Policy Research Working Paper 8870 Technology Adoption and the Middle-Income Trap Lessons from the Middle East and East Asia Rabah Arezki Rachel Yuting Fan Ha Nguyen Middle East and North Africa Region Office of the Chief Economist May 2019 Policy Research Working Paper 8870 Abstract This paper documents the existence of a “middle-income and North Africa having experienced a relatively slow pace trap” for the Middle East and North Africa region and of technology adoption in general-purpose technologies contrasts the evidence with that of the East Asia and Pacific and that a slower adoption pace of technology is associated region. The results are two-folds. First, non-parametric with lower levels of economic growth. The paper concludes regressions show that the average rate of economic growth that barriers to the adoption of general-purpose technol- in the Middle East and North Africa has not only been ogies related to the lack of contestability in key sectors significantly lower than that in the East Asia and Pacific constitute an important channel of transmission for the region, but it has also tended to drop at an earlier level of middle-income trap. income. Second, econometric results point to Middle East This paper is a product of the Office of the Chief Economist, Middle East and North Africa Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at rarezki@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Technology Adoption and the Middle-Income Trap: Lessons from the Middle East and East Asia Rabah Arezki, Rachel Yuting Fan and Ha Nguyen Updated May 30, 2020 JEL: O4 Keywords: middle-income, growth, technology adoption I. Introduction The term “middle-income trap” refers to the possibility that economies could get stuck at a certain level of income. The debate on the trap has thus far focused mostly on the East Asia and Pacific region (EAP). 1 While economies in Middle East and North Africa region (MENA) have stalled, they have largely been overlooked in the debate over the middle-income trap. Indeed, MENA has been characterized by pervasively low growth. In the 1980s and 1990s, GDP growth per worker in the region was less than 1 percent per year, with continuous decline in total factor productivity (Yousef, 2004). In recent decades, growth in MENA has remained relatively low (see Figure 1). 2 In the present paper, we document the existence of a middle-income trap for MENA and contrast the evidence with that of EAP. To do so, we adopt a non-parametric analysis of growth dynamics that helps flexibly capture sharp changes in growth. Results from non-parametric regressions show that growth in GDP per capita and total factor productivity (TFP) in MENA quickly decline as income levels rise. In contrast, growth in GDP per capita and TFP in EAP is not only higher on average along the income ladder but also decline at much higher levels of income. Importantly, we document that the slow pace of technology adoption of general-purpose technologies (GPT) is associated with lower levels of economic growth. We then examine the adoption of both older GPT and their applications such as electricity, and newer ones, such as broadband and internet. For all technologies, when controlling for the level of income, MENA falls behind EAP in terms of the adoption pace. Barriers to the adoption of general-purpose technologies thus constitute an important channel of transmission for the middle-income trap. This paper is most directly related to the strand of literature testing for the existence of a middle-income trap. For example, Aiyar et al. (2013) uncover that middle-income countries are more likely to experience growth slowdowns. Also, Eichengreen et al. (2013) determines that level of income within the $10,000- $11,000 and $15,000-$16,0000 ranges. The jury is however still out on the empirical validity of the 1 The term “middle income trap” was first coined by Gill, Kharas and others (2007). Policymakers and commentators have used the term abundantly in the media to characterize the risk of facing a ceiling on the level of economic growth for countries such as Malaysia, Vietnam and China. Also, researchers have investigated the risk associated with the trap in Asia and as well as the needed reforms to escape it (Ohno and Le, 2015 for Vietnam; Fragen et al, 2013 for Malaysia; Eichengreen et al, 2012 and Glaw and Wagner, 2017 for China). 2 Figure 1 shows that for the period from 2000 to 2021, MENA countries, with the exceptions of Djibouti and Morocco, are expected to experience lower growth in GDP per capita than the median of other countries in the same income group. The years from 2019 to 2021 are projections. 2 middle-income trap. 3 The contribution of this paper is to provide evidence that MENA is subject to much lower levels of growth along the income ladder compared to EAP. Figure 1: MENA Growth performance has been subpar 2000-2021 Lower 0.04 Upper Middle Middle 0.03 0.02 High Income 0.01 0.00 -0.01 -0.02 Yemen, Rep. Lebanon Libya Bahrain Egypt United Arab Emirates Kuwait Tunisia Oman Djibouti West Bank & Gaza Qatar Jordan Algeria Saudi Arabia Iran Iraq Morocco Note: The blue diamonds are country average growth in GDP per capita. The red lines capture the median growth in GDP per capita in non-MENA countries in the same income group. Source: World Economic Outlook and World Bank’s Macro and Poverty Forecasts This paper is also related to the literature on the link between innovation and economic growth. In Schumpeterian growth theory, faster growth is associated with higher rates of firm creation and destruction driven by R&D and innovation (Aghion and Howitt, 1992). In this environment, incumbent firms’ innovation and productivity growth would be stimulated by competition and entry, particularly in firms near the technology frontier (Aghion et al, 2014). 