the latter. The opposite exercise suggests that 12% more Economics candidates would have achieved a grade C or better if they had taken Business Studies. In order to render a Business Studies A‐level grade comparable with an Economics one in terms of relative difficulty, we estimate that a downward adjustment of 1.5 UCAS points should be applied to the former subject. This adjustment is lower than that suggested by correction factors based on conventional subject pair analysis for these two subjects

Keywords:

• Business Studies
• Economics
• subject difficulty
• ordered probit

Acknowledgements

The authors would like to thank Carol Taylor Fitz‐Gibbon and Paul Skinner of the Curriculum, Evaluation and Management Centre, University of Durham for permission to use ALIS data in this study, and for providing comments on an earlier draft of this paper. The constructive and helpful comments of two anonymous referees are gratefully acknowledged. In addition, Samer Al‐Samarrai, Mike Barrow, Tom Bourner, Peter Davis, Peter Kutnick, Mike Sumner and Alan Winters are also thanked for providing useful comments on earlier drafts. The authors are also grateful to Lynne Cahill for providing important technical assistance. In all cases, the usual disclaimer applies.

Notes

Adnett et al. (2002) examine in more detail the nature of the incentives created by the quasi‐market reforms in education and analyse some of their implications.

The substitution is also partly facilitated by the ease with which existing staff in most schools and colleges can move from Economics to Business Studies teaching.

For instance, enrolments on Communications Studies courses have risen by nearly 90% over the corresponding period (see DfEE, various issues).

The recent work of Machin and Oswald (2000) highlighted the decline in the number of UK graduate economists applying for PhD programmes at British universities. The trends in A‐level enrolments may have some indirect impact here.

In contrast to subject pair analysis we control, within an econometric framework, for a range of characteristics that could potentially affect academic achievement.

One potential econometric problem with the modelling approach adopted is that no attempt is made to correct for selectivity bias. Since the magnitude of selection bias effects cannot be known a priori, some caution needs to be exercised in the interpretation of our results.

All the estimation reported in this paper was undertaken using the LIMDEP 7.0 (1998) and the STATA 6.0 (1999) software packages.

The testing principle used is based on the outer‐product gradient form of the efficient score (or Lagrange Multiplier) tests (see Chesher and Irish, 1987). Orme (1990) has questioned the use of outer‐product gradient‐based tests in the context of a simple binary probit and demonstrated their poor finite sample properties in this setting. Orme (1990) demonstrated that the null hypothesis was rejected more often than suggested by the nominal size of the test. The sample sizes used in our application are relatively large. In addition, the implication of Orme's (1990) findings, if they extend to the ordered probit model, is that we are actually setting our estimated models a more stringent set of tests to pass.

The omitted variables’ tests will focus on the role of gender interactions given the potential influence that gender differentials in the effects of characteristics may exert on academic performance.

The pseudo‐functional form tests are based on the RESET testing principle conventionally applied in the linear regression model (see Ramsey, 1969). The test uses the predicted ordered probit standardised index raised to the second, third and fourth powers as auxiliary measures to capture model mis‐specification. Ramsey (1969) demonstrated the optimality of this polynomial order. Peters (2000) provides some empirical evidence on the power of this type of test for a number of different limited dependent variable models.

This correction to the variance–covariance matrix is comparable with adjustments undertaken by educational researchers in multi‐level analysis. In our application the data could be interpreted as comprising just two levels (i.e., the candidates and their institutions). There is insufficient information in the data to take the clustering any further (e.g., by teacher, tutor or class). Goldstein (1987) provides further details on multi‐level analysis for the interested reader.

The methodology used here is an extension to the ordered probit of the Oaxaca (1973) decomposition for a linear regression model, which has had some popularity in empirical labour economics.

On the basis of Likelihood Ratio Tests, the joint effects of 14 controls for parental labour force status when the student is 16 are statistically insignificant in both Economics and Business Studies models. The chi‐squared values are 6.3 and 14.1, respectively, with a critical value of 23.7. The joint effects of four parental educational‐level variables are statistically significant in the Economics model but not the Business Studies model. The chi‐squared values are 11.1 and 4.5, respectively, with a critical value of 9.5.

However, the significant RESET value may be attributable to the neglect of selection effects.

The estimated standard error for the point estimate of the gender effect in the Business Studies equation is 0.048, thus rendering the effect well determined with an asymptotic t‐ratio of 3.1. However, on the basis of a simple t‐test, there is no statistical difference in gender effects across the two subjects. The absolute asymptotic t‐test computed for this hypothesis is 1.3.

A unit increase in the average GCSE score represents a considerable academic achievement. If the average student takes eight GCSEs, a unit increase in the average score from say 5 to 6 could represent a movement from grade Cs to grade Bs in all eight GCSE subjects. Thus, if we wanted to roughly obtain the effect of a one letter grade increase in just one of the eight GCSE subjects, we would need to divide the estimated ordered probit coefficient by eight.

Thus, although being male raises performance in Business Studies, ceteris paribus, there is an interesting counter‐effect present in that, given the GCSE achievement level, the value‐added effect for females doing Business Studies is higher than for males, ceteris paribus.

One potential weakness of the counterfactual exercise is that the predicted performance for the sample of Business Studies (Economics) candidates in Economics (Business Studies) is based on the assumption that they are constrained to follow the same Examination Board as in their chosen subject. It is not implausible that educational institutions use a different board depending on the subject. Unfortunately, our analysis cannot deal with this particular problem.

We use the UCAS points score here to facilitate the computation of a correction factor and to allow a numerical interpretation of the decomposition analysis.

It should be stressed that our findings and conclusions relate to just one cohort drawn from 1998. It is conceded that a different picture in regard to say subject difficulty, Examination Board variability or gender might emerge in later years.

Ashworth and Evans (2001) note that the absence of a critical mass of female students studying Economics at secondary school level also acts as a deterrent for females in choosing this subject.

The weaker effect for Economics may be attributable to a stronger correlation between achievement in A‐level Mathematics and GCSE Mathematics for the Economics candidates. This might act to attenuate the estimated effect for the Mathematics A‐level subject in the Economics specification.

It is worth noting that independent earnings effects were detected for women for these particular degree subjects.