My Photo
Location: Singapore, Singapore

Not confident but not too shy; Not White but definitely not quite Oriental; Not religious but also not an athetist; Not sure whether 'athetist' is spelt correctly, but not that bothered about it; Not a naysayer, but not averse to saying no.

Tuesday, September 06, 2005

Reviewing Toh & Wong (1999)

Economic Value of Education
Spring Term

Course Assignment 1

‘Write a review of a paper reporting a study of the rate of return to education.’

Paper reviewed: Toh, M. & Wong, S. (1999). ‘Rates of return to education in Singapore’, Education Economics Vol 7, No. 3, 235-252.

Weijie Ng
MA Economics of Education

Section A: Introduction

The allocation of scarce resources between competing ends is the basic economic problem that any government faces. ‘Resources’ refer to inputs and notions of factors of production, such as Mill’s (1994) ‘three requisites of production … labour, capital and land’ while among the ‘competing ends’ are goods and services that one consumes for immediate satisfaction of needs and wants or that one invests in to create the capacity for future production of goods and services. Governments need to weigh competing ends against each other in order to decide where to spend scarce resources: in education, health, defence, etc. This is where studies to measure the rate of return to education come in useful, in that they may serve to inform policy-makers on the efficacy of investing in the education system and whether resources are being directed into educational stages and processes that are yielding a good return on the investments being made. The higher the relative social rate of returns to education, the more justifiable and efficient government expenditure on education would be. Toh & Wong’s (1999) article is one such paper that attempts to measure the rate of return to education, using a cost-benefit method. More specifically, it aims to compute both social and private rates of return to education in Singapore from 1980 – 1994, the results of which are claimed to justify shifting the burden of financing tertiary education to students and graduates who enjoy substantial private benefits.

In Section B, the study by Toh & Wong (1999) (henceforth referred to as T&W99) will be briefly described, with respect to their data, method, results and conclusions. Then, in Section C, their study will be evaluated, and its distinctiveness as compared to others in the field will also be explained. Section D will conclude this essay by outlining the probable impact of the weaknesses in the study on the rates of return calculated.

Section B: Brief description of T&W99

Data Sources:

As outlined in T&W99, there were four main reports from which data on incomes were drawn to estimate the financial benefits of education. These reports were:

a) the Census of Population of 1980, Singapore ( Khoo, 1981)
b) the annually published Report on the Labour Force of Singapore (Singapore MOL, 1981-1994)
c) the Report on the Survey of Employment of Graduates, 1980 (Singapore IRD, 1981)
d) the Report on Wages in Singapore, 1994 (Singapore MOL, 1994)

Adjustments to the data selected for use were needed, so as to include Central Provident Fund[1] pension contributions from employers and because different statistical measures of averages like the median and the mean were applied in the different reports.

Data sources for the private and social costs of education, as well as ability measures, which were presumably used for computations throughout the paper, were not stated in T&W99. Readers were instead referred to an unpublished paper by Wong (1996) for these.


A cost-benefit approach was used instead of a Mincerian one. Please see Section C for details.


Results from both cross-sectional and time series analyses were presented in a few subsections in T&W99. Below is a selection of T&W99’s findings:

From cross-sectional data (1980/1981 versus 1994):

Finding 1: Private rates of return were higher than social rates of return.

Finding 2: Rates of return to education when adjusted for ability were lower than that without adjustment.

Finding 3: Private and social rates of return for all educational levels and types in 1994 were lower than that in 1980.

Finding 4: Private and social rates of return to ‘primary’ education were the lowest, followed by ‘secondary, post-secondary and polytechnic’ education, and then ‘university’ education.

Finding 5: When ‘secondary, post-secondary and polytechnic’ education in 1995 was disaggregated, the vocation-based polytechnic education had the highest social and private rates of return.

Finding 6: Returns to professional university courses were higher than that to non-professional courses.

From time-series data (1981 – 1994):

Finding 7: Both social and private rates of return to university education declined over the period, while the rates of return to primary and secondary education has been relatively stable.

Finding 8: Income differentials between university graduates and secondary graduates declined over the period, while that between secondary graduates and primary graduates has remained relatively stable.