4 There is strong empirical evidence that competition and productivity growth display an inverted-U shaped relationship: starting at an initially low level of competition, higher competition stimulates innovation and growth; however, starting from a higher initial level of competition, higher competition may hurt innovation and productivity growth. 5 This paper documents MENA’s slow pace of adoption in GPT which can help explain the pervasively low economic growth and TFP. This paper also provides evidence that technology adoption is slower when concentration is higher in key (upstream) sectors of the economy. 3 Bulman et al. (2017) find that the fraction of countries “trapped” at the middle-income level is not larger than the fraction of countries “trapped” at the low-income level. Similarly, Han and Wei (2017) find that the probability of escaping from the middle-income level is not smaller than the probability of escaping from the low-income level. 4 See Aghion et al. (2014) for a recent review. 5 See for instance Aghion et al. (2005). 3 The remainder of the paper is organized as follows. Section II documents the evidence of a middle- income trap for MENA. Section III explores the link between technology adoption and economic growth. Section IV presents evidence of the relatively slow pace of technology adoption of GPT in MENA. Section V concludes. II. Empirical Evidence for the Middle East’s Middle-Income Trap MENA countries are less likely to escape the middle-income trap than other countries around the globe. Figure 2 illustrates that by comparing levels of income reached in 1975 to the ones in 2017. We follow Bulman et al. (2017) in grouping countries into three relative income groups, namely low-income, middle-income and high-income depending on their GDP per capita relative to that of the United States in the same year. 6 Countries in the middle-left quadrant escaped the low-income group in 1975 and shifted to the middle-income group in 2017. Countries in the top-middle quadrant escaped the middle-income group and shifted to the high-income group. Countries in the center quadrant have been trapped in the middle-income group for more than four decades. Among MENA countries, aside from the six countries that have remained high-income, five have been trapped in the middle-income group (Algeria, Egypt, Jordan, Morocco, and Tunisia), three fell from the high-income to middle-income group (the Islamic Republic of Iran, Lebanon, and Libya), while none have become “escapees”. The Republic of Korea; Hong Kong SAR, China; Cyprus and Portugal have become escapees. Figure 2: Illustrating the Middle-Income Trap log of real PPP GDP per capital relative to the U.S. in 2017 2.5 QAT KWT ARE HKG 2 SAU BHR OMN KOR CYP PRT IRN 1.5 LBN CHN DZA LKA IDN EGY JOR TUN LBY BTN MAR LAO VNM IND 0 .5 1 0 1 2 3 log of real PPP GDP per capital relative to the U.S. in 1975 Sources: International Monetary Fund, World Economic Outlook database, and authors’ calculations. Note: Data labels use International Organization for Standardization (ISO) country codes. Regions follows World Bank country groups. 6 A country is defined as low-income if its per capita GDP is lower than or equal to 10 percent of that of the United States; middle-income if between 10 percent and 50 percent of U.S. GDP, and high-income if above 50 percent. 4 To document more systematically the evidence of a middle-income trap for MENA relative to EAP, we use the non-parametric local-linear regression technique that give the mean and standard errors of the estimated growth rate of each region at each level of income: 7 ∆log (),+10 − ∆log (),+10 = � , � + where ∆log (),+10 and ∆log (),+10 are overlapping annualized decadal growth in GDP per capita (or TFP) of country and of the U.S. between time and time + 10, is the country’s relative income per capita relative to the U.S. at time . We use GDP per capita derived from output-side real GDP at chained PPPs and total factor productivity (TFP), both from Penn World Table 9.0. The regressions also include overlapping decade fixed-effect, , to control for common global shocks. Note that we are agnostic about the form of the function � , �. Unlike linear regression, a nonparametric regression is agnostic about the functional form between the outcome and the explanatory variables and is therefore not subject to misspecification error. In our context, a non-parametric regression could capture sharp changes in growth rates as relative income rises, a key advantage for us to identify an income trap. For each region, the non-parametric regressions (with 100 bootstrap replications) help provide the average predicted values and confidence intervals of annualized decadal growth in GDP per capita at different levels of relative income. Average predicted relative growth in GDP per capita relative to the U.S. (and its 95% confidence interval) for MENA and EAP are visually shown in Panel A of Figure 3, while those of absolute growth in GDP per capita are shown in Panel B 8. Their numerical values are reported in Appendix Table A1. The results for other regions are also reported in Appendix Figure A2, although not discussed in the text. For EAP countries, both average relative and absolute growth in GDP do not significantly decline until the countries reach 60 percent of U.S. GDP per capita. At 50 percent or below, EAP economies maintain a stable growth rate at 4 to 4.5 percent (Panel B), or 2 to 2.5 percent higher than the U.S. (Panel A) indicating that these countries are catching up. In contrast, the growth performance of MENA countries is much weaker. Although starting at the same level of growth as EAP, around 4 percent, growth for MENA 7 The STATA command is npregress kernel y x1 x2, where y is the dependent variable, and x1 and x2 are the explanatory variables. See Fan and Gijbels (1996) for a reference on local-linear regressions. 8 Note that we restrict the estimation at below 100 percent of U.S. income because we focus on the middle-income level. In addition, at above 100 percent of U.S. income, there are fewer observations making the estimations imprecise. 