Trade globalization, economic development and the importance of education-as-knowledge Salvatore J. Babones University of Sydney

Abstract It is widely asserted that globalization puts a premium on knowledge, but in fact there is no empirical correlation across countries between globalization and returns to education. One reason for this discrepancy may be that education is not everywhere coequal with knowledge. In this article the overall contribution of education to income is modelled as the sum of the contributions of two components of education, education-as-knowledge and education-as-credential. Assuming that the former dominates in developed countries while the latter dominates in developing countries, it is possible to separate these two effects. In a broadly comparative analysis of returns to education in 80 countries using World Values Survey data, globalization is found to be positively associated with education in developed countries but negatively associated with education in developing countries, consistent with the model. These results are robust in the face of controls for the supply and demand for education. Keywords: development, education, globalization, knowledge, stratification, trade

Does globalization increase the premium placed on knowledge in the world today? There certainly are many seemingly convincing arguments as to why it should. In poor countries, those who are literate and numerate seem more likely to secure jobs with multinational firms and their local contractors, while in rich countries globalization creates highly paid opportunities for those who work in such knowledge-economy sectors as banking, marketing, design and the arts. The ability to leverage the opportunities created by the rise of the internet surely increases the economic importance of technical

Journal of Sociology © 2009 The Australian Sociological Association, Volume 46(1): 45–61 DOI:10.1177/1440783309337674 www.sagepublications.com

46 Journal of Sociology 46(1)

knowledge. Globalization should also put a premium on those with foreign language skills in countries of all income levels. There are some countervailing arguments: that globalization in specific industries can lead to deskilling, that globalization is not so pervasive an influence as to have a detectable impact on local labour markets, or that globalization leads the most knowledgeable to leave poor countries altogether. Nonetheless, the general public policy discourse is one of the premium placed on knowledge by globalization. The promise (or threat) of globalization is used to promote education, not to retard it. Yet despite the widespread acceptance of the need to prepare workforces for the demands of today’s global knowledge economy, there has to date been little scientific research on the link between globalization and returns to knowledge. In fact, ‘knowledge’ as a concept is very difficult to measure, and even more difficult to measure cross-nationally. The OECD PISA (Programme for International Student Assessment) studies benchmark knowledge levels across a small group of rich countries, but only the knowledge of children, not of adults. Instead of knowledge, most social surveys measure the allied concept, ‘education’. If by ‘knowledge’ we mean ‘education’, then the evidence that globalization puts a premium on knowledge is weak indeed. Evidence will be presented later in this article that across a broad panel of 80 countries, national levels of the importance of education for income are actually negatively (though not significantly) related to national levels of trade globalization (see Table 4, Model 1, p. 000). This negative, non-significant result holds up even when controlling for proxies for the supply of and demand for education (Table 4, Model 2). At first blush, it appears that all the attention educators heap on globalization as a renewed rationale for the importance of their work is just so much hype. Globalization is not broadly and universally correlated with returns to education. This raw empirical fact, however, seems unsatisfying as a final conclusion. The lack of empirical correlation between globalization and returns to education does not necessarily mean that globalization is uncorrelated with returns to knowledge. Knowledge may typically be measured using levels of education, but this practice confounds knowledge with credentialing: education results in credentials that only loosely correspond to the actual knowledge gained in school. Credentials may not capture people’s actual levels of knowledge, may not all represent the same kinds of knowledge, and may not incorporate equivalent levels of knowledge between countries. Moreover, much important knowledge is gained outside of the school setting, perhaps most of the knowledge that is relevant for economic activities (Lucas, 1988). In order to estimate the importance of knowledge itself for individual income, without confounding knowledge with education, it is necessary to formulate a model that captures the effect of knowledge as distinct from the mere credentialing inherent in education. In the next section, such a model is elaborated, taking advantage of hypothesized differences in the relative importance of knowledge and

Babones: Trade globalization, economic development 47

credentialing in developed versus developing countries. Following this, in the third section of this article, data sources for an empirical test of the relationship between globalization and returns to knowledge are delineated. In the fourth section the statistical case for such a relationship is laid out. The fifth and final section includes a discussion of the implications of the results of this article both for policy and for future research.

