Was higher education a major channel through which the US became an economic superpower in the 20th century?

Cook, Adam; Ehrlich, Isaac Working Paper Was higher education a major channel through which the US became an economic superpower in the 20th century? ADBI Working Paper, No. 820 Provided in Cooperation with: Asian Development Bank Institute (ADBI), Tokyo Suggested Citation: Cook, Adam; Ehrlich, Isaac (2018) : Was higher education a major channel through which the US became an economic superpower in the 20th century?, ADBI Working Paper, No. 820, Asian Development Bank Institute (ADBI), Tokyo This Version is available at: https://hdl.handle.net/10419/190241 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/3.0/igo/ ADBI Working Paper Series WAS HIGHER EDUCATION A MAJOR CHANNEL THROUGH WHICH THE US BECAME AN ECONOMIC SUPERPOWER IN THE 20TH CENTURY? Adam Cook and Isaac Ehrlich No. 820 March 2018 Asian Development Bank Institute The Working Paper series is a continuation of the formerly named Discussion Paper series; the numbering of the papers continued without interruption or change. ADBI’s working papers reflect initial ideas on a topic and are posted online for discussion. Some working papers may develop into other forms of publication. The Asian Development Bank recognizes “China” as the People’s Republic of China, and “South Korea” as the Republic of Korea. In this report, “$” refers to US dollars. Suggested citation: Cook, A. and I. Ehrlich. 2018. Was Higher Education a Major Channel through which the US Became an Economic Superpower in the 20th Century?. ADBI Working Paper 820. Tokyo: Asian Development Bank Institute. Available: https://www.adb.org/publications/higher-education-major- channel-which-us-became-economic-superpower-20th-century Please contact the authors for information about this paper. Email: [email protected], [email protected] The paper is an extension of Isaac Ehrlich’s NBER Working Paper 12868 titled “The Mystery of Human Capital as Engine of Growth, or Why the US Became the Economic Superpower in the 20th Century”, as well as his chapter with the same title in The Mystery of Capital and the Construction of Social Reality, B. Smith, D. Mark, and I. Ehrlich, eds. Chicago: Open Court, 2008. A related version was presented in the Asian Development Bank Institute and Asian Growth Research Institute workshop on “Public and Private Investment in Human Capital and Intergenerational Transfers in Asia,” held at Hotel Harmonie Cinq, Kitakyushu City, 14–15 November 2017. We are indebted to Robert Tamura for sharing with us his data on historical enrollments in tertiary institutions. Adam Cook is a professor in the School of Business, Department of Economics at the State University of New York at Fredonia. Isaac Ehrlich is a SUNY and UB distinguished professor of economics and Melvin H. Baker professor of American Enterprise at the State University of New York, Buffalo. The views expressed in this paper are the views of the author and do not necessarily reflect the views or policies of ADBI, ADB, its Board of Directors, or the governments they represent. ADBI does not guarantee the accuracy of the data included in this paper and accepts no responsibility for any consequences of their use. Terminology used may not necessarily be consistent with ADB official terms. Working papers are subject to formal revision and correction before they are finalized and considered published. Asian Development Bank Institute Kasumigaseki Building, 8th Floor 3-2-5 Kasumigaseki, Chiyoda-ku Tokyo 100-6008, Japan Tel: +81-3-3593-5500 Fax: +81-3-3593-5571 URL: www.adbi.org E-mail: [email protected]g © 2018 Asian Development Bank Institute ADBI Working Paper 820 Cook and Ehrlich Abstract This paper offers a thesis for why the United States (US) overtook the United Kingdom (UK) and other European countries in the 20th century in both aggregate and per capita GDP as a case study of recent models of endogenous growth, where “human capital” is the engine of growth. By human capital we mean an intangible asset, best thought of as a stock of embodied and disembodied knowledge comprising education, information, entrepreneurship, and productive and innovative skills, which is formed through investments in schooling, job training, and health as well as through research and development projects and informal knowledge transfers (cf. Ehrlich and Murphy 2007). The conjecture is that the ascendancy of the US as an economic superpower in the 20th century owes considerably to its faster human capital formation relative to that of the UK and “old Europe.” This paper assesses whether the thesis has legs to stand on through both stylized facts and a supplementary quasi-experimental empirical analysis. The stylized facts indicate that the US led other major developed countries in schooling attainments per adult population member, beginning in the latter part of the 19th century and lasting throughout the 20th century, especially at the secondary and tertiary levels. The quasi-experimental analysis constitutes the first attempt to test the hypothesis that the US’s ascendancy to a major economic power stems largely from the impact of the first Morrill Act of 1862, which launched the public higher education movement in the US through the establishment of land grant colleges and universities across the nation during the latter part of the 19th century. The higher education movement appears to have spearheaded a higher long-term rate of growth in per capita income in the US relative to the UK and other major European countries. Keywords: human capital, endogenous growth, Morrill Act, higher education, treatment effects, US JEL Classification: H1, I2, N1, N3, O0, O4, C21 ADBI Working Paper 820 Cook and Ehrlich Contents PROLOGUE ............................................................................................................................ 1 1. THE “MYSTERY” OF GROWTH: THE HUMAN CAPITAL HYPOTHESIS .................. 2 2. SUPPORTING EVIDENCE ON EDUCATIONAL ATTAINMENTS .............................. 5 2.1 Data on Schooling Attainments in the US versus OECD Countries over the Last Century ...................................................................................... 5 2.2 Recent Evidence from the OECD’s Education at a Glance, 1998 and 2003 ... 5 3. HOW THE US SCHOOLING ADVANTAGE EMERGED: MAJOR SOURCES AND TRENDS ............................................................................................................. 7 3.1 The Secondary Schooling Advantage ............................................................. 7 3.2 The Morrill Acts and the Land Grant Institutions of Higher Learning ............... 7 3.3 The GI Bill of 1944 ........................................................................................... 8 3.4 Immigration and the Brain Drain ...................................................................... 8 4. WHENCE THE DIVERGENCE? CONTRIBUTING FACTORS ................................... 9 4.1 Educational Templates .................................................................................... 9 4.2 Economic Development ................................................................................ 10 4.3 Feedback Wealth Effects .............................................................................. 10 4.4 Educational Policies ...................................................................................... 10 4.5 The Political–Economic Systems .................................................................. 10 5. LINKING HUMAN CAPITAL FORMATION WITH ECONOMIC GROWTH ............... 11 5.1 The Endogenous Growth Hypothesis: Human Capital as the Engine of Growth ....................................................................................................... 11 5.2 The Special Role of Higher Education in Economic Growth ......................... 13 5.3 The Role of Underlying Factors ..................................................................... 14 6. EVIDENCE LINKING EDUCATION AND PRODUCTIVITY GROWTH ..................... 15 6.1 Evidence from Growth Accounting ................................................................ 15 6.2 Evidence from Rates of Return to Education ................................................ 16 6.3 Linking Investment in Schooling and Per Capita Income Growth .................. 16 6.4 The Role of the 1862 Morrill Act in Enhancing Human Capital Formation and the Pace of per Capita Income Growth in the US Relative to the UK ..... 18 EPILOGUE: LOOKING BACK AND LOOKING AHEAD ....................................................... 22 REFERENCES ..................................................................................................................... 25 APPENDIX A: TABLES ......................................................................................................... 27 APPENDIX B: Data Used in the DiD Regressions (Tables 11 and 11a) .......................................... 37 Figures ...................................................................................................................... 