5 quickly and steadily declines. At 20 percent of U.S. GDP, the average growth rate for MENA is about 3 percent (Panel B), only 1 percent higher than that of the U.S. (Panel A) as opposed to almost 3 percent gap as in EAP. At 40 percent of U.S. income, MENA relative growth in GDP per capita becomes insignificantly different to that of the U.S., and starting from 60 percent of U.S. income, MENA growth is lower than that of the United States. The steady decline in per capita GDP in MENA along the income ladder indicates stronger evidence of the middle-income trap for MENA than for EAP—the region most prominently associated with the debate about the middle-income trap. 6 Figure 3: Growth in PPP GDP per capita Panel A: Relative to the U.S. East Asia and Pacific Middle East and North Africa .01 .02 .03 .04 .05 .01 .02 .03 .04 .05 relative growth in income relative growth in income -.05 -.04 -.03 -.02 -.01 0 -.05 -.04 -.03 -.02 -.01 0 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 relative income to the U.S., percentage relative income to the U.S., percentage Panel B: Actual growth (not relative to the U.S.) East Asia and Pacific Middle East and North Africa -.05 -.04 -.03 -.02 -.01 0 .01 .02 .03 .04 .05 -.05 -.04 -.03 -.02 -.01 0 .01 .02 .03 .04 .05 absolute growth in income absolute growth in income 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 relative income to the U.S., percentage relative income to the U.S., percentage Note: MENA includes Algeria, Bahrain, Djibouti, the Arab Republic of Egypt, the Islamic Republic of Iran, Iraq, Jordan, Kuwait, Lebanon, Morocco, Oman, Qatar, Saudi Arabia, the Syrian Arab Republic, Tunisia, United Arab Emirates, and the Republic of Yemen; EAP includes Australia, Brunei Darussalam, Cambodia, China, Fiji, Hong Kong SAR-China, Indonesia, Japan, the Republic of Korea, Lao PDR, Macao SAR-China, Malaysia, Mongolia, Myanmar, New Zealand, Philippines, Singapore, Thailand, and Vietnam. A similar pattern emerges when we explore the evolution of TFP growth. Figure A1 in the Appendix shows the results of non-parametric regressions for relative and absolute TFP growth for EAP and MENA. In both relative TFP growth (Panel A) and absolute TFP growth (Panel B), MENA under- performs compared to EAP along the income ladder. MENA’s absolute TFP growth is downward-sloping and quickly falls below zero when the countries reach 20 percent of U.S. GDP per capita. EAP’s absolute TFP growth, on the other hand, is stable at 1 percent level. In relative terms, MENA’s TFP growth is almost always below that of the United States. 7 Figure 4: Growth in PPP GDP per capita – MENA sub-regions Panel A: Relative growth to the U.S. Panel B: Absolute growth. Note: Gulf Cooperation Council (GCC) consists of Bahrain, Kuwait, Oman, Qatar, Saudi Arabia and United Arab Emirates. Other oil exporting countries include Algeria, the Islamic Republic of Iran, Iraq, the Syrian Arab Republic and the Republic of Yemen. Other oil importing countries include Djibouti, the Arab Republic of Egypt, Jordan, Lebanon, Morocco, and Tunisia. The pattern of the middle-income trap is robust across three sub-regions in MENA. 9 All the sub-regions have experienced a decline in GDP per capita at early levels of income (Figure 4), consistent with the regional overall pattern shown in Figure 3. 10 GCC countries perform best in terms of growth. Growth in GDP per capita in the GCC does not drop to below zero when the countries are still below the U.S. level of per capita income. 11 In contrast, growth in per capita income of other oil exporting countries and oil importing countries quickly drops as their income rises. Specifically, growth falls below zero at about 30 9 Countries in the Gulf Cooperation Council (GCC) are Bahrain, Kuwait, Oman, Qatar, Saudi Arabia and United Arab Emirates. Other oil exporting countries are Algeria, the Islamic Republic of Iran, Iraq, the Syrian Arab Republic, and the Republic of Yemen. Other oil importing countries are Djibouti, the Arab Republic of Egypt, Jordan, Lebanon, Morocco, and Tunisia. 10 Unfortunately, the TFP data for MENA do not allow us to run non-parametric regressions at the sub-region level. 11 The focus of our paper being on middle-income, we do not examine the performance of the GCC when their per capita income is higher than that of the United States. 8 percent of the U.S. per capita income for other oil exporting countries and at about 20 percent of the U.S. per capita income for other oil importing countries. III. Technology Adoption and Economic Growth There are many possible causes for MENA’s relatively slow growth. In this section, we focus on poor technology adoption. The literature has identified technology adoption as one important cause for economic growth (see Parente and Prescott, 1994 and Temple, 1999). To do so, we simply regress decadal growth on the initial level of income, a measure of technology adoption that is Technology Readiness obtained from the World Economic Forum and the interaction between the latter two terms. Technology readiness captures availability of latest technologies, firm-level technology absorption, FDI and technology transfers, and other indicators of technology adoption. 12 The interaction allows to explore the importance of the technology adoption in driving growth at different levels of income. Results presented in Table 1 show that higher technology adoption is associated with higher economic growth and that the effect of technology also differ depending on the initial level of income. Indeed, the coefficient of the interaction term is significantly positive in all three columns, indicating that given the same initial income level, a high ranking of technology readiness is associated with higher economic growth. According to column (3), for a country whose initial GDP per capita is 50 percent of that of the US, increasing average technology readiness ranking by 10, would increase annual growth of GDP per capita in the next decade by 0.8 percent. 