Education, knowledge and income That education is closely related to income for individuals living in the same society is well established. Perhaps the most widely tested and confirmed model in sociology is the status attainment model (Blau and Duncan, 1967) linking education to income over the life course and across generations. The status attainment model has been applied repeatedly to regional and national samples in the United States, as well as internationally in countries such as Germany and Poland (Krymkowski, 1991), Hungary (Luijkx et al., 2002), Japan and Korea (Kim, 2003) and Australia (Marjoribanks, 1996). Sociologists have carefully verified that the relationship between individual education and income is robust and independent of such factors as individuals’ intelligence, individuals’ attitudes and aspirations, and parents’ social status (Alexander et al., 1975). They have not, however, paid much attention to how the education–income relationship might be conditioned on important macro-level social forces like technological change, increasing education levels, or increasing trade globalization (Breen and Jonsson, 2005). There is also an extensive literature in economics on the education–income relationship (characterized as ‘returns to education’ or the ‘skill premium’) that does take account of macro-level forces. This literature has focused primarily on the role of technological change in increasing the demand for skills, and thus the returns to education. Despite elaborate mathematical argumentation that technology increases the demand for skills, the only strong empirical evidence is for the United States (Katz and Autor, 1999). Notably, this increase in returns to education in the US has occurred despite a parallel increase in the supply of skills, operationalized as the educational attainment of the population (Acemoglu, 2002). Economic models also predict that trade globalization should increase returns to education in both developed and developing countries (Acemoglu, 2003), but these predictions have similarly not been borne out in empirical cross-national studies of the education– income relationship (Ethier, 2005). One reason it may be difficult to observe a relationship between globalization and the importance of knowledge for individual economic success is that individual education only imperfectly operationalizes individual knowledge. This may be especially true when comparing countries of vastly different levels of development. As the status attainment literature in sociology demonstrates, education is significantly related to adult income even after controlling for knowledge, as well as such potentially endogenous

48 Journal of Sociology 46(1)

factors as intelligence, aspirations and social support (Jencks et al., 1983). This suggests that the total effect of education on income incorporates the effects of both the knowledge gained in school and a second component, the educational credential itself. One possible key to disaggregating the effects of education-as-knowledge and education-as-credential is to surmise that their relative strengths differ systematically across countries. It might be suggested that in developed countries, education is relatively more valued as an indicator of knowledge, while in many developing countries, education is relatively more valued as a credential in itself. Just such a dichotomy was one of the few sources of agreement between classical modernization theorists (Inkeles and Smith, 1974) and their critics (Portes, 1973). If true, this tendency would be compounded by the fact that, in recent decades, developing countries have been encouraged to educate their populations well beyond the levels that are required by the productive needs of their economies (Easterly, 2001; Pritchett, 2001). In this situation, individuals are still likely to be stratified in labour markets according to their levels of education, but the knowledge embodied in that education would itself be increasingly irrelevant for individual income, since individuals are generally overeducated for the labour market roles they are expected to play. A schematic model disaggregating the importance of education for individual income into effects due to education-as-knowledge and effects due to education-as-credential is depicted in Figure 1. The horizontal axis in Figure 1 represents a country’s level of development. Lines Ed, Kn, and Cr represent the total effect of education, the effect of education-as-knowledge, and the effect of education-as-credential, respectively. In this model, the overall importance of measured education for adult income (Ed) is constant across countries at all levels of development: as the importance of education-as-knowledge for adult income (Kn) increases, the importance of education-as-credential (Cr) declines. This model is consistent with the observation that the overall relationship between education and income is constant across countries of all levels of development. It is also consistent with the empirical literature reporting that the strongest effect of technology change on returns to education is to be found in the US, which possesses the world’s richest economy. If it is true that globalization increases the economic returns to knowledge, this will be reflected in an increase in the absolute importance of educationas-knowledge for adult income. Such an effect is depicted in Figure 2 by the shift from Kn to Kn′. Globalization increases the importance of knowledge everywhere, but more so in developed countries, where knowledge is surmised to be relatively more important for determining people’s incomes than in developing countries, where knowledge is thought to be less important as a determinant of income. As a result, the slope of the line representing the importance of knowledge for income across countries of differing levels