38 ADBI Working Paper 820 Cook and Ehrlich 1 PROLOGUE Common to the bulk of the “new” economic growth and development literature is the idea that the process by which less-developed countries break out of a poverty trap and achieve steady, self-sustaining growth in their real per-capita income is predicated on the persistent production and accumulation of “human capital.” This powerful concept is wrapped up in three layers of mystery. First, unlike physical capital, human capital is not a tangible asset. How, then, can we account for it empirically? Second, what explains its continuous formation over time? Third, how is such formation transformed into growth in real output and personal income? One of the objectives of this essay is to unwrap this apparent mystery through the exposition of a general-equilibrium paradigm of economic development whereby human capital, or knowledge, is the engine of growth, parental and public investments in children’s education empower its accumulation, and institutional and policy variables enable its productive returns and impact on long-term growth. We develop the paradigm in the context of an institutional environment that ensures a well-functioning market economy that competitively rewards and efficiently allocates human capital, measured imperfectly using indicators of schooling and training, to productive activities. The model also recognizes, however the role of externalities, such as market imperfections that adversely affect the accessibility and financing costs of schooling for those with borrowing constraints, or informal knowledge spillover effects emanating from workers and entrepreneurs with superior education and skills, which enhance the productivity of others with whom they interact. The way in which the political and legal frameworks governing the economy internalize these externalities may vary across different economies and as a consequence of accommodating economic and educational public policies, especially concerning higher education. Such factors ultimately account for differential long-term growth patterns in different countries. A more specific objective of the presentation is to illustrate the power of the “human capital hypothesis” to explain the differences that we observe in the long-term growth dynamics across specific countries. The case in point is the emergence of the US as the world economic superpower, overtaking the UK and Europe in general. The US was a relatively poor country throughout much of the 19th century. In the last few decades of that century, and especially during the 20th century, however, the US overtook the UK and other major European countries and then developed a considerable advantage over them not only in gross domestic product but also in per-capita GDP. The comparison of the US with the UK is not just because the UK had reigned as the world economic superpower, at least through the early part of the 19th century, but also because the US had inherited its basic institutional and cultural setting from the UK since its inception as a colony of the UK and a destination country for largely English-speaking immigrants. What may be less well known is that, over the same period, the US developed a considerable lead over Europe in the schooling attainments of its labor force, especially at the higher education level. The gap remained significant throughout the entire 20th century, although it narrowed in the latter part of it and is continuing to narrow in the current decade. Largely accounting for this gap was the massive high school movement of 1915–1940, but an independent lead emerged as early as the 1860s with the US foray into tertiary education beginning with the first Morrill Act of 1862 and continuing especially with the massive higher education movement following World War II. A basic argument of this paper is that the US lead in knowledge formation, ADBI Working Paper 820 Cook and Ehrlich 2 imperfectly measured as higher educational attainments, was perhaps a major, if not the major instrument through which the US overtook Europe as the economic superpower in the 20th century. To illustrate the case empirically, it is worth noting that, according to popular measures of real income often used for international comparisons – GDP, adjusted by purchasing power parity – the US maintains a considerably larger level of per-capita income relative to practically all the top 25 countries in the world, excluding small countries with populations of fewer than 5 million in 2013 (see Appendix A, Table A). In the early 1800s, however, the US had levels of GDP and GDP per capita that were considerably below those of the UK, and it was not until 1872 for the GDP and 1905 for the GDP per capita that the US overtook the UK. Figures 1 and 2 (see Appendix B) illustrate the comparisons poignantly. Abstracting from year-to-year and cyclical fluctuations, both the US and the UK graphs relating the logarithm of the GDP or the GDP per capita to chronological time appear to resemble an upward-sloping straight line in the long term. The slope of each line represents the long-term annual growth rate of the GDP or GDP per capita. The fundamental difference is that the slopes are higher for the US than for the UK. In other words, the US has overtaken the UK, because its long-term growth rates have been higher: over the 142-year period 1871–2012 (starting at the point of overtaking), the US versus the UK GDP growth rates were 3.31% versus 1.88% per annum, while the corresponding percapita GDP growth rates were 1.8% versus 1.4%.1 In recent decades, these gaps have narrowed. For example, over the period 1961–2012, the comparative growth rates of the GDP in the US versus the UK were 2.99% versus 2.13%, while those for the per capita GDP were 1.95% versus 1.89%, respectively.2 Our basic thesis is that the differences in the long-term per capita income growth stem primarily not from differences in physical stocks, including land or other natural resources, but from differences in the rates of growth of human capital. Both human capital formation and its impact on growth, however, are ultimately conditional on supportive institutional and policy factors that reward knowledge formation and innovative entrepreneurship. In the following, we investigate whether this hypothesis is defensible. 1. THE “MYSTERY” OF GROWTH: THE HUMAN CAPITAL HYPOTHESIS The cause of the differences in wealth across nations has been a key puzzle of economic science since Adam Smith. Logically, the question involves both static and dynamic elements: why some nations perform better than others economically at a particular point in time, and why some nations become more successful than others over time. In the terminology of the current literature on economic growth and development, this two-part question relates to the determinants of the long-term rate of 1 We take these statistics from the Maddison Project Database updated by Bolt and Zanden (2014). We convert all the figures into 1990 US dollars using the Geary–Khamis purchasing power parity (PPP) method. For 2009–2012, we compute the GDP using the real GDP growth rate estimates from the IMF. Similar graphs apply to other major European countries as well. For example, the growth rates of the GDP and GDP per capita (in parentheses) over the period 1850–2012—starting when the US overtook other major European countries in per capita GDP—were 3.34 (1.74) for the US; 1.9 (1.42) for the UK; 1.97 (1.6) for France; 2.21 (1.66) for Germany; 2.15 (1.53) for Italy; and 2.36 (1.67) for Spain. 2 The shorter-term trends have been uneven for other major European countries. Over the period 1961–2003, for example, the percapita GDP growth rate in France and Italy was 0.21% and 0.40% higher than that in the US, respectively, while in Germany it was 0.14% lower. However, over the period 1976–2003, the US’s per – capita GDP growth was 0.28% higher than France’s, 0.47% higher than Germany’s, and .06% higher than Italy’s. ADBI Working Paper 820 Cook and Ehrlich 3 growth, as distinct from the level of per-capita real income or GDP, taking the latter to represent a scalar measure of personal economic welfare. A significant advance in the modern economic treatment of the problem emerged with the neoclassical growth model, which identifies the key factors contributing to a steady-state level of percapita income and its associated capital–labor (K/L) ratio under any exogenously given rate of population growth and level of production technology. The model thus attributes persistent growth in percapita income over time, which is a more relevant measure of private economic welfare than aggregate income, strictly to exogenous technological shocks. We can conveniently illustrate this inference through the following “neoclassical” aggregate production function: Y = B(T)F(L, K), (1) where Y is the economy’s aggregate output; F is a constant-returns-to-scale production function summarizing the impact of conventional labor (L) and physical capital (K) inputs on production; and B(T) represents a factor-neutral technological factor (T) which augments the impact of both inputs. In the standard neoclassical growth model, these inputs and percapita income can grow over time through a dynamic process involving sufficiently high levels of investment in physical capital that exceed population growth. If technology is exogenously determined, the standard model suggests that in a balanced growth equilibrium, the steady-state level of percapita real income (y) can approach the following steady state level: y* ≡ B(T0)f(k*), (1a) where f(k*) is subject to diminishing returns and k* ≡ (K/L)* is the “golden rule” or equilibrium capital to labor ratio under a given technology level, T0. Growth in the equilibrium percapita income level y* may thus occur, according to this analysis, through exogenous technological advances. We can interpret the role of technology, B(T), more broadly to include any and all factors that enhance the utilization of the labor and physical capital resources available to the economy at a certain point i[t]n time. In principle, therefore, this factor also subsumes the economic and regulatory policies that facilitate the operational efficiency of the market economy within which a country uses its economic resources—a point that we will further underscore in later sections. Like technology, we assume for simplicity that the economy obtains these factors exogenously. They affect the level of output per capita at a particular point in time. In the last two hundred years or so, however, the world has witnessed a relatively new phenomenon in economic history: persistent and seemingly self-sustaining growth in percapita real income over the long term in most of the so-called developed economies following the technological shock produced by the Industrial Revolution. Periodic and occasionally large business cycle disturbances