12 Data for Technology Readiness are from Global Competitiveness Index. The index captures: availability of latest technologies, firm-level technology absorption, FDI and technology transfer, individuals using internet, fixed broadband internet subscriptions, international internet bandwidth, and mobile broadband subscriptions. In the following we use the terms “technology adoption” and “technology readiness” interchangeably. 9 Table 1. Technology Adoption Readiness and Growth (1) (2) (3) Relative decadal growth Relative income -0.0130*** -0.0135*** 0.0302*** (0.00186) (0.00168) (0.00328) Average technology readiness (-) 0.000243*** 0.000243*** (0.0000146) (0.0000133) Relative income # Average technology readiness (-) 0.000335*** 0.000259*** 0.00169*** (0.0000459) (0.0000415) (0.0000846) Observations 6319 6319 6319 Country fixed effect no no yes Year fixed effect no yes yes R-square 0.115 0.286 0.451 Notes: Coefficient estimates from ordinary least squares regressions at the country-year level. Standard errors are given in parentheses. * p<0.1, ** p<0.05, *** p<0.01. The dependent variable is the relative annualized overlapping decadal growth of real GDP per capita, compared to the growth in the US. Relative income is the relative real GDP per capita from the initial year of the decade (US’s real GDP per capita at the same year equals 1). Average technology readiness in the regression represents the average ranking for technology readiness. A higher number means a better ranking and higher technology readiness. The main variable of interest in all columns are the technology readiness, interacted with relative income from the initial year. The coefficient estimates associated with the constant are not reported to save space. Column (1) has no fixed effects, while column (2) is added with year fixed effects. In column (3), we added country fixed effect to replace the linear term of average technology readiness, in order to capture country-specific characteristics in addition to technology readiness. See Appendix Table A2 for the list of countries. To address concerns about endogeneity associated with technology adopted, we instrumented Technology Readiness with variable capturing variables capturing the attitude toward innovation and risks presented in Hofstede et al (2010). 13 Attitudes toward innovation vary considerably across countries. These attitudes play a critical role in driving decision of governments, firms, individuals toward adoption of technology and innovation. Figure 5 provides illustrative evidence of the powerful relationship between attitude traits and technology readiness. The correlations validate that the most relevant psychological traits are power distance (the way in which power is distributed), avoidance of uncertainty, and individualism (see Figure 5). Other dimensions that might affect are tough versus tender, (short-term) normative versus (long-term) pragmatic, and indulgence versus restraint. We use all six dimensions to instrument technology readiness in the first table, and the results are presented in Table 2. 13 Data are from Hofstede Insights: https://hi.hofstede-insights.com/national-culture 10 Table 2. Growth and technology, OLS and IV regressions Relative decadal growth (1) (2) (3) (4) (5) (6) OLS IV OLS IV OLS IV Relative income -0.0525*** -0.0555*** -0.0558*** -0.0544*** -0.0654*** -0.0862*** (0.00297) (0.00366) (0.00266) (0.00323) (0.00656) (0.00721) Average technology readiness (-) 0.000383*** 0.000413*** 0.000407*** 0.000389*** (0.0000262) (0.0000347) (0.0000234) (0.0000305) Relative income x Average technology readiness (-) 0.000691*** 0.000582*** 0.000662*** 0.000524*** 0.00275*** 0.00202*** (0.0000756) (0.0000874) (0.0000672) (0.0000770) (0.000174) (0.000205) Observations 2794 2794 2794 2794 2794 2794 Year fixed effect no no yes yes yes yes Country fixed effect no no no no yes yes R-square 0.201 0.200 0.382 0.380 0.555 0.552 First-stage F-stat 249.5 255.1 1009.8 First-stage Sargan-stat 70.56 95.94 107.0 Notes: Coefficient estimates from ordinary least squares regressions at the country-year level. Standard errors are given in parentheses. * p<0.1, ** p<0.05, *** p<0.01. The dependent variable is the relative annualized decadal growth of real GDP per capita, compared to the growth in the US. Relative income is the relative real GDP per capita from the initial year of the decade (US’s real GDP per capita at the same year equals 1). Average technology readiness in the regression represents the average ranking for technology readiness. A higher number means a better ranking, and higher technology readiness. This variable is instrumented in column (2) (4) (6), by 6-dimensions of country specific attitudes. First stage F-stat and Sargan test for over-identification are both reported. Regressions in all columns have the same sample to ease comparison. The coefficient estimates on constant are not reported to save space. Column (1) and (2) has no fixed effects, while column (3) and (4) is added with year fixed effects. In column (5) and (6), we added country fixed effect to replace the linear term of average technology readiness, in order to capture country specific characteristics in addition to technology readiness. The main variable of interest in all columns are the technology readiness, interacted with relative income from the initial year. This coefficient has been significant through all columns. Countries involved this regression are listed below. The relationship between technology readiness and 6 dimensions of attitude are graphed in Figure 5. First stage regressions of column (4) is provided in Appendix Table A3. We have also conducted regressions in Table 2 with quadratic term of relative income, and confirmed the relationship between decadal growth and initial income to be negative in the segment of interest. 