Importance for adult income

Babones: Trade globalization, economic development 49

Ed Cr

Kn

Less developed

More developed

Figure 1: Importance of education by development status (schematic). Across countries at all levels of development, the importance of education for adult income is roughly the same (line Ed). The compositional breakdown of the importance of education, however, varies systematically by level of development. In less developed countries, the credential embodied in education is relatively more important for determining one’s income (line Cr), while in more developed countries the knowledge embodied in education is relatively more important (line Kn).

of development steepens when those countries globalize. Globalization would be expected to exacerbate the gap between rich and poor countries in the returns to knowledge. This is another way of saying that globalization creates wider economic opportunities for knowledgeable Europeans than it does for knowledgeable Africans. If it is true that globalization reduces the economic returns to mere credentials (aside from any knowledge they might embody), this will be reflected in a decrease in the absolute importance of education-as-credential for adult income. Such an effect is depicted in Figure 2 by the shift from Cr to Cr′. Globalization undermines the importance of (mere) credentials everywhere, but more so in developing countries, where credentials are surmised to be relatively more important for determining people’s incomes than in developed countries, where credentials are thought to be less important as a determinant of income. As a result, the slope of the line representing the importance of credentials for income across countries of differing levels of development becomes less steep when those countries globalize. Globalization would be expected to diminish the gap between rich and poor countries in the returns to (mere) credentials, pushing both towards (but not to) zero. This is another way of saying that globalization

50 Journal of Sociology 46(1)

Importance for adult income

Importance of education for income decines in developing countries

Importance of education for income increases in developed countries

Ed’ Kn’

Ed

Cr Up

Down Kn Cr’

Less developed

More developed

Figure 2: Effect of globalization on the importance of education (schematic). Globalization increases the relative importance of education-as-knowledge (line Kn shifts to line Kn′) and reduces the importance of education-as-credential (Cr shifts to Cr′). Since these two components of education are unequally represented in developed and developing countries, the net result is a rotation of the line Ed (the overall importance of education) to Ed′. The overall importance of education for income increases with globalization in developed countries and decreases with globalization in developing countries.

undermines the income earning power of highly credentialed Africans more than it does the income earning power of highly credentialed Europeans. Thus, if globalization does in fact increase the economic returns to knowledge, it should be possible to detect this by comparing the effects of globalization on returns to education in developed versus developing country panels. Though the slopes of lines Kn and Cr have been drawn symmetrically in Figure 1 and the multiplicative effects of globalization on each of them expressed as reciprocals of each other in Figure 2, the logic of the model does not depend on these particular parameters. In fact, all that must be assumed is that the effects depicted at Kn and Cr are linear and that the effects of globalization are multiplicative; the slopes of the lines and the magnitudes of the globalization effects are not important. A simple algebraic proof of these assertions is presented in Appendix 1 below. The differential effect of globalization on the importance of educationas-knowledge versus education-as-credential can thus be tested empirically using differences in the relationship between globalization and the overall importance of education in developed versus developing countries. The importance of education in a country can be operationalized as the

Babones: Trade globalization, economic development 51

correlation of individuals’ levels of education with their adult incomes. Based on the arguments presented above, globalization would be expected to be negatively associated with the importance of education in developing countries (where education compositionally represents mainly educationas-credential) but positively associated with the importance of education in developed countries (where education compositionally represents mainly education-as-knowledge). Since these two predicted effects are diametrically opposed, it should be relatively easy to detect the difference between them.