notwithstanding, this phenomenon is continuing, although at a different pace in different countries. Furthermore, over the last century or so, the world has experienced episodes of economic takeoffs by less developed countries, transforming them from countries with relatively stagnant, low income levels into regimes of self-sustaining growth (e.g., the Asian Tigers), as well as episodes in which a relatively poor economy has overtaken a much wealthier one (e.g., the US versus Europe). If “exogenous” factors, such as accidental technological discoveries, are the key to this mystery, what accounts for the smooth and continuous, but also variable, productivity growth in different countries, ADBI Working Paper 820 Cook and Ehrlich 4 especially when any country can rapidly imitate and adopt technological discoveries originating in another country? The answer which much of the recently developed “endogenous growth” literature offers (see, e.g., the articles in the Journal of Political Economy 1990, special issue, Ehrlich ed.), relies on identifying “technology” as “human capital” and modeling continuous and self-sustaining technological advances as the outcome of persistent investment in human capital treated as a decision variable within a dynamic, general-equilibrium framework. We can perhaps best define the concept of human capital as an intangible asset – a stock of embodied and disembodied knowledge, comprising education, information, health, entrepreneurship, and productive and innovative skills, which is formed through investments in schooling, job training, and health, as well as through research and development projects and informal knowledge transfers (see Ehrlich and Murphy 2007). Following this definition, human capital has two inherent dimensions: “embodied” and “disembodied.” The first is knowledge embodied in workers, or skill, which augments the productivity of labor and physical capital inputs at a point in time. The second is creative knowledge, which flows from the minds of scholars, scientists, inventors, and entrepreneurs and increases their capacity to accumulate new knowledge. This “disembodied” knowledge emerges in papers, books, patents, and algorithms and results in technological advances—product and process innovations—at the firm and industry levels. It is thus more likely that individuals will acquire and produce this form of knowledge in tertiary institutions of teaching and research. While these types of human capital are distinct, they are also complementary, as creative knowledge feeds on previously accumulated embodied knowledge and facilitates the acquisition of new knowledge. In this view, technology as people popularly understand it—inventions, innovations, and scientific discoveries—does not “fall from heaven”: it stems from decisions that families, firms, and governments make to invest in schooling, job training, and research and development, making human capital the relevant “engine” or facilitator of growth. The fuel that feeds this engine is the rewards or rates of return on investments in knowledge formation or human capital, set by market forces and influenced by government policies. Skills and creative knowledge can accumulate continuously in a given economy, however, only if the underlying reward system in that economy supports sufficient investment in skills and creative knowledge beyond a critical level. How does one measure human capital empirically? The empirical literature associated with this concept typically identifies it as a function of years of schooling and job experience. Corresponding measures of educational quality, however, must supplement these measures. Also missing are education and research efforts at the firm level and knowledge transfers via social media, which become more important at advanced stages of development. Indeed, systematic econometric studies have yet to verify the hypothesis that investment in schooling serves as an engine of long-term growth (but see Section 6.C for some empirical insights). Nevertheless, we venture to apply this hypothesis here using as a case study the comparative long-term real income growth and educational attainment paths of the US versus the UK and other major European countries over the last century. Our dual hypotheses are the following: first, the US’s economic overtaking of Europe beginning in the late 19th century and its continuing dominance through the 20th century are due largely to the faster and more widespread schooling attainments at the upper-secondary and especially the tertiary level; and second, these differential schooling attainments, whether domestically produced or imported, are ultimately attributable to the higher reward that the US economy has offered to human capital attainments owing to accommodating political ADBI Working Paper 820 Cook and Ehrlich 11 The preceding analysis attributes the gap in educational attainments favoring the US in the 20th century to the interplay of two main forces: first, the feedback effects on the private demand for education that the new industrial economy, economic growth, and personal wealth generated; and second, the impact of the more open economy and society in the US on the returns to human capital formation, whether produced domestically or imported, and thereby on economic growth. As the items in Sections 4.1–4.3 above show, economic growth and affluence lead to a greater demand for education and knowledge and to a greater ability to finance private educational investments by overcoming the inherent imperfections in the capital market. The items in Sections 4.4–4.5 above trace the growth in educational attainments to institutional, political, and economic policies that lower the costs or raise the potential returns to investment, especially in higher education, thus enabling individuals and firms to capture more fully any external effects generated by education. These factors also encourage the immigration of workers with superior skills, education, and entrepreneurial ability. Put differently, the democratic capitalism exercised in the US has contributed to a higher rate of return to individual investment in human capital generally and in tertiary education in particular. While the two groups of factors represent apparently opposite directions of causality regarding the association between human capital formation and economic growth, they are in fact complementary. Greater investment in human capital as a proportion of the total production capacity raises productivity growth, while the demand for human capital investments is partly a by-product of economic growth, and regression analyses aiming to explain productivity growth as a function of educational spending need to account for this. However, these would provide a partial-equilibrium view of economic development. The endogenous growth, general-equilibrium model discussed below sees both human capital formation and productivity growth as endogenous outcomes of the underlying legal and political factors. Moreover, the schooling level of the electorate affects prudent political and economic policies. This view traces the critical causal factors especially to those summarized in Section 4.5. 5. LINKING HUMAN CAPITAL FORMATION WITH ECONOMIC GROWTH 5.1 The Endogenous Growth Hypothesis: Human Capital as the Engine of Growth The literature on endogenous growth attempts to move beyond the neoclassical model of economic growth in two important ways: (a) explaining persistent growth as a result of factors that are endogenous to the economy rather than exogenous, unpredictable technological inventions; and (b) identifying “technology” as human capital or knowledge. According to this view, knowledge breeds greater knowledge. Some new knowledge translates into higher productivity of existing resources (process innovations) or skills (embodied human capital), and some emerges through new goods and machines (product innovation) or new ideas, patents, and manuscripts that account for what we may call “disembodied human capital.” Human capital is ultimately the source of both types of “technology,” and we can therefore consider it as the engine of growth (see Lucas 1988; Becker, Murphy, and Tamara 1990; Ehrlich and Lui 1991). ADBI Working Paper 820 Cook and Ehrlich 12 While major technological innovations may often be the result of discrete and unpredictable breakthroughs in knowledge, deliberate investment in both learning and new knowledge can effect human knowledge formation. The unique property of investment in human capital, however, is that it can lead to persistent growth in knowledge on the assumption that knowledge is the only instrument of production that is not subject to diminishing returns, as John Maurice Clark (1923) put it. It is possible to formalize the idea in a simple way by considering an overlapping generations or dynastic model in which labor is fully employed, human capital is the sole capital asset in the economy, and the law of human capital accumulation of the representative agent or family is: Ht+1 = A (He + Ht) htα (2) Here Ht and Ht+1 denote the human capital stocks of a representative agent in generations t and t + 1, respectively; A represents the technology of knowledge transfer; and (He + Ht) denotes the agent’s production capacity, with He representing a fixed personal endowment of ability or production capacity and Ht representing acquired knowledge at t. The control variable in this production law, 0 ≥ ht ≤ 1, represents the fraction of production capacity that the representative agent in generation t invests in the human capital formation of his or her offspring in generation t + 1. Although the rate of investment in human capital could in principle be subject to diminishing returns, if α in equation (2) is less than 1, we can specify the next generation’s human capital stock as a linear function of the human capital that the current generation, Ht, attains. The implicit argument is that the knowledge and skills that any given generation attains