The Relevance of Science in Development

1 Economic Behavior, Game Theory, and Technology in Emerging Markets. Editors: Christiansen B, Basilgan M; IGI Global Chapter1; 1-17 The Relevance of Science in Development Scientific development favors economic prosperity, but not necessarily through its effect on technological knowledge, in middle income countries. Klaus Jaffe Universidad Simón Bolívar, Caracas, Venezuela ABSTRACT Scientific knowledge and technical expertise promote the wealth of nations. The traditional view is that science allows the expansion of technology which in turn promotes economic development. Here we show that scientific productivity of a country correlates stronger with Gross National Income per capita than its technological sophistication; that science is important for economic growth among developed economies, whereas technical complexity is more important for the economic development of poorer countries; that scientific productivity of countries correlates stronger with present and future wealth than indices reflecting its financial, social, economic or technological sophistication; and that middle income countries with higher relative productivity in basic sciences such as physics and chemistry have the highest economic growth in the following five years compared to countries with a higher relative productivity in applied sciences. No simple direct causal relationship between scientific productivity and economic growth could be detected. The results are best explained by assuming that science favors economic development by providing society with a more rational atmosphere, allowing the implementation of sound policies and institutions, and/or that rational societies with succesful economic policies are also the ones giving priority to basic natural sciences.

2 INTRODUCTION Knowledge and wealth have been recognized to be related since ancient times (Bacon, 1620; Condorcet, 1794; Spencer, 1875; Marshall, 1890). Napoleon used to say that there cannot be a great nation without great mathematics. Yet how this relationship works in the modern world is still a sensitive political issue (Salter & Martin, 2001; Nelson, 1959; King, 2004; Royal Society, 2011). The traditional view is that scientific development is required for technological expansion which affects the productive apparatus of a country explaining a high correlation between scientific research and technological advances (Teitel, 1994; Wang, 2007 for example) explaining why scientific development and the wealth of nations are closely linked (Sachs, 2005). A significant recent contribution to the debate was made by Hidalgo et al (Hidalgo et al. 2007, Hidalgo & Hausmann 2009, Hausmann et al. 2011) who proposed a novel Economic Complexity Index (ECI) to account for knowledge embedded in society that produces wealth. In their words Modern societies can amass large amounts of productive knowledge because they distribute bits and pieces of it among its many members. But to make use of it, this knowledge has to be put back together through organizations and markets. Thus, individual specialization begets diversity at the national and global level. Our most prosperous modern societies are wiser, not because their citizens are individually brilliant, but because these societies hold a diversity of knowhow and because they are able to recombine it to create a larger variety of smarter and better products. This ECI is built based on the relative amount of exports of different products for each country and on an index as to the complexity or difficulty in producing each product. Although ECI reflects many different features of an economy, the authors (Hausmann et al. 2011) maintain that it mainly reflects the

3 composition of a country s productive output and its structures, which in turn is a strong reflection of the countries combined productive knowledge. On the other hand, scientific development and the wealth of nations have been postulated to be closely linked. Scientific productivity showed to be a much better predictor of economic wealth of a nation than all educational variables tracked by the United Nations Development Program and the World Bank (Jaffe 2005, Jaffe et al 2013a,b). The research output of a country showed to be a much better predictor of future economic growth than technological complexity of the country as measured with ECI. Scientific development was shown to correlate with tolerance and openness of a society, reflecting the fact that attitudes favoring science are related to valuation of empirical facts over personal convictions, which lay at the base of modern scientific progress (Jaffe 2009). Here we want to answer some questions raised by the above described research: Do certain areas of science promote economic development more than others? Are more applied sciences better in advancing economic development than more general basic sciences? Does science promotes economic development through its effect on technological development, or does science act through other more general mechanisms?

4 Box 1: Definition of abbreviations of variables analyzed Economic wealth: GDPc: Gross Domestic Product per capita at constant US$ 2000 (World Bank) GNIc: Gross National Income per capita (World Bank Atlas method) Growth: Δ: difference in the respective variable between 1998 and 2008 %: difference expressed in percent between 1998 and 2008 Knowledge: ECI: Economic Complexity Index (Hausmann et al. 2011) SP: Scientific productivity irrespective of its measure as SPW or SPS SPc: Scientific productivity per capita SPWc: Scientific Productivity measured as the number of publications in scientific and engineering journals published in physics, biology, chemistry, mathematics, clinical medicine, biomedical research, engineering and technology and earth and space sciences, per capita. (World Bank) SPSc: Scientific Productivity measured as the number of academic documents per capita published, compiled by Scopus (SCImago. 2007)