Data and data sources This study takes advantage of a unique cross-national dataset, the World Values Survey (WVS), to estimate the strength of the correlation between education and adult income across 80 countries (World Values Survey, 2006). WVS waves 1 and 2 are excluded because of the limited numbers of cases with data available on the key variables (just six countries). WVS waves 3 and 4 (1994–2004), on the other hand, collectively include national probability sample data on education and income for 100,383 adults in 81 countries around the world, including 53 developing and 28 developed countries. Singapore is excluded from the analyses reported here because it is a strong positive leverage point in the developed country panel. Including Singapore strengthens the results reported below; excluding Singapore is thus a conservative decision. Waves 3 and 4 are combined, rather than used separately in an unbalanced, multiple-observation design because the majority of countries were surveyed in only one of the two waves. Interviews for waves 3 and 4 were conducted over the years 1994–2004; the midpoint of this period (1999) is used as the reference year for this study. Education is operationalized using WVS variable X025, an eight-point ordinal scale of the respondent’s highest level of educational attainment. Income is operationalized using WVS variable X047, a 10-point ordinal scale of respondent income designed to approximate the income deciles in each WVS country. Only data for unambiguously working-age adults (age 25–54) were used. The resulting dataset includes education and income data for 100,383 adults in 81 countries. The importance of education in status attainment is operationalized as the simple correlation of education and income in each country. Correlations were computed based on combined data from both waves 3 and 4, using the sampling weights provided in the WVS database. Respondents from the Northern Ireland and Great Britain surveys have been combined to form a single United Kingdom total. The resulting correlations for 81 countries are reported in Table 1. All correlations are significant at the p < .001 level, and all are in the expected direction (positive) except for that for Armenia, which is mildly negative. The Armenian survey was conducted in 1997, a time of dramatic political upheaval; this may explain the perverse result.

52 Journal of Sociology 46(1)

Table 1: Countries’ levels of correlation between education and income (c. 1999) Country

r

Albania Algeria Argentina Armenia Australia Austria Azerbaijan

0.446 0.185 0.455 −0.096 0.305 0.295 0.154

Bangladesh Belarus Belgium Bosnia and Herzegovina Brazil Bulgaria Canada Chile China Colombia Croatia Czech Republic Denmark Dominican Republic Egypt, Arab Rep. El Salvador Estonia Finland France Georgia

Country

r 0.034 0.251 0.289 0.329 0.380 0.162 0.214

Country

r

0.346 0.186

Germany Greece Hungary Iceland India Indonesia Iran, Islamic Rep. Iraq Ireland

0.266 0.485

0.349 0.353

Israel Italy

0.339 0.361

0.406 0.300 0.400 0.509 0.181 0.584 0.289 0.344

Japan Jordan Korea, Rep. Kyrgyz Republic Latvia Lithuania Luxembourg Macedonia, FYR

0.196 0.303 0.292 0.240 0.277 0.190 0.405 0.393

Pakistan Peru Philippines Poland Puerto Rico Romania Russian Federation Saudi Arabia Serbia and Montenegro Singapore Slovak Republic Slovenia South Africa Spain Sweden Switzerland Taiwan Tanzania Turkey

0.623 0.416 0.243 0.282 0.297 0.241 0.139

0.145 0.307

Malta Mexico

0.368 0.375

Uganda Ukraine

0.188 0.167

0.359

Moldova

0.230

United Kingdom

0.241

0.690 0.282 0.247 0.306 0.106

Morocco Netherlands New Zealand Nigeria Norway

0.447 0.313 0.269 0.434 0.181

United States Uruguay Venezuela, RB Vietnam Zimbabwe

0.323 0.540 0.380 0.077 0.575

0.193 0.361 0.640 0.360 0.551 0.508 0.340 0.251 0.325 0.400 0.567 0.465

Countries were categorized by development status (developing versus developed) on the basis of the official World Bank categorization for 1999, the midpoint of the WVS study period (World Bank, 2000). There is no significant difference in the strength of the education-income correlation in developing versus developed countries (t79 = .332). There are, however, significant differences among developing countries by region (also defined using official designations from World Bank, 2000). Mean levels of the education–income correlation are reported in Table 2. The strongest levels are found in the developing countries of Africa, Latin America and south Asia, where the correlation between education and income is stronger than in developed countries. The correlation between