enhance both the creation of new knowledge and the productivity of intergenerational knowledge transfer to the overlapping future generation, thus escaping diminishing returns. Human capital can thus grow perpetually from one generation to another essentially because the level of productive knowledge that the current generation attains serves as an input into the production of knowledge in the succeeding generation. However, whether and the extent to which the latter exceeds the former (or Ht+1 > Ht) critically depend on whether the investment in human capital exceeds a threshold level: according to equation (2), if the optimal investment rate, ht, is not sufficiently high, the knowledge that generation t + 1 attains will be stuck at the level of generation t, Ht, producing a stagnant equilibrium level of output. In a decentralized market economy and a well-functioning free-market system, individuals and families, as well as the level of public spending that they demand from their local and federal government, affect investment in human capital directly. This is particularly important in the case of investment in higher education (a major component of (ht), in which the direct and opportunity costs of investment are high and the access is limited for credit-constrained individuals and families. Government subsidization generally involves tuition and cost- of-living subsidies but may be in the form of capital endowments (unimproved lands), as in the case of the Morrill Act of 1862 (see Section 6.4). Formally, (He + Ht+1) / (He + Ht) ≡ (1 + gt) = Aht + [He / (He + Ht)] > 1 iff Aht > 1. (3) In a growth equilibrium, with Aht > 1 and t → ∞, the gross rate of growth of knowledge capital (1 + g) would converge on the steady-state level (1 + g*) = Ah*, which would be the same as the rate of growth of the agent’s production capacity or the per capita income growth. ADBI Working Paper 820 Cook and Ehrlich 13 The production of human capital, however, is a necessary but not a sufficient condition for inducing productive economic activity. Implicit in this analysis is the assumption that accumulated human capital contributes to expansion in a desired output (Y) through the aggregate production function that Section 1 introduced and the accommodating role of efficient markets, which assure the allocation of skill and creative knowledge to their most productive uses. The endogenous growth paradigm indicates that, in a steady state of continuous growth, physical capital accumulation, including natural resources and productive land, would adjust to the pace of human capital accumulation, making the latter the economy’s engine of growth. At a given fertility level, continuous human capital formation will then lead to continuous expansion in real output per capita (y). Human capital (H) thus replaces the concept of “technology” (T) in equation (1). The model outlined in the preceding discussion is a closed-economy model. In an open economy, the expansion of output is also conditional on the ability of the economy to retain the human capital that it produces. The US was not the first to take off: the Industrial Revolution began in Europe. However, we can attribute the emergence of the US as an economic superpower to the ability of the US market to provide a high reward for human capital investments and thus both to retain domestically produced human capital and to attract human capital produced abroad (see Ehrlich and Kim [2015] for further analysis of this process). 5.2 The Special Role of Higher Education in Economic Growth The previous analysis also rests on the simplifying assumption that workers are homogeneous. In reality, people are heterogeneous in terms of both their innate ability and their contributing family endowments. A more complete view of endogenous growth and development, based on human capital as an engine of growth, must recognize differences among individuals and families in terms of their capacity both to acquire and to implement knowledge. This is the framework that recent work on income growth and income inequality (Ehrlich and Kim 2007) uses to explain the dynamic pattern of both income growth and income distribution over different stages of economic development. The story is simple: human capital, measured using average schooling attainments, has a direct effect on the skills and productivity of the existing labor force as well as an indirect effect on the emergence of new ideas and thus technological innovations and productivity growth. Those who are in a position to acquire more human capital, especially higher education, because of their personal ability or family inputs, are likely to be the “first movers” when it comes to creating new knowledge or implementing advances in knowledge that technological shocks trigger through innovative entrepreneurship, which contributes directly to the pace of economic growth (see Ehrlich, Li, and Liu 2017). Both schools and the labor market also allow for the socialization of knowledge, whereby the achievements of workers with superior knowledge can spill over to, and be shared by, other workers. These “spillover effects” tie population groups of different human capital attainments together over the development process as well as in a regime of persistent growth and ultimately produce stable income distributions. The existence of spillover effects and imperfections in the capital market also justify governments’ subsidization of education, especially higher education, to maximize social income and welfare. ADBI Working Paper 820 Cook and Ehrlich 14 5.3 The Role of Underlying Factors The endogenous growth models described above are general-equilibrium models. In such models, both human capital accumulation and income growth are “endogenous” choice variables: they attain self-sustaining growth as a consequence of the individual choices about optimal investments that individuals make for themselves and their offspring, motivated by a desire to maximize the return that they obtain on these investments. Individual welfare maximization in a decentralized market system thus leads to continuous, self-sustaining growth for the average person in the economy—a dynamic restatement of Adam Smith’s basic proposition. However, this also means that human capital accumulation and income growth are two sides of the same coin: while production functions (1) and (2) represent a causal relation flowing from per capita human capital formation (H) to per capita income (y), a faster-growing economy can also enhance the returns to investment in and accumulation of human capital. The primary causal factors are the underlying “parameters” that influence both variables: most importantly, the factors enhancing the incentives that individuals and families have to invest in their own and their offspring’s knowledge as well as the ability of the domestic economy to utilize the human capital that it generates or imports through immigration effectively in domestic production (see Ehrlich and Kim 2015). The basic parameters affecting both output and knowledge accumulation are knowledge production and transfer technologies—A and B (T) in equations 2 and 1—and population longevity (see Ehrlich and Lui 1991), which enable those investing in learning and training to recoup the benefits of their investments over a longer lifetime horizon. Equally important, however, are “institutional” factors, such as the “rule of law,” a legal system that protects intellectual and property rights, and a free-enterprise system in which entrepreneurial human capital and competitive market forces rather than bureaucratic intervention determine the wages and rates of return on investment (see Ehrlich, Li, and Liu [2017]). They also include accommodating public educational policies that help to overcome capital market constraints in education financing and internalize spillover effects that basic science generates. These accommodating factors, including government regulations and tax policies, can greatly affect output growth through the way in which they enhance or discourage the incentives to invest in human capital. For example, under a heavily regulated system, let alone a command economy, bureaucracy rather than free markets determines the allocation and remuneration of resources, including education. The Soviet Union invested heavily in basic sciences, which it used largely to promote military might, not necessarily economic might. Its command economy system also fostered investment in “political capital,” promising bureaucratic power to apparatchiks, rather than investment in market-driven productive human capital (see Ehrlich and Lui 1999). A free-market system is better geared to reward human capital of the productive type through the market mechanism and is thus more likely to produce self-generating growth. Free trade and an open economy create greater opportunities for human capital accumulation but also greater challenges. Greater opportunities arise because investment in “disembodied knowledge,” such as new production processes or new products, is subject to scale economies, which make their returns higher in a larger market that is open to free trade. Greater challenges emerge because opportunities to migrate from one region or country to another mean that investment in human capital made in one place may actually wind up benefiting another. Public investment in human capital in Peru or in Ireland before 1986, for example, did not bring about an ADBI Working Paper 820 Cook and Ehrlich 15 economic takeoff and self-sustaining growth partly because graduates of institutions of higher learning sought employment in the US market rather than in their own countries. Nonetheless, this does not refute the thesis that investment in human capital is the key to economic growth. It simply reflects the fact that we cannot expect investment that is not supported by a market system that assures an adequate reward for knowledge to yield its full economic benefits. A final underlying factor is the role of externalities inherent in both the production and the transfer of human capital. Private human capital, unlike physical capital, cannot serve as collateral in financial markets, which limits borrowing opportunities. This justifies a public role in the financing of education at all levels but especially in higher education, where investment is substantial, which enhances the accessibility to such educational opportunities according to talent rather than social class and borrowing constraints. Moreover, since higher education can generate spillover effects on the productivity of less educated workers that are not fully internalized through a private reward system, subsidizing it becomes an especially important role of government. In this context, the establishment of a public higher education system can serve as a means of internalizing the range of externalities to which the preceding paragraph alluded. In Section 6.4, we attempt to test this theory by examining the role of the 1862 Morrill Act in bolstering the pace of human capital formation and accumulation and triggering a significant increase in the rate of economic growth. As our analysis in Section 6.4 indicates, the launching of the land grant public university system may have been a significant factor in explaining the higher rate of economic growth in the US relative to the UK and other major European economies and the ascendance of the US to the status of economic superpower in the 20th century. 