11 Figure 5. Correlations between Technology Readiness and Attitude Traits Technology Readiness Technology Readiness 0 0 50 50 100 100 150 150 0 20 40 60 80 100 0 20 40 60 80 100 Power Distance Individualism Technology Readiness Technology Readiness 0 0 50 50 100 100 150 150 0 20 40 60 80 100 0 20 40 60 80 100 Indulgence Long-term Orientation Technology Readiness Technology Readiness 0 0 50 50 100 100 150 150 0 20 40 60 80 100 0 20 40 60 80 100 Masculinity Uncertainty Avoidance Source: World Economic Forum, The Global Competitiveness Index dataset 2007-2017; and Hofstede Insights. 12 Note: Technology readiness in y-axis represents the average ranking for technology readiness. A smaller number means a better ranking, and higher technology readiness. The y-axis is reversed. The results from the instrumental regressions using attitude traits as instruments for technology readiness confirm that there is a causal relationship between technology adoption and economic growth. Indeed, Table 2 shows that the individual coefficients associated with technology readiness and interactions with the level of initial income are statically significant and with the expected signs. Due to the lack of complete 6-dimensions of attitude for some countries, regressions in Table 2 are conducted again with only two dimensions, namely long-term orientation and indulgence. The coefficients of the interaction term remain significantly positive, indicating a causal relationship between technology on economic growth (Appendix Table A4). To streamline the instrumentation, we use the first component of the 6-dimensions of the attitude using a principal component analysis in the IV regressions presented in Table 2. The regression table is provided in Appendix Table A5. The first stage regression is provided in Appendix Table A6, and the weights in the first principle component is reported in Appendix Table A7. The results confirm the causal relationship between technology adoption and economic growth. 13 IV. Empirical Evidence for the Slow Pace of Technology Adoption in MENA In this section, we show that MENA’s technology adoption in general purpose industries (GPT) has been poor. We do so within a framework of cross-country panel regressions, specifically contrasting technology adoption between MENA and EAP. The specification is as follows: ℎ −1 −1 −1 = 0 + 1 + 2 + 3 + 4 × + 5 × + + ℎ, ,−1 ,−1 ,−1 ℎ where ℎ captures technology adoption of technology in county at time ; hence ℎ captures , technology adoption relative to the U.S. Technology adoption depends on the country’s development −1 level, proxied by lagged per capita income relative to the U.S. ( ), time fixed effects , region ,−1 fixed effects, and the interaction of the region fixed effects and per capita income relative to the United States. The EAP and MENA fixed effects capture region-specific difference in technology adoption relative to the rest of the world. The interactions capture the speed of technology adoption in MENA and EAP, relative to other countries with the same level of income, as income rises. Technology adoption is proxied by (1) bandwidth per internet user (bits per second), (2) number of self- contained computers designed for use by one person, (3) internet users in percentage of population, (4) number of ATMs per million capita (5) number of payments by credit and debit cards per million capita, (6) tractors used in agriculture per million capita, and (7) gross output of electric energy per million capita. Data are mainly from the CHAT database (Comin and Hobjin, 2010), except that bandwidth and internet users in percentage of population are from World Telecommunication Database (ITU). For MENA, the pace of technology adoption for all technologies as income rises is slower compared to other countries with the same income. The results are shown in Table 3 and illustrated graphically in Figure 6. The coefficients associated with the interaction between the MENA regional dummy and relative income are all negative and statistically significant, which translates into downward-sloping lines for MENA (in red) in Figure 6. For EAP, the coefficients associated with the interaction between the EAP regional dummy and relative income are also negative and significant, but the magnitudes are much smaller than those for MENA. That implies the speed of technology adoption as income rises in EAP is larger than that in MENA. This is shown by the gaps between the blue lines (EAP) and the red lines (MENA) in Figure 6. It is noteworthy that both MENA and EAP have positive and significant coefficients for the region fixed effects (translating into positive intercepts of the blue and red lines in Figure 7). That suggests that at (very) low levels of income, EAP and MENA countries have a faster pace of technology adoption relative 14 to the rest of the world. However, when income rises, that initial advantage quickly fades away because of the lower speed of adoption. Results are robust to using all regional dummies (see Appendix Table A9). Table 3: Technology adoption in MENA and EAP (1) (2) (3) (4) (5) (6) (7) Bandwidth per Computer Per Internet Users Number of ATM Payments by Tractor per Mil Electricity Internet User Mil Capita (%) per Mil Capita Credit and Capita Production per Debit Cards per Mil Capita Mil Capita Relative Income to the US 6.554*** 0.718*** 0.835*** 0.621*** 0.649*** 1.307*** 0.848*** (1.227) (0.0181) (0.0214) (0.0539) (0.0569) (0.0339) (0.0317) EAP 114.2*** 3.716*** 9.477*** 200.6*** 38.39** -2.956* 1.678* (30.67) (1.108) (1.336) (39.10) (15.99) (1.676) (0.871) MNA 149.8*** 7.814*** 11.98*** 60.16* 15.19*** 11.83*** 14.08*** (32.20) (0.653) (1.311) (32.83) (5.540) (0.851) (1.193) EAP * Relative Income -4.136*** 0.0430 -0.361*** -1.951*** -0.416** -0.323*** -0.0633 (1.358) (0.0395) (0.0431) (0.377) (0.166) (0.0859) (0.0396) MNA * Relative Income -6.382*** -0.640*** -0.653*** -1.205*** -0.709*** -1.317*** -0.786*** (1.228) (0.0185) (0.0275) (0.320) (0.0775) (0.0340) (0.0338) Constant -76.26* 14.97* -21.00*** 4.214 -27.73** -25.35*** -0.178 (42.48) (8.378) (1.883) (9.867) (13.66) (5.750) (10.90) Observations 2985 1281 4193 368 372 4115 3151 r2 0.161 0.796 0.673 0.432 0.262 0.536 0.579 * p<0.10 ** p<0.05 *** p<0.01 Note: (1) Bandwidth per internet user (bits per second), (2) number of self-contained computers designed for use by one person, (3) internet users in percentage of population, (4) numbers of ATM per million capita (5) payments by credit and debit cards per million capita, (6) tractors used in agriculture per million capita, (7) Gross output of electric energy per million capita. Time-fixed effects are included in all regressions. See the list of countries included in these regressions presented in Appendix Table A8. 