Comparing across countries:

Finding 9: Social rate of returns in 1994 resembled that of South Korea in the 1980s while that in 1980 resembled Hong Kong and Taiwan in the 1970s.


Because the private rates of return have been higher than social rates, and also higher than the market rate of interest, there is a strong efficiency case to shift the burden of financing higher education to students and graduates.

Section C: Evaluation of T&W99

Section C will attempt to evaluate T&W99 by addressing these issues in the following order: a) the choice of the cost benefit analysis method versus the Mincerian approach, b) the assumptions embedded that education has a causal link with earning differentials c) the measurements of the inputs (costs) and outputs (benefits) of education and d) the distinctiveness in the results of the study in its similarity with Carnoy’s (1972) work rather than that of Psacharopoulos’ (1991; 1994).

As mentioned in Section B and elsewhere, the method used in T&W99 was that of cost-benefit analysis rather than the Mincerian approach. The rationale given was the limited public availability of data. Indeed, the Mincerian approach, which involves the formulation of a wage equation and uses regression analysis to fit this function to individual data on earnings, years of schooling, educational attainment, gender, race, etc (Carnoy, 1995), requires a lot of data. If such disaggregated and rich data does not exist, then researchers have to either start from scratch and do their own surveys to obtain the required information, or do the cost benefit approach. Unfortunately, the cost benefit approach is also problematic, since it uses aggregated data, which means that individuals and their characteristics are grouped together in amorphous lumps, losing in the process individualised ‘special knowledge’ (Hayek, 1945), and renders the differing impact of education on separate individuals unclear.

In the footnotes of T&W99, it was stated that while education was assumed to be the ‘major determinant of earning differentials’[2], non-educational and inter-correlated factors were also integral. Denison (1964) had studied gross earnings differentials between college and high school graduates, and accrued 66 percent of such differentials to education alone, while 3, 6, and 7 percent of the remaining 33 were accrued to IQ, rank in high school class and father’s occupation respectively. This 66 percent has then been taken to be the alpha factor, ‘that coefficient which expresses the proportion of the observed differentials which can be directly attributable to extra education’ (Vaizey, 1973), and used to adjust observed differentials in seminal works such as Morris and Ziderman (1971), as well as in T&W99[3]. There are a few problems with T&W99’s use of the alpha factor. Firstly, it is by no means clear how the alpha factor of 0.66 has been applied in T&W99, as no calculations were shown or even explained in the main text. T&W99’s Table 1 presented columns for social and private rates that were ‘unadjusted for ability’ and ‘adjusted for ability’. One might presume that one could multiply the unadjusted rate by 66 percent to get the adjusted rate, but this appears not to be the case. Also, if the alpha factor has indeed been used to derive the ‘adjusted for ability’ results, this ‘adjusted for ability’ tag is a confusing misnomer since the alpha factor is really an adjustment for background factors rather than for ability solely. There is moreover no good reason for the alpha factor of 66 percent to be directly transplantable to Singapore, a country in which the stock of education, contexts of social divisions and various other social, economic and political forces are so different. It is also extremely unlikely for it to be the same between each stage of education, e.g. between a secondary education and a university education, etc, or to be the same from year to year. A proper measurement and comparison of rates of returns to education, whether over time or cross-sectional would need alpha factors to be measured for each instance. If resources are not available for such an extensive measurement, then the alpha factor would probably have to be induced, specifically considering Singapore’s arguably unique context. A final criticism is that it is tautological to claim that ‘(a)s expected, the rates of return to education adjusted to ability are lower than that without adjustment’[4], if the adjustments that T&W99 were actually doing was to multiply the unadjusted figures by a factor less than 1.

The assumption that education is the major determinant of earnings differentials is itself questionable. According to the screening hypothesis as expounded by Spence (1973), while education may identify productive capacities, it does not necessarily create or enhance them. Formal education may just confer credentials that employers can use to select better workers and to determine relative wages. If the screening hypothesis is true, then earning differentials are not a result of differing educational attainments per se, but underlying innate differences in ability that are brought to light by having or not having the appropriate credentials. Given that screening does manifest itself in the jobs market in some form and to some degree, all rate of return studies, including T&W99, which do not take into account screening effects, are likely to over-estimate the rates of return to education and schooling.