6. EVIDENCE LINKING EDUCATION AND PRODUCTIVITY GROWTH 6.1 Evidence from Growth Accounting The estimates of the role of schooling in explaining per worker income variations or growth rely on a “growth accounting methodology,” following the works of Denison (1974) and Solow (1957). The technique ascribes changes in the aggregate economy (GDP per capita) to variations in aggregate measures of capital utilization and labor employment, with the labor employment index weighted by measures of the educational attainments of workers. Claudia Goldin and others estimate that, over the 20th century (actually since 1915), the expansion in the educational index has accounted for nearly a quarter of the 1.62% per year increase in US labor productivity. Hall and Jones (1999) estimate that, in 1988, educational attainments accounted for over 20% of the international variation in labor productivity across different countries. Studies using the growth accounting methodology invariably find a substantial unexplained residual variation in productivity, known as the “Solow residual.” They generally attribute it to “technological growth.” However, much of this residual variation may be ascribed to the indirect role of education in inducing technological advancements, as technology is a derivative of special knowledge or specific human capital. Indeed, this is the crux of the “endogenous growth” literature that identifies human capital as the engine of growth. ADBI Working Paper 820 Cook and Ehrlich 16 6.2 Evidence from Rates of Return to Education Human capital theory and related empirical work establishes well that education is the critical factor explaining the differences in earnings across individuals at a point in time. The human capital earnings-generating function that Jacob Mincer formulated links the logarithm of individual earnings to the number of years of schooling and a quadratic specification of the number of years of job market experience. This specification allows the measurement of the “rate of return to human capital” as the regression coefficient associated with the number of years of schooling. Table 9 (see Appendix A), based on a study by Heckman, Lochner, and Todd (2008), indicates that the real rate of return to schooling thus measured has been stable at over 10% for six decennial years but approached 13% in 1990. More import, by estimating separate regressions for white and black males, this study shows that, over the period 1940–1990, the rates of return to black workers, initially lower than those of white workers, more than caught up with the latter in 1990, indicating that the US labor markets have become more competitive over time and better able to reward human capital regardless of race. The Mincerian linear regression model does not allow for the separate estimation of rates of returns using alternative levels of schooling. By relaxing various linearity restrictions implicit in the Mincer model, however, Heckman, Lochner, and Todd (2008) also estimate the rates of return for primary, secondary, and tertiary levels of schooling. Their results indicate that the rates of return are considerably higher for those actually completing high school and college education than for other levels of schooling.6 Other studies indicate that the rate of return, especially to college education, shoots upwards at times of rapid technological innovation, essentially because people with higher skill levels adapt more quickly to changes in technology. These studies focus on returns to education captured in market earnings. New work in economics indicates that these may greatly understate the individuals’ full returns to education, which are derived from various nonmarket activities as well, such as improved health, longevity, and implicit individual assessments of their own life-saving values. Ehrlich and Yin (2005), for example, estimate that both age-specific life expectancies and implicit private values of life saving are substantially higher for those with tertiary relative to high-school education. 6.3 Linking Investment in Schooling and Per Capita Income Growth Empirical studies linking educational attainments and economic growth do not reach uniform conclusions, partly because of disagreements about the quality of the available schooling data. Barro and Lee’s (1993) study, for example, indicates a positive but weak correlation between the overall schooling data that they assemble and the growth rates. Following Ehrlich and Kim (2007), we attempt here to offer a different perspective on the link between education and growth by stressing the correspondence between investments in education, rather than the level of educational attainments, and long-term growth rates of per capita income. According to our theoretical analysis, the steady-state rates of investment in human capital, which are endogenous outcomes of the underlying demographic, institutional, and public policy 6 International comparisons using Mincer’s model or related techniques are hampered by the absence of comparable data. The existing evidence suggests, however, that the estimated rates of return in the US tend to be higher than those in other highly developed countries (see, e.g., Psacharopoulos and Patrinos 2002). Less developed countries may show unusually high rates of return to schooling during a takeoff period from stagnation to a continuous, self-sustaining growth regime. ADBI Working Paper 820 Cook and Ehrlich 17 variables, are the critical determinant of the corresponding long-term growth rates of both per worker human capital stocks and per capita real output in a growth equilibrium regime. While the reported data on educational outlays are incomplete, we can impute the investment levels from time-series evidence on relatively long-term rates of growth of schooling attainments in different countries. We thus expect a systematic link between the equilibrium values of the average growth rates of schooling attainments per worker (H) and the per capita GDP (GDPPC) over relatively lengthy periods in countries experiencing persistent growth. To test this hypothesis, we first estimate the expected growth rates of per capita GDP, [1 + g(GDPPC)*], and schooling attainments, [1 + g(H)*], which we predict from the underlying country-specific factors through the regression model described below, and then we compute their association using the following log-linear regression specification: Log[1 + g(GDPPC)]* = α + βlog[1 + g(H)]* (3) Specifically, we use Barro and Lee’s (2001) data on average schooling years attained by the population aged 15–65 and Summers and Heston’s (1991) estimates of the real GDPPC as proxies for our endogenous variables, along with data on the explanatory variables listed below, to construct a panel of 57 developing and developed countries over an intermediate-length period of 31 years (1960–1991). We first run fixed-effects regressions relating each of our two endogenous variables to a set of underlying country-specific factors. These include demographic variables (population longevity measures) and public policy variables (the share of government spending in GDP and a measure of the social security tax rate) as well as the chronological time and the interaction terms of these explanatory variables with time. (For an explanation of the role of these explanatory variables, see Ehrlich and Kim 2007.) The fixed-effects specification also accounts for the role of idiosyncratic institutional factors, which are unchanging over the sample period. This method allows us to generate multiple predicted values of g(GDPPC)* and g(H)* in each country over our sample period. We can then estimate equation (3) using an OLS regression model. Variant 1 of the model imposes a common intercept term (α) representing the same technology linking human capital formation to output growth in all countries, whereas variant 2 allows for variation in the latter, using a fixed-effects regression specification.7 The idea behind this experiment follows the basic thesis underlying our endogenous growth model. If human capital is the engine of growth, the equilibrium rates of growth of the two endogenous variables of the model—human capital attainments g(H) and real income g(GDPPC)—should be outcomes of the economy’s institutional and demographic factors, including the degree of government intervention in private economic activity. If these two variables are predicted separately from these underlying 7 The analysis involves the following steps. In step 1, we run fixed-effects