5 COMPARING DIFFERENT INDICES OF KNOWLEDGE AND SOCIO- TECHNICAL DEVELOPMENT AS TO ITS CLOSENESS TO GDPc A large number of indices have been developed in order to measure different aspects of the social, cultural, political and economic aspects of a nation (see World Bank Data repository for example). Despite this large number of indices, the Gross Domestic Product per capita (GDPc) is still the most widely used measure to estimate overall development of a country. This is partly due to the fact that GDPc is a measure based on quantitative data, that it is available for most countries in the world for long periods of time. But how GDPc relates to these other measures? Do all these indices measure different things or are there groups of indices track similar phenomena? A cluster analysis of these indices might answer these questions. In Figure 1 we present a Joining Tree Cluster Analysis (Statistica 7) comparing GDPc with a list of widely used indices using the weighted pair-group average and its Euclidean distance to compute a matrix from this distances. The tree was then drawn from the data in the matrix. This cluster analysis revealed that scientific productivity is a much better predictor of economic wealth and Human Development of a nation than other variables tracked by a number of commonly used indices proposed worldwide. Figure 1 show that the number of publications per capita of a country (Publication) is the index closest the GDP per capita and to the Human Development Index (HDI) of the country. That is, Publication correlates much stronger with the wealth per capita of a nation than any of the other indices tested. This cluster is joined by indices such as HDI, Doing Buisiness, Economic Freedom and Transparency. Other clusters group around Democracy and Press Freedom, and some indices seem to be disconnected from the rests, such as Happy

6 Planet, Scientific Complexity and a Web Index. More details of this study are published in Jaffe et al. (2013b) Figure 1: Cluster analysis of the country ranks of common econometric indices (sources for the indices are detailed in the appendix). COMPARING THE FORCE OF SCIENTIFIC VS. TECHNOLOGICAL KNOWLEDGE IN PREDICTING FUTURE ECONOMIC GROWTH Technology is clearly related to economic growth and so is scientific development. When we compare indices measuring each of these factors we found that scientific productivity is a much better predictor of economic wealth than technological knowledge reflected as ECI. For example, in order to estimate the power of ECI and SPSc, assessed in a given year, in predicting economic growth during the following 10 years, we compared two different general linear models (Table 1), using variables

7 defined in Box 1: In Model 1, build to estimate the effect of ECI on future economic growth, the percent change increase in GNIc (%ΔGNIc) during 1998-2008 is regressed against loggnic and ECI, both assessed in 1998. In Model 2, built to estimate the effect of SPSc on future economic growth, %ΔGNIc during 1998-2008 is regressed against loggnic in 1998 and logspsc in 1998. Results (Table 1a) show that both models predict future economic growth with high probability, but Model 2 was better than Model 1 in rejecting the null hypothesis - that no correlation between economic growth and the selected variables exists. In Model 3 we included logspsc, ECI and loggnic assessed in 1998 in the general regression to predict %ΔGNIc during 1998-2008. The effect sizes and power of the variables is given in Table 1b. The results in Table 1 suggest that adding ECI to log SPSc does not increase the predictive value of Model 2, but rather decreases it slightly. Using the Akaike information criterion (AIC) to compare model 1, 2 and 3 we obtained lower AIC values for Model 2 compared to Model 1, and very similar AIC values for Model 2 and 3. These results indicate that Model 1 was 0.36 times as likely as Model 2 to minimize information loss, and that Model 2 was 0.7 times as likely as Model 3 to minimize information loss (Akaike 1974). Thus, ECI estimated for 1998 carried much less information compared to SPSc in predicting economic growth in the decade 1998-2008.

8 Table 1a: Akaike information criterion (AIC) and regression coefficients of the three models showing R² (regression coefficient), SS (sum of squares), MS (median value), df (degrees of freedom), F (Fisher statistics), and p (probability of null hypothesis) %ΔGNIc 1998-2008 AIC Multiple - R² Adjusted - R² SS - Model df - Model MS - Model SS - Residual df - Residual MS - Residual Model 1 210.88 0.146 0.126 8.824 2 4.412 51.77 85 0.609 7.245 0.00120 Model 2 206.89 0.183 0.164 11.10 2 5.55 49.49 85 0.582 9.529 0.00018 Model 3 206.17 0.208 0.180 12.60 3 4.20 47.99 84 0.571 7.36 0.00020 F p Table 1b: Effect sizes and power of the variables of model 3 SS Deg. of Freedom MS F p Observed power (alpha=0.05) Intercept 16.08 1 16.08 28.14 0.000001 0.999 Log GNI 1998 12.00 1 12.00 21.00 0.000016 0.995 ECI 1998 1.51 1 1.51 2.64 0.107938 0.362 Log SPSc 1998 3.78 1 3.78 6.62 0.011855 0.720 Error 47.99 84 0.57 A series of similar tests, using the Akaike information criterion corroborated the findings presented here (Interciencia). Thus, scientific knowledge as measured though the number of scientific publications of each country is a much better correlate to economic development as measured by GNDc than technological knowledge as estimated with ECI. A more detailed analysis is given in Jaffe et al. (2013a) RELATIONSHIP BETWEEN KNOWLEDGE AND ECONOMIC GROWTH When comparing scientific with technological development in its relationship with economic growth, it is not enough to asses which of the two knowledge indicators is better in tracking economic growth, but we need to know if both track the same think.