The lack of methodological clarity in T&W99 is not isolated to the use of the alpha factor. More serious is the inexplicable omission of measurements of costs. Despite claiming to measure social and private rates of return, nowhere did T&W99 explain what they have taken to constitute social and private costs, and how they have measured them. Instead, readers are referred in no more than a footnote to an obscure and inaccessible unpublished academic exercise[5] by Wong (1996). From OECD (2002) and similar studies of private and social rates of return to education, one presumes that in T&W99, for the private costs of education, transport costs, school fees, a measure of opportunity cost, including income forgone during education, and other costs accrued to the individual were summed up, while social costs also included government subsidies to education.

On the benefits side of the cost-benefit analysis, T&W99 had focused on earnings, while acknowledging that to do so would be to treat education as purely an investment good and disregarding its consumption benefits, i.e. the positive utility that one derives from consuming education for its own sake (Vaizey, 1973). This neglect of education as also a consumption good would necessarily understate the private rate of return to the individual. Furthermore, because of the omission of external and non-pecuniary benefits to education, such as its positive impact on law and order (Feinstein, 2002), social cohesion (Bowles & Gintis, 1976), both the health of oneself and others (Feldman et al., 1989), etc, both private and social rates of return would tend to be underestimated, especially since negative externalities to education were unlikely to be large. In T&W99’s words, this ‘suggests that the return on education calculated from earnings is a ‘minimum return’, the lower bound of a true return.’[6] Therefore, even if the return to educational programmes calculated from earnings is very low, it does not necessarily mean that individuals and the state should halt their consumption and investment in them. For example, in a country with massive chronic rates of unemployment, a primary school education might not yield a higher future income to an individual, but the basic knowledge of basic hygiene, basic linguistic and mathematical literacy, etc. that he derives from his primary schooling are of great external and non-pecuniary benefit and arguably well worth the cost.

Moving on to the findings of the article, T&W99 outlined two diverging patterns of results from rates of return studies: the first characterised by Psacharopoulos (1991) and the other by Carnoy (1972). Essentially, the former postulated and showed that across all regions, the rates of return to education are higher for lower levels of education (Psacharopoulos, 1994), i.e. primary education has a higher rate of return than secondary education, which is in turn higher than that for tertiary education. This is due to the law of diminishing returns: the returns to initial investments in education, e.g. in primary schooling, would be very high but returns decline with each subsequent level of investment, such as in the form of lower rates of return to degrees (Psacharopoulos, 1994). The latter predicted that levels of economic development have an integral role in determining the rates of return to each stage of education. As Mingat & Tan (1996) asserted and proved, in low income countries, because productive processes and economic transactions are relatively simple, primary education provides sufficient skills for most jobs and therefore has a high rate of return. There is at the same time a low demand for labour with secondary level schooling, meaning that these workers would only have a small, if at all, wage premium, over their primary school educated counterparts, and therefore the rates of return to secondary education is lower. In middle income countries, the greatest demand is for workers with secondary level education and skills, and therefore the rates of return to secondary education is highest. Then, in high income nations, the returns are highest for tertiary education. T&W99 claimed that ‘(t)he experience of Singapore is more in agreement with the view of Carnoy’[7], on the basis that in Singapore, the returns to tertiary education is higher than that to primary education. Since independence in 1965, Singapore has undergone rapid industrialisation and trade liberalisation in a bid for economic survival as a nation-state (Huff, 1994), which suggests that there has been a huge demand for tertiary educated engineers and other professionals since then, and which supports the predictions of Carnoy (1972).