regressions of log(GDPPC) or log(H) as a dependent variable on a set of regressors as follows: t, t*log(Pi1), t*log(Pi2), t*log(G), t*log(PEN), log(Pi1), log(Pi2), log(G), and log(PEN), where t is the chronological time in years, PEN is a measure of the social security tax rate, Pi1 and Pi2 are probabilities of the survival of children to adulthood and of adults to old age, respectively, and G is the share of government spending in the GDP. (For details, see Ehrlich and Kim 2007.) In step 2, we compute multiple predicted country-specific growth rates of GDP and H over the entire sample period, g(GDPPC)* and g(H)*, based on the estimated regression coefficients involving t and the interaction terms of the basic explanatory variables with t from step 1. This produces a large scatter of observations for 1 + g(GDPPC)* and 1 + g(H)*, allowing a meaningful estimation of equation (3). In step 3, we then estimate variants 1 and 2 of equation (3) via OLS and fixed-effects regressions. Since the countries in our panel are at varying development stages, in additional regressions, which we skip here for simplicity, we also allow the intercept terms in variants 1 and 2 to drift downwards over time, which our model predicts to occur over the development process. These regressions produce very similar results to those that Table 10 reports and have even greater explanatory power. ADBI Working Paper 820 Cook and Ehrlich 18 country-specific “parameters,” they should be closely related within countries. We present the results in Figure 3 (Appendix B) and Table 10 (Appendix A). Figure 3 shows the noisy scatter of the estimated expected growth rates of the per capita GDP and average schooling attainments within countries. The line through this scatter represents the estimated regression line of variant 1 of equation (3). Table 10 also shows the estimated results of variant (2) of equation (3), which we cannot depict graphically. The results in Table 10 indicate the existence of a statistically significant correlation between the predicted growth rates of per capita schooling attainments and the real income within countries in our panel. These results are experimental and preliminary. More complete measures of human capital formation and productivity growth over longer periods, and more elaborate sensitivity analyses, would be necessary to confirm the findings. 6.4 The Role of the 1862 Morrill Act in Enhancing Human Capital Formation and the Pace of per Capita Income Growth in the US Relative to the UK 6.4.1 The Morrill Act as a Natural Experiment Prior to the 1860s, higher education was mostly a privilege for the offspring of just a tiny fraction of the population in both the UK and the US. Both the UK and the early US universities (Harvard, est. 1636, and Yale, est. 1701, which were modeled after Oxford, est. 1096, and Cambridge, est. 1209) had strong ties to theological organizations. While tuition costs were low in both countries, the university required students (only males) to provide their own living arrangements. These expenses alone excluded all but the wealthiest families from sending their children to university. Similarly, the underlying growth rates of per capita GDP in the US and the UK were comparable (both were approximately 1.2‒1.4% per annum) during the period spanning from 1820 until around 1860. From 1860 onwards, however, the US developed a distinct advantage in the number of both universities per capita and student enrollments, largely due to the Morrill Act of 1862, and the per capita GDP growth rate, which lasted not just through the latter part of the 19th century but throughout much of the 20th century and beyond. In this section, we explore the hypothesis that the Morrill Act of 1862, and the consequent rapid expansion of the land grant public higher education system in the US, exerted an important influence on this development. The push to establish a secular public higher education system in the US started in the early 1850s. In 1853, Rep. Justin Morrill of Vermont introduced an act calling for such a system, but the bill failed due to congressional opposition to increased federal spending. President Buchanan vetoed the second attempt, which Rep. Morrill made in 1857, on similar grounds. In 1862, Morrill was finally successful in passing his “Land Grant Act.” There seems to be a general consensus among historians about three coincident factors that converged in 1862 to help pass the Act. First, the US was already in the first phase of the Civil War, and the Southern “states’ rights” delegation, which was strongly opposed to the act, left Congress. The mission of the land grant university system—which included a focus on agriculture and mechanical arts (engineering)— also included military sciences, which increased the backing of the bill as a means of supporting the Union army. Second, President Lincoln was supportive of the bill. Third, a major motivating factor was the ingenious financing plan that Justin Morrill devised. Instead of direct financial support, Morrill suggested granting US states “scrips” of unimproved federal lands in exchange for building universities. Each Congressmen and ADBI Working Paper 820 Cook and Ehrlich 19 Senator was given 30,000 acres of federal land that could be combined to finance the building effort and school expenses. Within the course of a decade, higher education became vastly more accessible and enrollments soon rose as well. To our knowledge, there was no comparable effort in the UK to pursue the establishment of a public university system until 1900. We assume that the random confluence of these unique factors, along with the strong cultural, legal, and economic similarities of the UK and the US and the comparable trends in per capita income growth in the US and the UK over the period preceding the Act, justifies the use of the Morrill Act as a quasi-natural experiment in access to higher education in the United States relative to the UK. We thus apply a quasi-experimental econometric design to measure the impact of the Morrill Act on human capital formation and per capita income growth in the US compared with the UK. In this context, we consider the Act as the “treatment,” the US as the treatment group, and the UK as the control group. We measure the impact of the 1862 Morrill Act on universities per capita, tertiary enrollments, and alternative measures of long-term GDP growth rate differences between the US and the UK and other control groups using the differencein-differences methodology. 6.4.2 The Difference-in-Differences Regression Models We first estimate the MA’s effect on two indicators of higher education formation per capita: the number of universities founded per capita in the US vs. the number in the UK following the act, limiting the time horizon to avoid potential confounding effects from the “redbrick” university expansion in the UK.8 We augment the regression equation that we use to measure the treatment effect with a set of correlates designed to help make the differences between the treatment and the control group “as random as possible.” The data sources that we use to measure all the variables are discussed in Appendix A. Since GDP data are available only for decennial years until 1870, we also use decennial data to measure the corresponding higher education indicators. The DiD estimator δ in the regression equations below represents the interaction between the post-Morrill Act period dummy and the US dummy variable. The land grant dummy variable takes the value of one after the ratification of the Morrill Act on 2 July 1862 and zero before that time. Using the data on university founding across countries that Foldvari (2014) compiled, we test the impact of the first Morrill Act on the “number of universities per capita,” UNIVPC, using reduced-form regression equation (4): () T it t t i i t USA it it it UNIVPC Y country MA country αβ γ δ ξ ε = + + + × +Χ + (4) In equation (4), βt represents year (Yt) fixed effects, γi are country fixed effects, and δ measures the impact of the Morrill Act policy change on the stock of universities per capita in the US relative to the UK. We add a set of covariates, denoted by the vector XitT, to control for the effects of possible non-random differences between the US and the UK (or alternative control groups that we use in some of our DiD regressions), which may influence the Act’s treatment effect. The variable that we include as a covariate in Tables 11 and 11a is the average real GDP per capita.9 8 Beginning in 1900, the redbrick movement was a civic expansion of British higher education, similar in many ways to the earlier 1862 Morrill Act expansion. 