9 A comparison of these data is presented in Figures 2 to 7 which show the close correlation between scientific knowledge, technological knowledge and economic growth. ISO abbreviations (ISO) for the 90 countries with complete data sets ALB Albania GIN Guinea NLD Netherlands ARG Argentina GRC Greece NOR Norway AUS Australia GTM Guatemala NZL New Zealand AUT Austria HND Honduras OMN Oman BEL Belgium HUN Hungary PAK Pakistan BGR Bulgaria IDN Indonesia PAN Panama BOL Bolivia IND India PER Peru BRA Brazil IRL Ireland PHL Philippines CAN Canada IRN Iran PNG Papua New Guinea CHE Switzerland ISR Israel POL Poland CHL Chile ITA Italy PRT Portugal CHN China JAM Jamaica PRY Paraguay CIV Cote d'ivoire JOR Jordan ROU Romania CMR Cameroon JPN Japan SAU Saudi Arabia COG Congo, Rep. KEN Kenya SDN Sudan COL Colombia KOR Korea, Rep. SGP Singapore CRI Costa Rica KWT Kuwait SWE Sweden CUB Cuba LBN Lebanon SYR Syria DNK Denmark LBY Libya THA Thailand DOM Dominican Rep. LKA Sri Lanka TTO Trinidad and Tobago DZA Algeria MAR Morocco TUN Tunisia ECU Ecuador MEX Mexico TUR Turkey EGY Egypt MLI Mali TZA Tanzania ESP Spain MNG Mongolia UGA Uganda ETH Ethiopia MOZ Mozambique URY Uruguay FIN Finland MRT Mauritania USA United States FRA France MUS Mauritius VEN Venezuela GAB Gabon MYS Malaysia VNM Vietnam GBR United Kingdom NGA Nigeria ZMB Zambia GHA Ghana NIC Nicaragua ZWE Zimbabwe

10 Figure 2: Relation between GNI per capita and ECI for 2008. Countries are indicated using their ISO abbreviations. 2.5 ECI 2008 = -4.0975+1.1111*x; 0.95 Conf.Int. JPN ECI 2008 2.0 1.5 1.0 0.5 0.0-0.5-1.0-1.5-2.0 UGA TZA ETH ZWE MOZ MLI GIN AUT SWE CHE SGP FIN GBR HUN KOR FRA USA ITA IRL DNK MEX ISR POL BEL NLD CHN ROU ESP THA PAN MYS NOR PRT BGR CAN JOR COL CRI LBNTUR IND TUN BRA NZL GRC ZMB ALB ARG URY PHL EGY IDN GTM DOM VNM KEN LKA SAU SYR CHL AUS JAM MUS TTO PAK MAR CUB HNDPRY ECU PER NIC OMN CIV GHA BOL MNG VEN IRN DZA CMR GAB LBY NGA PNG SDN COG MRT -2.5 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 Log GNIc 2008 Figure 3: Relation between indices of GNI per capita and Scientific Productivity for 2008 1.0 Log SPSc 2008 = -5.3684+1.1905*x; 0.95 Conf.Int. Log SPSc 2008 0.5 CHE AUS FIN DNK ISR NZLSGP CAN BEL GBR NLD AUT IRL NOR 0.0 GRC ESP FRA USA PRT ITA KOR JPN HUN POL -0.5 BGR TUN ROU TUR CHL JOR IRN LBN MYS ARG URY CHN BRA TTO OMN CUB THA CRI MEX SAU JAM -1.0 PAN MNG EGY COL MUS DZA GAB VEN MAR IND COG KEN CMR LKA LBY -1.5 PAK ALB BOL ECU ZWE GHA NGA PER UGA PNG VNM ZMB CIV SYR TZA NIC -2.0 MLI PHL PRY ETH MRT SDN GTM MOZ HND IDN DOM -2.5 GIN 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 Log GNIc 2008

11 Figure 4: Relation between ECIc and SPSc. 1.0 Log SPSc 2008 = -0.9652+0.7089*x; 0.95 Conf.Int. 0.5 0.0 AUS NZL GRC CHE CAN NOR NLD DNK SGP FINSWE BEL ISR HKG IRL GBR AUT PRT ESP USA FRA ITA KOR HUN JPN Log SPSc 2008-0.5 IRN BGR CHL TUN TUR JOR MYS OMN TTO URY ARG BRA CHN CUB JAM SAU CRI THA MEX PAN -1.0 GAB MNG MUS EGY DZA VEN COL MAR IND COG CMR -1.5 PAK KEN LKA NGA BOL ECU ALB PER GHA ZWE PNG ZMB LAO UGA CIV SYR NIC -2.0 MRT SDN MDG MLI PRY PHL ETH GTM DOM MOZ HND IDN -2.5 GIN -2.5-2.0-1.5-1.0-0.5 0.0 0.5 1.0 1.5 2.0 2.5 ECI 2008 Figure 5: Relationship between ECI, SPSc in 1998 and economic wealth expressed as GNIc for 2008. The size of the bubbles is proportional to GNIc in 2008.