Section D: Conclusion

In reviewing T&W99, one concludes that it is a weakly written paper. This is in part because of the lack of good publicly available data, and largely because of the authors’ own lack of clarity in explaining their methods. Because consumption benefits as well as external and non-monetary effects of education have been omitted, the rate of returns as calculated in T&W99 is very likely to be lower than the true value. It is probable that the rates of return to different stages and type of education would be mis-estimated to different degrees. For example, arguably the health benefits from basic hygiene awareness inculcated through primary education is far higher than other marginal health benefits from tertiary education. Therefore, if external benefits to primary education has been under-estimated to a larger degree than that of tertiary education, the predictions of Carnoy (1972) and Mingat & Tan (1996) may well be undermined, rendering Psacharapoulos’s (1991) results more plausible.

Further work needs to be done in order to ascertain the true rate of return to education in Singapore. More data, including on gender, family background, prior ability, has to be collected on a regular basis to facilitate the calculation of the alpha factor or the use of the Mincerian approach, and to allow for more robust time series comparisons.

Asher, M. & Newman, D. (2001). ‘Hong Kong and Singapore: Two Approaches to the Provision of Pensions in Asia’, Journal of Pension Management, Vol 7, No. 2.

Bowles, S. & Gintis, H. (1976). Schooling in capitalist America: educational reform and the contradictions of economic life, New York: Basic Books

Carnoy, M. (1972). ‘The rate of return to schooling and increase in human resources in Puerto Rico’, Comparative Economic Review, Vol. 16, 68-86

Carnoy, M. (1995). ‘Rates of return to education’. In Carnoy, M. (Ed.) International Encyclopedia of Economics of Education, Oxford: Pergamon Press.

Denison, E. (1964). Measuring the contribution of education (and the residual) to economic growth, Paris: OECD

Feinstein, L. (2002). The quantitative estimates of the social benefits of learning 1 (Crime), London: Centre for Research on the Wider Benefits of Learning.

Feldman, J., Makuc, D., Kleinman, J., Cornoni-Huntley, J. (1989). ‘National trends in educational differentials in mortality’, American Journal of Epidemiology, Vol 129.

Hayek, F. (1945). ‘The use of knowledge in society’, American Economic Review, Vol 19, No. 4, 519-530.

Khoo, C (1981). Census of population of 1980, Singapore, Singapore: Department of Statistics

Mill, J. S. (1994). Principles of political economy and chapters on socialism, Oxford: Oxford University Press.

Mingat, A. & Tan, J. (1996). The full social returns to education: estimates based on countries’ economic growth performance, World Bank

Morris, V. and Ziderman, (1971). ‘The economic return on government intervention in higher education in England and Wales’, Economic trends, 211, 20-28

OECD (2002). ‘Returns to education: private and social rates of return to education and their determinants’, Education at a glance, Paris: OECD.

Huff, W. (1994). The economic growth of Singapore: trade and development in the 20th century, Cambridge: Cambridge University Press.

Psacharopoulos, G. (1994). ‘Returns to investment in education: a global update’, World Development, Vol. 22, No. 9, 1325-1343

Psacharopoulos, G. (1991). The economic impact of education: lessons for policy makers, San Francisco: ICS.

Singapore Ministry of Labour (1981-1994). Report on the labour force of Singapore, Singapore: Ministry of Labour

Singapore Ministry of Labour (1994). Report on wages in Singapore, Singapore: Ministry of Labour.

Singapore Inland Revenue Department (1981). Report on the Survey of Employment of Graduates, 1980, Singapore: Inland Revenue Department

Spence, M. (1973). ‘Job market signalling’, Quarterly Journal of Economics, Vol 87, No. 3, 355-374

Toh, M. & Wong, S. (1999). ‘Rates of return to education in Singapore’, Education Economics Vol 7, No. 3, 235-252.

Vaizey, J. (1973). The economics of education, London: Macmillan

Wong, C. (1996). A study on the income and educational qualifications in Singapore, National University of Singapore unpublished academic exercise.

[1] Also social security and mandatory savings scheme (Asher & Newman, 2001)
[2] Footnote 10 in T&W99, page 250
[3] As stated in Footnote 10 in T&W99, page 250
[4] T&W99 page 238
[5] I have attempted to obtain the article via email correspondence with the author, to no avail.
[6] T&W99, Page 238
[7] T&W99, Page 237


Post a Comment

<< Home