15 Figure 6: Technology adoption in MENA and EAP 16 Generally, there could be several reasons behind to the lack of technology adoption. The literature has identified many factors that could affect a country’s technology adoption, such as human capital (Wozniak, 1987; Benhabib and Spiegel, 2005; Che and Zhang, 2018), trade and FDI (see Keller, 2004 for a review), and competition (Aghion et al, 2005; Seim and Viard, 2011). With a cross-country regression framework, we show that the lack of competition could be one of the reasons behind MENA’s lack of technology adoption. Our measure for market competition is market concentration. The argument is that, comparing within the same industry, countries or regions with higher market concentration tend to have weaker competition (see Berger and Hannan, 1989 and Bikker and Haaf, 2002 in the banking industry, and Sung, 2014 for the telecom industry). Market concentration is widely used to proxy for market competition. The calculation of market concentration indices such as Herfindahl-Hirschman Index (HHI) has been a starting point for assessing the state of market competition (see for example, U.S. Department of Justice and the Federal Trade Commission, 2010). Obviously, market concentration is just one indicator and does not contain all relevant information about competition. However, given our data limitation in cross-country regressions, it is our best choice. In general purpose industries (GPT) such as telecom and finance, there is a high level of market concentration in MENA. Table 4 and the associated graphical illustration in Figure 6 show that market concentration for Mobile Operators and Banking in MENA increases significantly faster as income rises than other countries with the same income (see the red lines in Figure 7) 14. To account for possible non- linearities we include a quadratic term. Figure 7 shows that for mobile operators, while market concentration is smaller in MENA when income is low, it quickly increases with a steep positive slope, while the slope for EAP is negative. For banking, asset concentration for both EAP and MENA is rising faster than the rest of the world, but MENA is above EAP in levels of asset concentration. This evidence is consistent with a popular notion that MENA does not fare well in market competition. For example, according to the World Bank’s Doing Business data, MENA countries are generally ranked very low in starting a business (e.g. Saudi Arabia is ranked 141, Egypt 109, Algeria 150, Iraq 155). 15 Results are robust to using regional dummies for all regions of the world (see Appendix Table A10). 14 For Mobile Operators, market concentration is calculated as annual average of quarterly HHI, based on market share of mobile operators provided by GSMA. For Banking, market concentration is calculated as assets of three largest banks as a share of assets of all commercial banks, data source is World Bank Database on Financial Development and Structure which was first constructed by Beck et al (2000). 15 As Arezki et al. (2018) argue, “MENA governments seeking to protect incumbents, especially in sectors like banking and telecommunications, impose excessive and outdated regulations that deter new actors from entering the market. This short-circuits competition, undermines the diffusion of general-purpose technology, and blocks the type of adaptation and evolution that underpins a vibrant private sector”. 17 Table 4: Market Concentration in Telecom and Finance (1) (2) Mobile Bank Operators Concentration Concentration Relative Income to the US -13.72*** -0.0472*** (1.394) (0.0146) EAP 1477.3*** -6.485*** (161.0) (1.894) MNA -112.6 -1.523 (168.7) (1.559) EAP * Relative Income -5.327* 0.105*** (3.123) (0.0255) MNA * Relative Income 23.21*** 0.0655*** (2.166) (0.0205) Constant 6791.0*** 72.84*** (207.0) (2.118) Observations 3321 2983 r2 0.194 0.0246 * p<0.10 ** p<0.05 *** p<0.01 Note: Time-fixed effects are included in all regressions. Figure 7: Visual illustration for Table 2 18 A large literature has helped document both theoretically and empirically that weak competition is arguably harmful for innovation and productivity growth (see Aghion and Hewitt, 1992 and Aghion et al, 2014). Results from cross-country regressions presented in Table 5 show that given the same level of relative income, mobile and banking concentrations are negatively correlated with technology adoption in the corresponding sector. In other words, a higher level of concentration is associated with lower penetration of the technology. Figure 8 provides a graphical illustration of the results in Table 5. The lines for bandwidth, internet users and ATMs are downward-sloping. For credit and debit card payments, the line slopes upward when bank concentration is very high. However, the confidence band becomes large. To summarize, this section shows that market concentration in GPT such as banking and telecom in MENA becomes higher than other comparators as income rises. This translates to lower adoption of the GPT technologies in MENA. Table 5: Concentration in mobile and banking operators and technology adoption Number of Payments by credit bandwidth per internet users ATMs per mil