GIN CMR NGA COG PNG MLI CIV NIC TZA GHA UGA KEN HND SDN ETH MRT SYR ZWE GAB GTM ECU DZA JAM DOM PAK MNG IRN LKA MUS CUB VNM BOL CRI MAR MOZ ALB OMN PHL EGY PER PRY IDN PAN TTO TUN LBN CHL THA AUS IND SAU VEN GRC COL TUR CHN ARG ZMB JOR URY NZL MYS PRT ROU BGR BRA MEX KOR NOR HUN POL CAN ESP SGP ISR NLD ITA DNK IRL BEL FRAUT USA FIN GBR SWE CHE JPN -2.0-1.5-1.0-0.5 0.0 0.5 1.0 1.5 2.0 2.5 ECI 1998-3.0-2.5-2.0-1.5-1.0-0.5 0.0 0.5 Log SPSc 1998 GIN CMR NGA COG PNG MLI CIV NIC TZA GHA UGA KEN HND SDN ETH MRT SYR ZWE GAB GTM ECU DZA JAM DOM PAK MNG IRN LKA MUS CUB VNM BOL CRI MAR MOZ ALB OMN PHL EGY PER PRY IDN PAN TTO TUN LBN CHL THA AUS IND SAU VEN GRC COL TUR CHN ARG ZMB JOR URY NZL MYS PRT ROU BGR BRA MEX KOR NOR HUN POL CAN ESP SGP ISR NLD ITA DNK IRL BEL FRAUT USA FIN GBR SWE CHE JPN B D A C 12

13 Figure 6: Relationship between ECI, SPSc in 1998 and economic growth expressed as the % change in GNIc for 1998-2008. Bubbles with negative data are revalued to 0. The size of the bubbles is proportional to %ΔGNIc for 1998-2008. 0.5 0.0 A AUS NZL GRC SWE CHE ISR DNK NOR CAN NLD FIN GBR B BEL USA SGP FRAAUT IRL ESP ITA JPN HUN -0.5 PRT BGR POL KOR Log SPSc 1998 TTO CHL CUB TUR ARG JOR ROU -1.0 OMN SAU URY CRI TUN LBN BRA GAB JAM MEX VEN MYS MAR EGY CHN -1.5 MUS PAN COG ZWE THA IND PNG KEN MNG IRN ALB CMR DZA COL ECU NGA CIV GHA LKA BOL ZMB -2.0 MLI NIC TZA SYR PAK UGA PER PRY HND GTM PHL SDN ETH MRT VNM -2.5 DOM IDN MOZ GIN C D -3.0-2.0-1.5-1.0-0.5 0.0 0.5 1.0 1.5 2.0 2.5 ECI 1998 Figure 7: Correlation between growth in ECI and SPSc during 1998-2008 and economic development. The size of the bubbles indicates % Growth in GNIc during 1998-2008 1.6 1.4 A SGP B 1.2 CHE Δ SPSc 1998-2008 IRL 1.0 AUS NOR FIN 0.8 CAN NLD NZL GRC BEL AUT PRT DNK KOR 0.6 GBR ESP SWE ITA 0.4 FRA ISR POL ROU TUN IRN TUR HUN CHL BGR MYS 0.2 USA JOR LBN CHN BRA OMN URY TTO JPN MNG DZA ARG VEN MRT MOZ SDN KWT LBY GAB SAUPER BOL ZMB COG COL IND CUB NGA MAR PRY PNG 0.0 ZWE ETH IDN ECU GHA PAK CMR THA LKA PHL GIN MLI VNM JAM ALB HND CIV TZA UGA DOM NIC GTM CRI KEN PAN C D -0.2-1.6-1.4-1.2-1.0-0.8-0.6-0.4-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Δ ECI 1998-2008

14 Clearly, SPSc and ECI are closely correlated, but their differences reveal the countries peculiar mix of knowledge (Figure 4). Countries with a higher score for ECI than what the mean correlation would suggest, such as Japan, Germany (not shown in Fig 4), Mexico, China and India are countries with important non-scholar knowledge acquisition systems in place that complement scholar knowledge acquisition practiced in universities. Countries with a higher score for SPSc than the corresponding mean, such as Switzerland, Denmark and Chile, are countries with a strong long term investment in formal education. Countries such as Australia, Kuwait and Congo whose national wealth is based mainly on the exploitation of natural resources with a low dependence on industry, also show a relative higher SPSc compared to ECI at their corresponding level of economic development. The strong relationship between wealth (GNIc), scientific knowledge (SPSc), and to a lesser degree, technical knowledge (ECI), is evidenced in Figures 5-6. Figure 5 shows a cluster of countries with high scores in these indices (quadrant B). No country with low scores in two of these indices scored high in the other (quadrants A & C). The outlier data points for high GNIc, such as Australia (quadrant A ), reflects the fact that high commodity prices affect the wealth of a nations exploiting natural resources adding to that created by science and technology (Hausmann et al. 2011). Development of economic wealth and/or economic complexity without a simultaneous development of science was achieved by no country (quadrant D). Thus, a tightly knit ensemble of technical knowledge, scientific knowledge and wealth seems to exist in our economies. Figure 6 shows the relationship between the knowledge indicators ECI and SPSc in 1998 with percent economic growth in the following decade (1998-2008). All fast growing countries were found in quadrant C, where countries with low SPSc and ECI in 1998 are grouped. Transforming economic growth data in percentages, normalize for