9 We also test government spending as a fraction of the GDP as an additional covariate to control for the possible role of government expenditure beyond the land grant endowments in promoting the ADBI Working Paper 820 Cook and Ehrlich 20 We use a similar specification to test the impact of the Act on tertiary enrollment rates (per 100 18‒24-year-old population members), Tertrate, using data from the National Center of Education Statistics (NCES), the US Census, and Tamura et al. (2015): () T it t t i i t USA it it it Tertrate Y country MA country αβ γ δ ξ ε = + + + × +Χ + (5) In equation (5), βt now represents the decennial year fixed effect (1820, 1830 ...) and δ captures the impact of the Morrill Act event on the difference in tertiary enrollment rates between the US and the UK.10 We then complete the analysis by estimating the MA’s impact on alternative specifications of the “long-term” real GDP per capita rate of growth in the US relative to that in the UK (and other control variables), using the same reduced-form regression model: , log(1 ) ( ) iy T T y t t i i t USA it it it g Y country MA country αβ γ δ ξ ε + = + + + × +Χ + (6) In equation (6), βt is again the decennial year fixed effect (1820, 1830 ...), γi are country fixed effects, and δ measures the impact of the Morrill Act policy on the per capita income growth in the United States over a T-year forward-looking time horizon. We add the vector of covariates, denoted XitT, to this equation as well. More specifically, we measure the dependent variable as the natural logarithm of the average gross rate of real per capita income in country i in year t, , log(1 ) iy Ty g+ ,11 over alternative “forward-looking” time intervals, T (for T = 10, 20, 30, 50, or 100 years).12 Appendix B explains the method of calculating these forward-looking rates. The coefficient of interest, δ, now measures the impact of the Morrill Act on the natural log of the growth rate of real per capita income in the United States relative to that in the UK across the pre- and post-Morrill Act periods. We present the regression results concerning the impact of the Morrill Act on all three outcome indicators in Table 11 (see Appendix A).13 6.4.3 The Findings The results in Table 11 indicate that, after 1862, the United States experienced rapid expansion of its university system following the ratification of the Morrill Act. Compared with the UK, the US added 0.78 universities per 1000 population during the post-Morrill period. The estimated Morrill Act effect on the number of universities created per capita supports our choice of the Morrill Act as the exogenous policy shift (and the use of the quasi-experimental framework). subsequent expansion of the US public university system. While the inclusion of this added correlate does not affect markedly the estimated magnitude of δ in Table 11, the data limitations causing a reduction in the sample size lower its associated t-statistic substantially. 10 The years reported vary by country, which creates sample size issues when comparing different countries. To resolve this problem without affecting the sample size, we round the years to the nearest decennial. 11 The log transformation is useful for “drawing down” the long tail of the distribution (normalizing the distribution) of the growth rates. 12 We calculate T-year average growth rates in the GDP per capita, which Barro and Sala-i-Martin (1992) calculate as g = (1/T)[log(yt+T)/yt)], where y = GDP/N. 13 We correct the per capita GDP growth rate dependent variable in equation (6) for potential “forwardlooking” biases, as Appendix B, Section II, describes. ADBI Working Paper 820 Cook and Ehrlich 27 APPENDIX A: TABLES Table A: Comparison of Real GDP per Capita for the Top 26 Countries (US Dollars Converted Using Purchasing Power Parity)* Country Per Capita GDP Estimate Year Country Per Capita GDP Estimate Year Singapore 62,400 2013 Denmark 37,800 2013 Norway 55,400 2013 Belgium 37,800 2013 United States 52,800 2013 United Kingdom 37,300 2013 Hong Kong, China 52,700 2013 Japan 37,100 2013 Switzerland 46,000 2013 Finland 35,900 2013 Canada 43,100 2013 France 35,700 2013 Australia 43,000 2013 Israel 34,900 2013 Austria 42,600 2013 Republic of Korea 33,200 2013 Netherlands 41,400 2013 Saudi Arabia 31,300 2013 Ireland 41,300 2013 Spain 30,100 2013 Sweden 40,900 2013 UAE 29,900 2013 Taipei,China 39,600 2013 Italy 29,600 2013 Germany 39,500 2013 Czech Republic 27,200 2013 * The table excludes countries with a population of fewer than 5 million in 2013. Sources: Central Intelligence Agency. The World Factbook. http://www.cia.gov/library/publication/the-world-factbook Table 1: Average Years of Formal Educational Experience of the Population Aged 15–64 in 1913 and 1989 (Maddison’s Data) 1913 Country Total (Rank) Primary (Rank) Secondary (Rank) Higher (Rank) France 6.18 (4) 4.31 (5) 1.77 (4) 0.10 (3) Germany 6.94 (2) 3.50 (6) 3.35 (1) 0.09 (4) Japan 5.10 (6) 4.50 (4) 0.56 (6) 0.04 (6) Netherlands 6.05 (5) 5.30 (1) 0.64 (5) 0.11 (2) United Kingdom 7.28 (1) 5.30 (1) 1.90 (2) 0.08 (5) United States 6.93 (3) 4.90 (3) 1.83 (3) 0.20 (1) 1989 Country Total (Rank) Primary (Rank) Secondary (Rank) Higher (Rank) France 11.61 (3) 5.00 (5) 5.29 (2) 1.32 (2) Germany 9.58 (6) 4.00 (6) 5.20 (3) 0.38 (6) Japan 11.66 (2) 6.00 (1) 4.95 (4) 0.71 (3) Netherlands 10.51 (5) 6.00 (1) 3.82 (6) 0.69 (4) United Kingdom 11.28 (4) 6.00 (1) 4.75 (5) 0.53 (5) United States 13.39 (1) 6.00 (1) 5.72 (1) 1.67 (1) Source: Data from Maddison (1991, 64). ADBI Working Paper 820 Cook and Ehrlich 28 Table 2: Percentage of the Population that has attained at least Tertiary Education Type A, by Age Group, 1998 and 2011 Country 1998 2011 25–64 25–34 35–44 45–54 55–64 25–64 25–34 35–44 45–54 55–64 France 11 15 10 10 6 18 27 21 13 12 Germany 14 14 16 15 10 16 18 18 15 15 Italy 9 9 11 9 5 15 21 16 11 11 Norway 24 27 25 22 17 36 46 39 31 26 Spain 14 21 16 11 6 22 27 25 20 15 United Kingdom 15 17 17 15 11 30 39 32 24 22 United States 27 27 26 29 22 32 33 34 30 31 29 OECD 14 16 15 13 9 23 30 25 19 17 Country mean * The data from Japan correspond to 2003. Sources: OECD. 2000. Education at a Glance 2000, 36 (Table A2.2b) and 35 (Table A2.2). http://www.oecd.org/ dataoecd/22/35/35282639.xls (Tables A.1.3a and A.1.2a). OECD. 2013. Education at a Glance 2013. http://dx.doi.org/10.1787/888932847982 (Table T_A1.2a). Table 3: Distribution of the Population that has attained at least Upper-Secondary Education, by Age Group, 1998 and 2011 Country 1998 2011 25–64 25–34 35–44 45–54 55–64 25–64 25–34 35–44 45–54 55–64 France 61 75 63 56 41 72 83 78 68 58 Germany 84 88 87 84 76 86 87 87 87 84 Italy 41 55 50 35 19 56 71 60 52 40 Japan* 80 93 91 77 57 84 94 94 82 65 Spain 33 53 38 23 12 54 65 61 50 34 United Kingdom 60 63 62 58 53 77 84 80 75 67 United States 86 88 88 87 80 89 89 89 89 90 29 OECD 61 72 65 57 44 75 82 78 73 64 Country mean * The data from Japan correspond to 2003. Sources: OECD. 2000. Education at a Glance 2000, 36 (Table A2.2b) and 35 (Table A2.2). http://www.oecd.org/ dataoecd/22/35/35282639.xls (Tables A.1.3a and A.1.2a). OECD. 2013. Education at a Glance 2013. http://dx.doi.org/10.1787/888932847982 (Table T_A1.2a). ADBI Working Paper 820 Cook and Ehrlich 29 Table 4: Expected Years of Tertiary Education for All 17-Year-Olds (1998) Country Full and Part Time Rank Full Time Only Rank Australia 2.3 5 1.4 16 Austria 1.8 14 1.8 8 Belgium 1.3 21 1.2 20 Canada 1.9 10 1.4 16 Czech Republic 0.9 24 0.8 24 Denmark 1.3 21 1.3 19 Finland 2.9 1 2.9 1 France 1.9 11 1.9 5 Germany 1.7 15 1.7 11 Greece 1.7 15 1.7 11 Hungary 1.6 19 0.9 23 Iceland 1.6 19 1.6 14 Ireland N/A ~ N/A ~ Italy 2.2 7 2.2 4 Japan N/A ~ N/A ~ Republic of Korea 1.9 11 1.9 5 Luxembourg N/A ~ N/A ~ Mexico 0.8 25 0.8 24 Netherlands 2.2 7 1.9 5 New Zealand 2.1 9 1.5 15 Norway 2.7 2 2.4 2 Poland 1.9 11 1.0 22 Portugal 1.7 15 1.7 11 Spain 2.5 4 2.3 3 Sweden 2.3 5 1.8 8 Switzerland 1.1 23 1.1 21 Turkey 0.8 26 0.8 24 United Kingdom 1.7 15 1.4 16 United States 2.7 2 1.8 8 OECD average 1.8 1.6 Source: OECD. 2000. Education at a Glance 2000, 158, Table C3.2. ADBI Working Paper 820 Cook and Ehrlich 30 Table 5: Expenditure on Educational Institutions as a Percentage of the GDP for All Levels of Education by Source of Funds (1990, 2002, and 2010) 2010 2002 1990 Country Public Private Total Public Private Total Public Private Total Australia 4.6 1.5 6.1 4.4 1.5 6.0 4.2 0.8 5.0 Austria 5.6 0.2 5.8 5.4 0.3 5.7 N/A N/A N/A Belgium 6.4 0.2 6.6 6.1 0.3 6.4 N/A N/A N/A Canada 5.0 1.6 6.6 N/A N/A N/A N/A N/A N/A Czech Republic 4.1 0.6 4.7 4.2 0.2 4.4 N/A N/A N/A Denmark 7.6 0.4 8.0 6.8 0.3 7.1 N/A N/A N/A Finland 6.4 0.1 6.5 5.9 0.1 6.0 N/A N/A N/A France 5.8 0.5 6.3 5.7 0.4 6.1 5.1 0.5 5.7 Germany m m m 4.4 0.9 5.3 N/A N/A N/A Greece m m m 3.9 0.2 4.1 N/A N/A N/A Hungary 4.6 m m 5 0.6 5.6 N/A N/A N/A Iceland 7.0 0.7 7.7 6.8 0.6 7.4 N/A N/A N/A Ireland 6.0 0.5 6.4 4.1 0.3 4.4 N/A N/A N/A Italy 4.3 0.4 4.7 4.6 0.3 4.9 N/A N/A N/A Japan 3.6 1.5 5.1 3.5 1.2 4.7 N/A N/A N/A Republic of Korea 4.8 2.8 7.6 4.2 2.9 7.1 N/A N/A N/A Luxembourg m m m N/A N/A N/A N/A N/A N/A Mexico 5.1 1.1 6.2 5.1 1.1 6.3 N/A N/A N/A Netherlands 5.4 0.9 6.3 4.6 0.5 5.1 N/A N/A N/A New Zealand 6.0 1.3 7.3 5.6 1.2 6.8 N/A N/A N/A Norway 7.5 m m 6.7 0.3 6.9 8.1 N/A N/A Poland 5.0 0.8 5.8 5.5 0.7 6.1 N/A N/A N/A Portugal 5.4 0.4 5.8 5.7 0.1 5.8 N/A N/A N/A Slovak Republic 4.0 0.6 4.6 4.0 0.2 4.2 4.8 0.3 5.1 Spain 4.8 0.8 5.6 4.3 0.5 4.9 4.4 0.7 5.1 Sweden 6.3 0.2 6.5 6.7 0.2 6.9 5.1 N/A 5.1 Switzerland 5.2 m m 5.7 0.5 6.2 N/A N/A N/A Turkey m m m 3.4 0.4 3.8 2.8 N/A 2.8 United Kingdom 5.9 0.6 6.5 5.0 0.9 5.9 4.2 0.1 4.3 United States 5.1 2.2 7.3 5.3 1.9 7.2 4.9 2.2 7.1 Country mean 5.4 0.9 6.3 5.1 0.7 5.8 ~ ~ ~ OECD total 5.0 1.5 6.5 4.9 1.2 6.1 ~ ~ ~ Source: OECD. Education at a Glance 2005 and 2013, Indicator B2: Expenditure on Educational Institutions Relative to Gross Domestic Product. http://www.oecd.org/dataoecd/2/11/35286380.xls (Table B2.1a). http://dx.doi.org/10.1787/ 888932849616 (Table B2.3). ADBI Working Paper 820 Cook and Ehrlich 31 Table 6: Annual Expenditures on Educational Institutions per Student (US Dollars Converted Using PPP) by Levels of Education Based on Full-Time Equivalents (2010) Country Primary All Secondary Tertiary Type A All Tertiary United States 11,193 12,464 NA 25,576 United Kingdom 9,369 10,452 NA 15,862 France 6,622 10,877 15,997 15,067 Germany NA NA NA NA Italy 8,296 8,607 9,576 9,580 Spain 7,291 9,608 14,072 13,373 Austria 10,244 12,551 15,101 15,007 Belgium 8,852 11,004 NA 15,179 Denmark 10,935 11,747 NA 18,977 Finland 7,624 9,162 16,714 16,714 Netherlands 7,954 11,838 17,172 17,161 Norway 12,255 13,852 NA 18,512 Sweden 9,987 10,185 20,750 19,562 Switzerland 11,513 14,972 23,457 21,893 Japan 8,353 9,957 17,544 16,015 Republic of Korea 6,601 8,060 11,271 9,972 OECD mean 7,974 9,014 ~ 13,528 Source: OECD. 2013. Education at a Glance 2013. http://dx.doi.org/10.1787/888932849350 (Table B1.1a). ADBI Working Paper 820 Cook and Ehrlich 32 Table 7: Expenditure per Student (Private and Public) Relative to the GDP per Capita by Level of Education Based on Full-Time Equivalents (2010) All Secondary Education All Tertiary Education Australia 25 37 Austria 31 37 Belgium 29 40 Canada N/A 56 Czech Republic 26 30 Denmark 29 47 Finland 25 46 France 32 44 Germany N/A N/A Greece N/A N/A Hungary 22 42 Iceland 22 25 