and debit cards per internet user (%) capita mil capita relative income (t-1) 3.505*** 0.616*** 0.478*** 0.490*** (0.690) (0.0253) (0.0722) (0.0691) mobile concentration -0.0213 -0.00595*** (0.0224) (0.00121) mobile concentration^2 0.000000682 0.000000167* (0.00000167) (9.10e-08) bank concentration -1.198** -2.410*** (0.547) (0.781) bank concentration^2 0.00649 0.0181*** (0.00425) (0.00657) Constant 66.30 31.01*** 73.31*** 73.07*** (79.92) (4.211) (19.88) (24.41) Observations 2464 2658 178 182 year fixed effects y y y y r2 0.108 0.621 0.385 0.344 19 Figure 8: Visualization for Table 3 V. Conclusion This paper documented the existence of a “middle-income trap” for the Middle East and North Africa region (MENA). It argued that MENA economic woes offer new insights into the debate on the trap, which has thus far focused on the East Asia and Pacific region (EAP). The results are two-folds. First, non-parametric regressions show that the average rate of economic growth in MENA has not only been significantly lower than EAP but has also tended to drop at an earlier level of income. Second, a slower pace of technology adoption is associated with slower levels of economic growth and MENA has experienced a relatively slow pace of technology adoption in general purpose technologies (GPT). These results suggest that barriers to GPT adoption constitute an important channel of transmission for the middle-income trap. Indeed, the pervasive lack of market contestability in MENA markets and the resulting slow pace of technology adoption including in key sectors can help explain why more generally economies tend to get stuck. To the extent that governments play a key role in the regulation of entry including in key “upstream” sectors, the literature focus on firm level dynamics only shed lights on 20 “downstream” matters. Instead, the lack of availability of frontier GPT can seclude firms into low productivity activities, limiting trade and economic growth. Further research on the interplay between the causes and consequences of lack of (government induced) GPT adoption would help understand the nature and consequences of upstream factors impeding productivity gains and growth. From a policy perspective, one proposal put forward by Arezki et al (2018) to break with “business as usual” in the MENA region is for the authorities to embrace a “moonshot approach” to the adoption of information technology and communications. MENA countries could emulate President John F. Kennedy’s 1961 decision to unleash an extraordinary collective national effort that achieved its seemingly impossible goal: a manned lunar landing in mid-1969. A MENA moonshot would involve a collective regional commitment to achieve parity with advanced economies in information and communications technology by 2021. MENA countries would seek to equal or better OECD countries in terms of their level of access to the internet, capacity to transmit data (bandwidth) and the number of financial transactions carried out electronically. This would unleash the potential of the young and educated population—who have been subject to abnormally high levels of unemployment—and spur growth. 16 16 World Bank (2019) shows that MENA has the highest youth unemployment rates in the world and these rates are highest among the educated. 21 References Akcigit, Ufuk & Salomé Baslandze & Francesca Lotti, 2018. "Connecting to Power: Political Connections, Innovation, and Firm Dynamics," NBER Working Papers 25136, National Bureau of Economic Research, Inc. Acemoglu, D., Verdier, T., 2000. The choice between market failures and corruption. American Economic Review 90 (1), 194–211. Aghion, Philippe, Ufuk Akcigit and Peter Howitt, 2014. "What Do We Learn From Schumpeterian Growth Theory?," Handbook of Economic Growth, edition 1, volume 2, chapter 0, pages 515-563 Aghion, Philippe, Nick Bloom, Richard Blundell, Rachel Griffith and Peter Howitt, 2005. “Competition and Innovation: An Inverted-U Relationship” Quarterly Journal of Economics, 120, 701-728. Aghion, Philippe and Peter Howitt, 1992. A Model of Growth Through Creative Destruction., Econometrica, 60, 323-351. Arezki, Rabah, Lili Mottaghi, Andrea Barone, Rachel Yuting Fan, Amani Abou Harb, Omer M. Karasapan, Hideki Matsunaga, Ha Nguyen, and Francois de Soyres, 2018. “A New Economy in Middle East and North Africa” Middle East and North Africa Economic Monitor (October), World Bank, Washington, DC. Ades, A., Di Tella, R., 1999. Rents, competition, and corruption. American Economic Review. 89 (4),982–993. Aiyar, Shekhar, Romain Duval, Damien Puy, Yiqun Wu, and Longmei Zhang, 2013. “Growth Slowdowns and the Middle-Income Trap”, IMF Working Paper WP/13/71 Barro, Robert, 2016 “Economic Growth and Convergence, Applied to China” China and The World Economy, Vol. 24, pp 5-19 Benhabib, J. and Spiegel, M.M, 2005. “Human capital and technology diffusion”, in (P. Aghion and S.N. Durlauf, eds), Handbook of Economic Growth, pp. 936–66, Amsterdam: North Holland. Berger, Allen N., and Timothy H. Hannan, 1989. "The Price-Concentration Relationship in Banking." Review of Economics and Statistics 71, 291-2 Bikker, Jacob and Katharina Haaf, 2002, “Competition, concentration and their relationship: An empirical analysis of the banking industry”, Journal of Banking & Finance, Volume 26, Issue 11, pp 2191-2214. 