Babones: Trade globalization, economic development 53

Table 2: Strength of the education–inequality correlation by region (c. 1999)

Developing countries East Asia and Pacific Europe and Central Asia Latin America and Caribbean Middle East and North Africa South Asia Sub-Saharan Africa Developed countries All countries

Mean

SD

N

0.327 0.213 0.261 0.466 0.281 0.450 0.454 0.316 0.323

0.153 0.120 0.123 0.115 0.096 0.151 0.159 0.118 0.141

53 5 23 10 7 3 5 28 81

education and income is weakest in the developing countries of east Asia, eastern Europe and the Middle East. In all regions, however, it is highly significantly greater than zero. Data for country-level ecological variables come from a mix of published sources. Following the most common practice in the globalization literature (Babones, 2007), globalization is operationalized as trade globalization, using foreign trade as a proportion of GDP. Trade as a proportion of national income is taken for most countries from World Development Indicators (WDI) database. Data are for the year 1999. In three cases (Iraq, Singapore, Taiwan) data were not available in the WDI, so data from the CIA World Factbook were used instead (CIA, 2007). World Factbook data for 1999 are not reported in the same units as the WDI trade series, so data for 2006 were used instead for these three countries. Singapore is a 3.5 standard deviation positive outlier in trade as a ratio to national income. The ratio of trade to national income exhibits a strong positive skew. This is natural, since trade is bounded by zero on the left but is unbounded on the right (a country’s level of trade can, and in many cases does, exceed its level of national income). As a result, the ratio of trade to national income exhibits a positive skew on the order of 1.5. This skew can be reduced effectively to zero through logging. Trade has typically not been logged in the globalization literature, but the empirical and methodological arguments for logging the data are incontrovertible. Two control variables are used in addition to the main dependent and independent variables. These are the supply of, and demand for, education. A loose proxy for the demand for education (both knowledge and credentials) in a country is its level of economic development: more advanced economies require more highly educated workforces. Development is operationalized here as national income per capita. National income per capita data are taken from the World Bank (2007) WDI Atlas series for gross national product per capita. Figures for 1999 are used. As above, for Iraq, Singapore and Taiwan CIA World Factbook data for 2006 are used. National income per capita is logged to eliminate positive skew.

54 Journal of Sociology 46(1)

A country’s supply of education can be operationalized directly using its average educational level. Cross-national data on average educational attainment are notoriously incomplete and error-prone, but the secondary school gross enrolment ratio serves as a reasonable proxy. The gross enrolment ratio is the ratio of the population of secondary school age to the number of children actually enrolled in secondary school. It tends to be highly correlated with average educational attainment where such data are available. The secondary school enrolment data used here are mainly 1999 figures from UNESCO’s Global Education Database (UNESCO Institute for Statistics, 2007). Where 1999 data were unavailable the next available year (2000–5) was used. In one case (Sweden) the published 1999–2003 figures were clearly aberrant, so the 2004 entry was used instead. Four countries (Bosnia and Herzegovina, Puerto Rico, Singapore and Taiwan) were completely missing in the UNESCO data. For these countries, the secondary school gross enrolment ratio was estimated through a regression on logged national income per capita (R2 = .566). Since education and national income are not primary focuses of this study, the minor colinearity between them introduced by this procedure is not a major concern. Raw correlations of all variables used in this study are reported in Table 3.

Table 3: Correlations of key variable for complete, developing country and developed country panels (c. 1999) – Singapore excluded as an influential point in correlational results (0) (0) (1) (2) (3)

Importance of education for income Trade globalization Supply of education Demand for education

1 −0.124 −0.237* 0.009

(0) (1) (2) (3)

Importance of education for income Trade globalization Supply of education Demand for education

(0) 1 −0.331* −0.260+ 0.199

(0) (1) (2) (3)

Importance of education for income Trade globalization Supply of education Demand for education

(1) 1 0.439* −0.173 −0.282

+

p