15 the absolute amounts of ΔGNIc between countries. Many poor countries with low ECI and/or SPc showed high % growth in GNIc (quadrant C) and many rich counties showed low %ΔGNIc (quadrant B). That is, caching up in economic development seems to require less knowledge than expanding the economic frontier of humanity. Poorer countries have much more space to grow than richer ones, so that growth expressed as % favors poorer countries over richer ones. This may explain why no rich country with high scientific output (quadrant B) showed high %ΔGNIc in Figure 6. Figure 7 illustrates the divergence between ΔSPSc and ΔECI during 1998 to 2008. It shows that no country with a ΔECI scores of less than -0.6 increased its scientific productivity during this period (quadrants A). All countries with large increases in scientific productivity showed ΔECI between -0.6 and 0.1 (quadrant B). That is, countries with large increases in scientific productivity also showed relative high ΔECI, and countries with a large %ΔGNIc were countries with low ΔSPSc and relative high levels of ΔECI during the last decade (quadrant D). The countries with the largest ΔECI score (Panama, Kenya and Costa Rica) had only a modest %ΔGNIc and low ΔSPSc. The results in Figure 7 show us those rich countries with high scientific productivity and high ECI that got somewhat richer in the decade 1998-2008 and also increase their scientific productivity in this period (quadrant B). The poorer countries that showed important % increase in GNIc in the last decade got richer while avoiding drastic decreases in diversification of their economic activity (ΔECI>-0.6) but maintaining low growth in SP (ΔSPSc) in this period (quadrant D). No country managed high scientific productivity with low technical development (quadrant A). Countries with high increases in SPSc and healthy ECI (quadrant B) showed intermediate values of %ΔGNI. In this group, the outlier country with the lowest

16 %ΔGNI and highest ΔSPSc was Singapore, which showed high %ΔGNI in other periods. All fast growing countries were grouped in quadrant D. A few countries managed to achieve conspicuous industrial and scientific contraction simultaneously (quadrant C). DETERMINING THE SCIENTIFIC DISCIPLINES WITH THE CLOSEST LINK TO ECONOMIC DEVELOPMENT The picture revealed so far is still somewhat puzzling as many questions remain unanswered. What areas of academic activity are related to technological development and which areas are not? What kind of scientific development is actually measured with the total publication count as recorded by Scopus? In order to tackle these questions we compared the relative publication effort made by each country regarding research in different areas of knowledge, with its present and future national wealth. Data of the number of publication by area for each country for the years starting 1998 came from the database of Scopus complied by SCImago (2007), whereas data for 1982 and 1992 was compiled manually by us from the Web of Science. We calculated the relative research effort of each scientific subject area as the percentage of the total number of publications of that country published in journals of that area in a year. For example, the number of publications in mathematical journals of that country, divided by the total number of publications in all subject areas of that country, multiplied by 100, served as the estimate of relative research effort in mathematics for that country. This number was used to calculate the Revealed Comparative Advantages (RCA) of the scientific publication effort, adapted from the economic literature (Laursen 1998). RCA is a ratio of two shares. The numerator is the share of a country s publications in a given discipline or area of science in its total

17 number of publications. The denominator is the share of the world s number of publications in that same discipline in the total world s publications. In order to avoid statistical pitfalls due to non-linearity in our data, we used only nonparametric statistics for the analysis of the relationship between RCA and economic growth. Only countries with more than 100 publications in 1982, or 200 in 1996, and which had their GDPc data for the required years in the World Bank database, were taken into account. Economic wealth was estimated using the Gross National Product per capita (GDPc) as calculated by the World Bank (GDP per capita based on purchasing power parity at constant 2005 US $). Percentage growth in wealth was estimated by calculating the perceptual increase of GDPc during a given period of time. Countries with over 100 publications in 1998 recorded by Scopus and with GDPc data provided by the World Bank were used for the present analysis. Only 101 countries fulfilled these criteria. Rich countries with high GDPc show higher research efforts certain scientific disciplines, whereas poor countries with low GDPc show higher research efforts in other disciplines (Table 2). The table shows the correlations between RCA or the relative research effort in each discipline assed by the publication record of the year 2010 of each country, with its GDPc of the same year. The table shows that richer countries publish more and therefore probably invest more research effort in neurosciences, computer sciences and psychology than poorer ones; whereas poorer countries publish more research in agriculture and multidisciplinary sciences. Correlations between the RCA of the publication effort of scientific disciplines during 2000 with economic growth in the following years, estimated as percent increase in GDPc during the periods 2000-2005 (Table 3) shows a different