Ireland 28 39 Italy 27 30 Japan 28 45 Republic of Korea 28 35 Luxembourg 21 m Mexico 17 52 Netherlands 28 41 New Zealand 28 35 Norway 31 41 Poland 27 44 Portugal 35 41 Slovak Republic 21 30 Spain 30 42 Sweden 26 50 Switzerland 31 45 Turkey 16 N/A United Kingdom 30 45 United States 27 55 Country mean 27 41 Source: OECD. 2013. Education at Glance 2013, Indicator B1: Educational Expenditure per Student. http://www.oecd-ilibrary.org/education/education-at-a-glance-2013_eag-2013- en (Table B1.4). ADBI Working Paper 820 Cook and Ehrlich 33 Table 8: Cut Scores of 15-Year-Old Students on the PISA Mathematics Literacy Scale at Selected Percentiles and Percentile (2012) Country Percentile 10th 25th 50th 75th 90th 90th to 100th United States 368 418 477 543 600 233 France 365 429 497 565 621 256 Germany 385 447 516 583 637 252 Italy 366 421 485 550 607 241 Spain 370 424 486 546 597 228 Austria 384 440 506 572 624 240 Belgium 378 444 518 589 646 268 Denmark 393 444 501 556 607 214 Finland 409 463 520 577 629 219 Netherlands 397 457 529 591 638 242 Norway 373 428 490 552 604 231 Sweden 360 415 478 543 596 236 Switzerland 408 466 534 597 651 243 Japan 415 473 538 603 657 242 Republic of Korea 425 486 557 624 679 254 OECD average 375 430 494 558 614 239 Shanghai, People’s Republic of China 475 546 622 685 737 262 Singapore 432 501 579 650 707 275 Hong Kong, China 430 499 569 629 679 249 Note: This table shows the threshold (or cut) scores for the following: (a) 10th percentile—the bottom 10% of students; (b) 25th percentile—the bottom 25% of students; (c) 50th percentile—the median (half the students scored below the cut score and half scored above it); (d) 75th percentile—the top 25% of students; and 90th percentile—the top 10% of students. The percentile ranges are specific to each education system’s distribution of scores, enabling users to compare cut scores across education systems. The table orders education systems by cut score gap. The OECD average is the average of the national averages of the OECD member countries, with each country weighted equally. The table reports the scores on a scale from 0 to 1,000. Italics indicate non-OECD countries and education systems. The results for Connecticut, Florida, and Massachusetts are for public school students only. Source: The National Center for Education Statistics (NCES). http://nces.ed.gov/surveys/pisa/pisa2012/pisa2012 highlights_3b.asp; Organization for Economic Cooperation and Development (OECD). 2012. Program for International Student Assessment (PISA). ADBI Working Paper 820 Cook and Ehrlich 34 Table 9: Estimated Coefficients from Mincer Log-Earnings Regressions for Males White Males Black Males Coefficient Std Error Coefficient Std Error 1940 Intercept 4.4771 0.0096 4.6711 0.0298 Education 0.1250 0.0007 0.0871 0.0022 Experience 0.0904 0.0005 0.0646 0.0018 Experience squared –0.0013 0.0000 –0.0009 0.0000 1950 Intercept 5.3120 0.0132 5.0716 0.0409 Education 0.1058 0.0009 0.0998 0.0030 Experience 0.1074 0.0006 0.0933 0.0023 Experience squared –0.0017 0.0000 –0.0014 0.0000 1960 Intercept 5.6478 0.0066 5.4107 0.0220 Education 0.1152 0.0005 0.1034 0.0016 Experience 0.1156 0.0003 0.1035 0.0011 Experience squared –0.0018 0.0000 –0.0016 0.0000 1970 Intercept 5.9113 0.0045 5.8938 0.0155 Education 0.1179 0.0003 0.1100 0.0012 Experience 0.1323 0.0002 0.1074 0.0007 Experience squared –0.0022 0.0000 –0.0016 0.0000 1980 Intercept 6.8913 0.0030 6.4448 0.0120 Education 0.1023 0.0002 0.1176 0.0009 Experience 0.1255 0.0001 0.1075 0.0005 Experience squared –0.0022 00.000 –0.0016 0.0000 1990 Intercept 6.8912 0.0034 6.3474 0.0144 Education 0.1292 0.0002 0.1524 0.0011 Experience 0.1301 0.0001 0.1109 0.0006 Experience squared –0.0023 0.0000 –0.0017 0.0000 Source: Heckman, Lochner, and Todd (2008). Table 10: Correlating Predicted Growth Rates in Per Capita GDP and Average School Years of the Adult Population (Based on Ehrlich and Kim 2007) Intercept (α) Slope (β) t-value (β) Adjusted R2 Variant 1* 0.00567 1.67458 21.23 0.3036 Variant 2** ** 1.25854 11.40 0.3682 Number of observations = 1,032. * OLS regression estimates of equation (3). ** OLS fixed-effects regression estimates of equation (3) allowing for country-specific intercepts, which are not reported in this table. Econometric procedure: see the text and footnote 6. ADBI Working Paper 820 Cook and Ehrlich 35 Table 11: The 1862 Morrill Act Treatment Effect on Higher Education Formation and per Capita GDP Growth in the US Relative to the UK, 1820–1900 (the UK Serving as the Control Group) Dependent Variable Estimated Morrill Effect (SE)† [DF] Universities per capita 0.00078*** (0.00005) [96] Tertiary enrollment rate (per 100) 0.838*** (0.167) [6] 10y per capita GDP growth rate 0.0154*** (0.0033) [95] 20y per capita GDP growth rate 0.0101*** (0.0015) [94] 30y per capita GDP growth rate 0.0098*** (0.0008) [93] 50y per capita GDP growth rate 0.0058*** (0.0011) [96] 100y per capita GDP growth rate 0.0014*** (0.0004) [96] Notes: *** = 99% significance, ** = 95% significance, * = 90% significance. The table reports HAC robust standard errors in parentheses. The treatment group is the United States. All the regressions include the covariate initial real income as a proxy for capital per unit of labor. We censor the growth rate regressions to avoid the contamination bias discussed in the text. Accordingly, the 10-year specification omits one observation; the 20-year specification omits two observations; and the 30-year specification omits three observations. † The degrees of freedom (DF) reflect the sample size that we use in the individual DiD estimation. The DFs for the tertiary enrollment rate are lower due to the data limitations in Tamura et al. (2015). Source: The data on the number of universities founded across countries come from Foldvari (2014). The data on tertiary enrollment rates are from Tamura et al. (2015.) The tertiary enrollment rate is a fraction of 100 population members. We take the data on per capita real income from Maddison (2013). Table 11a: The 1862 Morrill Act Treatment Effect on Higher Education Formation and per Capita GDP Growth in the US Relative to the UK, 1820–1900 (Alternative Control Groups) Dep. Variable Est. Morrill Effect – WBO Control Coefficient (SE)/[DF] Est. Morrill Effect – EU5 Control Coefficient (SE)/[DF] Est. Morrill Effect – All Control Coefficient (SE)/[DF] Universities per capita 0.00053** (0.00024)/[246] 0.00087*** (0.00007)/[320] 0.00021 (0.00017)/[550] Tertiary enrollment rate (per 100) 1.022*** (0.253)/[30] 1.050*** (0.263) [31] 1.060*** (0.246) [63] 10y per capita GDP growth rate 0.0143*** (0.0034)/[243] 0.0188*** (0.0032)/[315] 0.0179*** (0.0030)/[543] 20y per capita GDP growth rate 0.0117*** (0.0025)/[244] 0.0164*** (0.0026)/[315] 0.0137*** (0.0016)/[545] 30y per capita GDP growth rate 0.0118*** (0.0027)/[243] 0.0120*** (0.0016)/[317] 0.0123*** (0.0012)/[547] 50y per capita GDP growth rate 0.0059*** (0.0011)/[246] 0.0051*** (0.0018)/[320] 0.0065*** (0.0016)/[550] 100y per capita GDP growth rate 0.0018*** (0.0005)/[246] † † Notes: *** = 99% significance, ** = 95% significance, * = 90% significance. The table reports HAC robust standard errors in parentheses. The treatment group is the United States. The WBO control uses (UK, AN, AUS, NZL) as the control group, and the EU5 combines (FRA, SPA, UK, ITA, GER) as the control group. The all-country control group is composed of: (AUS, CAN, FRA, GER, ITA, NED, NZL, SPA, UK). All the regressions include the covariate initial real income as a proxy for capital per unit of labor. We censor the growth rate regressions to avoid the contamination bias discussed in the text. Accordingly, the 10-year specification omits one observation; the 20-year specification omits two observations; and the 30-year specification omits three observations. † Indicates unreported results in the 100-year regression. This is because we observe significant deviations in the pre- Morrill period between the US and the UK, which violate the criteria required for DiD analysis. The degrees of freedom (DF) reflect the sample size that we use in the individual DiD estimation. The DFs for the tertiary enrollment rates are lower due to the data limitations in Tamura et al. (2015). Source: The data on the number of universities founded across countries come from Foldvari (2014). The data on tertiary enrollment rates are from Tamura et al. (2015). The tertiary enrollment rate is a fraction of 100 population members. We take the data on per capita real income from Maddison (2013). ADBI Working Paper 820 Cook and Ehrlich 36 Table 12: Relative Percentage Differences in Educational Attainments (US = 100) by Level and Age Group (1998 and 2011) Attaining at Least Tertiary Education Type A: 1998 2011 Country 25–64 25–34 35–44 45–54 55–64 25–64 25–34 35–44 45–54 55–64 US 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 France 40.7 55.6 38.5 34.5 27.3 55.9 81.8 61.4 43.0 38.2 Germany 51.9 51.9 61.5 51.7 45.5 49.7 54.5 52.7 49.6 47.7 Italy 33.3 33.3 42.3 31.0 22.7 45.3 62.8 47.7 36.4 33.6 Spain 51.9 77.8 61.5 37.9 27.3 69.2 81.0 74.6 65.3 46.2 United Kingdom 55.6 63.0 65.4 51.7 50.0 92.0 119.1 93.7 79.9 70.1 EU5* 46.7 56.3 53.8 41.4 34.5 62.4 79.8 66.0 54.9 47.2 29 OECD 51.9 59.3 57.7 44.8 40.9 70.8 89.4 72.3 64.1 52.9 Attaining at Least Upper-Secondary Education: 1998 2011 Country 25–64 25–34 35–44 45–54 55–64 25–64 25–34 35–44 45–54 55–64 United States 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 France 70.9 85.2 71.6 64.4 51.3 80.7 93.2 87.6 76.1 64.7 Germany 97.7 100.0 98.9 96.6 95.0 96.3 97.7 97.7 97.4 93.7 Italy 47.7 62.5 56.8 40.2 23.8 62.7 79.7 67.4 58.2 44.6 Spain 38.4 60.2 43.2 26.4 15.0 60.5 72.8 68.6 55.9 37.7 United Kingdom 69.8 71.6 70.5 66.7 66.3 86.1 94.7 90.1 83.6 74.4 EU5* 64.9 75.9 68.2 58.9 50.3 77.3 87.6 82.3 74.3 63.0 29 OECD 70.9 81.8 73.9 65.5 55.0 83.8 92.4 88.2 81.3 71.1 * EU5: The simple average of the normalized data for France, Germany, Italy, Spain, and the United Kingdom. Source: Own computations based on Tables 2 and 3.