18 result. Here relative research efforts in physics and chemistry were the best predictors for future economic growth, and efforts in medicine and psychology the best predictors for poor future economic growth. A part, but certainly not all, of the correlation between relative productivity in physical and chemical science and future economic growth could be explained by an additional correlation with development of technological knowledge. The ECI, as calculated in by Hausmann et al (2011), mirrors some but not all of the patterns of correlation between RCS in scientific publications and GDPc growth in the following 5 years. For example, RCA in physics and material sciences was positively correlated to both, the economic complexity index achieved 8 years later and the economic growth achieved 5 years later. RCA in chemistry, however, did not correlate significantly with economic complexity but did correlate positively with economic growth. RCA in computer science, health, biochemistry and neuroscience, for example, correlated with future economic complexity but not with economic growth. In summary, the results show a very strong correlation between the relative importance given to the basic natural science in a country and its future economic growth. This correlation is not evidenced when using the absolute productivity of a given discipline, but appears only when this effort is represented as a percentage of total productivity or as RCA. That is, it is not that science stimulates technology that in turn stimulates economic development, it is that a given scientific socio-political culture favors both, scientific development, technological development and economic growth. A more detailed analysis of these relationships is presented in Jaffe et al (2013b)

19 Table 2: Correlations between the research efforts in the various disciplines in each of 85 countries during 2010, measured as the RCA of publications pertaining to that area; and the wealth of that country in GDPc for 2010. n indicates number of countries with data for the corresponding analysis. Spearman correlations ECI 2008 (n=76) p-level GDPc 2010 (n=80) p-level Neuroscience 0.676 0.0000 0.718 0.0000 Psychology 0.361 0.0014 0.632 0.0000 Computer Science 0.531 0.0000 0.586 0.0000 Biochemistry, Genetics & Molecular Biology 0.642 0.0000 0.569 0.0000 Arts & Humanities 0.269 0.0188 0.493 0.0000 Health Professions 0.365 0.0012 0.477 0.0000 Decision Sciences 0.231 0.0444 0.442 0.0000 Business, Management & Accounting 0.225 0.0502 0.411 0.0002 Economics, Econometrics & Finance 0.103 0.3781 0.369 0.0008 Dentistry 0.184 0.1110 0.262 0.0197 Engineering 0.308 0.0068 0.253 0.0244 Medicine 0.045 0.6993 0.204 0.0719 Social Sciences -0.074 0.5272 0.151 0.1848 Earth & Planetary Sciences -0.067 0.5637 0.144 0.2057 Nursing -0.106 0.3617 0.068 0.5516 Physics & Astronomy 0.284 0.0130 0.068 0.5530 Mathematics 0.076 0.5166 0.052 0.6515 Veterinary -0.059 0.6109 0.008 0.9470 Chemical Engineering 0.102 0.3807 0.007 0.9509 Materials Science 0.158 0.1729-0.040 0.7249 Chemistry 0.175 0.1306-0.059 0.6034 Energy -0.112 0.3337-0.133 0.2435 Immunology & Microbiology -0.137 0.2363-0.172 0.1306 Pharmacology, Toxicology & Pharmaceutics -0.149 0.1975-0.197 0.0824 Environmental Science -0.292 0.0104-0.240 0.0329 Multidisciplinary -0.315 0.0056-0.280 0.0126 Agricultural & Biological Sciences -0.362 0.0013-0.341 0.0021 Total Publication/capita (n=79) 0.79 0.000 0.90.000

20 Table 3: Correlation between the relative research effort (RCA) of a given discipline in the year 2000 in 80 countries with the economic growth of these countries as % difference in GDP per capita during the years 2000-2005. n indicates number of countries with data for the corresponding analysis. Spearman correlations ECI 2008-1998 (n=61) p-level GDPc 2005-2000 (n=89) p-level Materials Science -0.029 0.8231 0.498 0.0000 Chemistry -0.117 0.3713 0.439 0.0000 Physics & Astronomy -0.132 0.3095 0.399 0.0002 Mathematics -0.263 0.0402 0.259 0.0206 Engineering -0.137 0.2909 0.241 0.0316 Energy 0.321 0.0117 0.176 0.1179 Chemical Engineering -0.194 0.1346 0.139 0.2190 Earth & Planetary Sciences 0.049 0.7062-0.024 0.8358 Decision Sciences -0.209 0.1068-0.050 0.6574 Business, Management & Accounting 0.046 0.7270-0.102 0.3698 Computer Science -0.267 0.0373-0.122 0.2817 Health Professions -0.277 0.0308-0.260 0.0201 Environmental Science 0.145 0.2657-0.272 0.0148 Dentistry -0.106 0.4141-0.287 0.0097 Social Sciences 0.108 0.4075-0.301 0.0067 Agricultural & Biological Sciences 0.231 0.0729-0.303 0.0063 Arts & Humanities -0.102 0.4354-0.305 0.0059 Veterinary -0.036 0.7823-0.328 0.0030 Economics, Econometrics & Finance -0.094 0.4706-0.345 0.0017 Immunology & Microbiology 0.019 0.8841-0.382 0.0005 Biochemistry, Genetics & Molecular Biology -0.380 0.0025-0.403 0.0002 Pharmacology, Toxicology & Pharmaceutics -0.148 0.2560-0.404 0.0002 Neuroscience -0.549 0.0000-0.468 0.0000 Nursing -0.147 0.2596-0.483 0.0000 Psychology -0.440 0.0004-0.502 0.0000 Multidisciplinary -0.294 0.0213-0.504 0.0000 Medicine -0.232 0.0719-0.579 0.0000 Total Publication/capita (n=80) -0.52 0.0000-0.31 0.006

21 CONCLUSIONS Many more factors than those analyzed here affect economic growth and may explain differences in the wealth of nations. Here we focused only on the difference in the contribution of scientific knowledge and technical expertise (industrial economic complexity) in powering the wealth of nations. In order to give a comprehensive interpretation to the results, I want to pinpoint the following aspects: 1. A tight relationship between economic prosperity, scientific development and economic complexity. We showed that technological knowledge required to power industrialization in an economy, measured with ECI, differs from that produced by science, measured with SPc. Both measures correlate with the wealth of nations, but SPc does so to a larger extent than ECI. Short term economic growth in economically less developed countries may not require autochthonous production of science or technology, but sustained economic well-being seems to be inseparable from both types of knowledge. 2. Repeating this analysis for other time periods with different macro-economic conditions would be desirable, but publication counts in the past were much more skewed to English language journal that they are today, distorting any rigorous historical analysis. Thus we used other methods for a historical analysis, such as comparing the changes in the relative importance of the different scientific disciplines in each country with its economic growth. These studies showed that the development of the natural sciences correlates with future economic growth in most historical periods examined.

22 3. Clearly, correlation is not causation. Even if some of the variables may be good in predicting future economic growth, this does not assure that there is a causal effect between them. No direct correlation between development in basic science and economic growth, or vice versa, exists (Jaffe et al. 2013b). We suggest that relative investment in basic science is a reliable indicator of a rational decision making atmosphere in a country. This atmosphere, if other factors allow, promotes economic growth (Jaffe, 2009). Detailed case studies, not attempted in this work, are required for this. The relationship between economic wealth, scientific knowledge and economic prosperity seems to be multi-directional. Rich countries can afford more science and more technology and more science and technology allows increasing and maintaining wealth. However, scientific knowledge affects the nation s economy only after a certain economic development has been achieved (Royal Society, 2011), whereas technical expertise may favor in some cases growth in poor countries. More technology allows for more and better science and more science advances technology. This multi-mutual relationship allows proposing the existence of a Wealth-Science-Technology Complex, whose fine workings are still open to more research. 4. Academic scientific productivity is a much better predictor of future economic growth than economic complexity. If we accept that science is the mother of technology, i.e. supports technological development, and that science affects other aspects of live such as services, governability, rational thinking, attitudes, etc. and of the economy besides technological development, this result seems congruent. 5. Science is essential for economic growth among developed economies, whereas technical complexity may be more important for the economic development of poorer

23 countries. Rich nations seem to slow their exponential growth in scientific productivity per capita as they seem to reach an upper limit in the intensity of scientific productivity. This trend presages that future worldwide scientific growth will relay more on the economic development of poorer countries. 6. We have to remark that the present study excluded countries with low scientific productivity, which include all poor countries. Previous studies (Jaffe 2005) showed that the correlation between science and wealth of a country appears only after a threshold of economic development has been reached and that a rapid increase in scientific productivity was normally observed after a previous increase in economic development. On the other hand, the relative effort to support academic activity in rich countries seems to be close to the maximum tolerated by society. Rich countries have completed their scientific and industrial revolution in the past and focus now on other aspects of the well-being of their citizens, as they have to manage low economic growth. This would explain the low correlations found between scientific publications and future economic growth in rich countries. Therefore, the present conclusions are valid only for middle income countries. Having all these information in mind, we can now analyze the recommendations by the controversial economist Jeffrey Sachs (2005) for developping countries. He recommended health, energy, agriculture, climate and ecology as the areas of science where investments were most likely to promote economic growth. None of them came out as positively correlated here. On the contrary, countries that knowingly or unknowingly complied with Sachs s recommendations achieved very poor economic growth. It is investment in hard sciences and basic sciences, such as physics and

24 chemistry that correlate strongest with economic growth. Material sciences are normally considered to be part of physics although Scopus computes the publication in this area separately. Our results show correlations between basic natural science and economic development that is not due to direct causal chains. This is in agreement with more recent empirical explorations in economics (Azoulay et al. 2011) that revealed an intricate network of reciprocal relationships between knowledge, services, environment and finance. Here we propose that scientific development works in an analogous way, affecting multiple aspects of the economy and in turn being affected by many of these aspects producing positive feedback cycles. Hirschman (1958) postulated the high development theory, as the view that development is a virtuous circle driven by external economies -- that is, that modernization breeds modernization. Some countries, according to this view, remain underdeveloped because they have failed to get this virtuous circle going, and thus remain stuck in a low level trap. Our data would support the proposition that investing in basic scientific research seem to be the best way a middle income country can foment fast economic growth, triggering Hirschman s virtuous cycle. This proposition is also used by Lin (1995) to solve the Needham Puzzle: Why the Industrial Revolution did not originate in China. The scientific revolution needs a profound conceptual revolution which is achieved by the development of basic natural sciences (Jaffe 2009). An empirical test of the predictive power of the ranking of RCA is that for 2010 they showed that the countries with an RCA value in Physics above 2.0 were Armenia, Ukraine, Moldova, Uzbekistan, Russia, Belarus, Bulgaria, Kazakhstan and Georgia. Most of them are among the fastest growing economies in 2012. In a few years time we

25 might know if they can serve as an example for countries in Africa or Latin America, as none of them are on this list of countries valuing basic natural science. REFERENCES Akaike, H. (1974). A new look at the statistical model identification". IEEE Transactions on Automatic Control 19 (6): 716 723. doi:10.1109/tac.1974.1100705. Azoulay, P., Graff-Zivin, J., Sampat B. (2011). The diffusion of scientific knowledge across time and space: Evidence from professional transitions for the superstars of medicine. National Bureau of Economic Research Working Paper #16683. Bacon, F. (1620). Novum Organum Scientiarum. In The Novum organon, or a true guide to the interpretation of nature, trans. G. W. Kitchin. Oxford: Oxford University Press, 1855. Benjamini, Y., Hochberg, Y. (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society, Series B, 57, 289-300. Condorcet, M.J.A. (1794). Sketch for a historical picture of the progress of the human mind: tenth epoch. Daedalus 133 (3): 65-82. Granger C.W.J. (1969). Investigating causal relations by econometric models and crossspectral methods. Econometrica 37, 424-438. Hausmann R., Hidalgo C.A., Bustos S. Coscia M. Chung S. Jimenez J., Simoes A., Yıldırım M.A. (2011). The Atlas of Economic Complexity: mapping paths to prosperity. Hidalgo C.A. and Hausmann R. (2009). The building blocks of economic complexity PNAS 106: 1057-1075. Hidalgo C A, Klinger B, Barabási Al and Hausmann R. (2007). The Product Space Conditions the Development of Nations. Science 317: 482-487. Hirschman, A. (1958). The Strategy of Economic Development. New Haven, Conn.: Yale

26 University press. Jaffe K. (2005). Science, religion and economic development. Interciencia 30: 370-373. Jaffe K. (2009). What is Science? An interdisciplinary perspective. University Press of America. Jaffe K. (2011). Do countries with lower self-citation rates produce higher impact papers? Or, does humility pay. Interciencia 36: 696-698. Jaffe K., Rios, A. & Florez, A. (2013a). Statistics shows that economic prosperity needs both high scientific productivity and complex technological knowledge, but in different ways. Interciencia 38: 150-156 Jaffe, K., Caicedo, M., Manzanares, M. Rios, A., Florez, A. Montoreano, C., Davila, V. (2013b). Productivity in physical and chemical science predicts the future economic growth of developing countries better than other indices for knowledge. PLoS One (in press). King D.A. (2004). The scientific impact of nations. Nature, 430: 311-316. Laursen K. (1998). Revealed Comparative Advantage and the Alternatives as Measures of International Specialization. DRUID Working Papers 98-30, DRUID, Copenhagen Business School, Department of Industrial Economics and Strategy/Aalborg University. Lin J.Y. (1995). The Needham Puzzle: Why the Industrial Revolution did not originate in China. Economic Development and Cultural Change 43: 269-292. Marshall A. (1890). Principles of Economics. Amazon Kindle edition. Nelson R. R. (1959). The Simple Economics of Basic Scientific Research. Journal of Political Economy, 67: 297-306. Royal Society (2011). Knowledge, Networks and Nations: Global Scientific Collaboration in the 21st Century. RS Policy document 03/11. London, UK: http://royalsociety.org/uploadedfiles/royal_society_content/influencing_policy/r eports/2011-03-28-knowledge-networks-nations.pdf Sachs J. (2005). The End of Poverty. Penguin Press 416 pp. Salter A.J., Martin B.R. (2001). The economic benefits of publicly funded basic research: A critical review. Research Policy, 30: 509 532. SCImago (2007). SJR SCImago Journal & Country Rank. Retrieved April 18, 2012, from http:www.scimagojr.com Spencer, H. (1875). Progress: Its law and cause. In: Illustrations of universal progress: A

27 series of discussions. pp. 1-60. New York, NY, US: D Appleton & Company, xxv, 451 pp. Teitel, S. (1994). Scientific publications, R&D expenditures, country size, and per capita income: A cross section analysis. Technological Forecasting and Social Change, 46 Pages 175 187. Wang, E.C. (2007). R&D efficiency and economic performance: A cross-country analysis using the stochastic frontier approach. Journal of Policy Modeling, 29: 345 360.