22 Bulman, David, Maya Eden and Ha Nguyen, 2017. “Transitioning from low-income growth to high- income growth: is there a middle-income trap?” Journal of the Asia Pacific Economy, volume 22, issue 1, pp 5-28 Beck, Thorsten, Aslı Demirgüç-Kunt and Ross Levine, 2000. "A New Database on Financial Development and Structure"", World Bank Economic Review 14, 597-605 Che, Yi and Lei Zhang, 2018 “Human capital, Technology Adoption and Firm Performance: Impacts of China’s Higher Education Expansion in the Late 1990s” The Economic Journal 128:614, 2282-2320 Comin, Diego and Bart Hobjin, 2010 “An Exploration of Technology Diffusion” American Economic Review, vol 100: 2031-2059 Diwan, Ishac and Marc Schiffbauer, 2018 “Private banking and crony capitalism in Egypt” Business and Politics, vol 20, pp 390-409 Djankov, Simeon & Rafael La Porta & Florencio Lopez-De-Silanes & Andrei Shleifer, 2002. "The Regulation Of Entry," The Quarterly Journal of Economics, MIT Press, vol. 117(1), pages 1-37, February. Eichengreen, Barry & Donghyun Park & Kwanho Shin, 2013. "Growth Slowdowns Redux: New Evidence on the Middle-Income Trap," NBER Working Papers 18673 Eichengreen Barry, Donghyun Park, and Kwanho Shin, 2012 “When Fast-Growing Economies Slow Down: International Evidence and Implications for China” Asian Economic Papers, pp 42-87 Fan, Jianqing and Irene Gijbels. 1996. Local Polynomial Modelling and Its Applications. London: Chapman and Hall Flaaen, Aaron; Ghani, Ejaz; Mishra, Saurabh, 2013. How to Avoid Middle-Income Traps? : Evidence from Malaysia. Economic Premise; no. 120. World Bank, Washington, DC Han Xuehui and Shang-Jin Wei, 2017. Re-examining the middle-income trap hypothesis (MITH): What to reject and what to revive? Journal of International Money and Finance, Volume 73, Pages 41-61. Hofstede, Geert, Gert Jan Hofstede and Michael Minkov (2010) “Cultures and Organization: Software of the Mind : intercultural cooperation and its importance for survival”, 3rd edition Gill, Indermit, Homi Kharas and Others, 2007. “An East Asian Renaissance: Ideas for Economic Growth.” World Bank, Washington, DC. Gill,Indermit S. & Kharas,Homi, 2015. "The middle-income trap turns ten," Policy Research Working Paper Series 7403, The World Bank 23 Glawe, L. and H. Wagner, 2017. “The People’s Republic of China in the Middle-Income Trap?” ADBI Working Paper No. 749. Tokyo: Asian Development Bank Institute (ADBI). Global Times, 2015. China may hit middle-income trap: minister, 4/26/2015 Keller, Wolfgang, 2004. “International Technology Diffusion”. Journal of Economic Literature, 42(3), 752-782. Kharas, Homi and Harinder Kohli, 2011. “What is the Middle Income Trap, Why do Countries Fall into It, and How Can It Be Avoided?” Global Journal of Emerging Market Economies, Volume 3, No. 3: 281- 289. Ohno, Kenichi and Le Ha Thanh, 2015 “Bẫy thu nhập trung bình tại Việt Nam: thực trạng và giải pháp” Tạp chí Khoa học xã hội Việt Nam, vol 7 (92), pp 31-47 Parente, Stephen., and Edward Prescott. 1994. “Barriers to Technology Adoption and Development” Journal of Political Economy, 102(2), 298-321. Rijkers, Bob & Freund, Caroline & Nucifora, Antonio, 2017. "All in the family: State capture in Tunisia," Journal of Development Economics, Elsevier, vol. 124(C), pages 41-59. Seim, K., & Viard, V., 2011. “The Effect of Market Structure on Cellular Technology Adoption and Pricing”. American Economic Journal: Microeconomics, 3(2), 221-251. Shleifer, A., Vishny, R.W., 1993. Corruption. Quarterly Journal of Economics 108 (3), 599–617. Shleifer, A., Vishny, R.W., 1994. Politicians and firms. Quarterly Journal of Economics 109 (4), 995– 1025. Sung, Nakil, 2014 “Market concentration and competition in OECD mobile telecommunications markets” Applied Economics, 46:25, 3037-3048 Temple, Jonathan. 1999. “The New Growth Evidence” Journal of Economic Literature, 37(1) pp 112-156 U.S. Department of Justice and the Federal Trade Commission. 2010. Horizontal Merger Guidelines. Available at http:// www.justice.gov/atr/public/guidelines/hmg-2010.html World Bank, 2019. “Expectations and Aspirations: A New Framework for Education in the Middle East and North Africa.” Overview booklet. World Bank, World Economic Forum. The Global Competitiveness Index dataset 2007-2017. Wozniak, Gregory D. 1987. “Human Capital, Information, and the Early Adoption of New Technology.” Journal of Human Resources 22:101-112 24 Yousef, Tarik, M, 2004. "Development, Growth and Policy Reform in the Middle East and North Africa since 1950."Journal of Economic Perspectives, 18 (3): 91-115. 25 Table A1: Coefficients for the non-parametric regressions: Relative growth versus relative income EAP Average Percentile relative Predicted Std.Err z P>|z| income [95% Conf. Growth Interval] 0 0.021 0.001 14.4 0 0.018 0.021 10 0.022 0.001 19.76 0 0.021 0.023 20 0.026 0.002 16.71 0 0.025 0.028 30 0.026 0.001 28.82 0 0.025 0.028 40 0.025 0.001 25.62 0 0.024 0.026 50 0.023 0.001 22.65 0 0.022 0.024 60 0.016 0.001 14.9 0 0.013 0.016 70 0.008 0.001 11.25 0 0.006 0.008 80 0.004 0 8.88 0 0.003 0.004 90 0.006 0.002 3.89 0 0.004 0.009 100 0.009 0.003 2.96 0.003 0.003 0.011 MENA Average Percentile relative Predicted Std.Err z P>|z| income [95% Conf. Growth Interval] 0 0.024 0.003 8.123 0 0.018 0.029 10 0.016 0.002 6.683 0 0.011 0.02 20 0.01 0.002 4.693 0 0.007 0.015 30 0.006 0.002 2.509 0.012 0.002 0.011 40 0.002 0.002 0.869 0.385 -0.002 0.007 50 -0.001 0.003 -0.242 0.809 -0.005 0.004 60 -0.004 0.003 -1.241 0.215 -0.009 0.001 70 -0.007 0.003 -2.117 0.034 -0.013 -0.002 80 -0.011 0.004 -2.775 0.006 -0.018 -0.004 90 -0.015 0.004 -3.497 0 -0.023 -0.007 100 -0.02 0.004 -4.784 0 -0.028 -0.013 26 Figure A1: Growth in TFP Panel A: Relative TFP Growth (to the U.S.) East Asia and Pacif ic Middle East and North Africa -.05 -.04 -.03 -.02 -.01 0 .01 .02 .03 .04 .05 -.05 -.04 -.03 -.02 -.01 0 .01 .02 .03 .04 .05 relative growth in TFP relative growth in TFP 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 relative income to the U.S., percentage relative income to the U.S., percentage Panel B: Absolute TFP growth East Asia and Pacif ic Middle East and North Af rica -.05 -.04 -.03 -.02 -.01 0 .01 .02 .03 .04 .05 -.05 -.04 -.03 -.02 -.01 0 .01 .02 .03 .04 .05 absolute growth in TFP absolute growth in TFP 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 relative income to the U.S., percentage relative income to the U.S., percentage 27 Figure A2: Relative GDP growth for other regions Europe and Central Asia Latin America and the Caribbean -.01 0 .01 .02 .03 .04 .05 relative growth in income relative growth in income -.01 0 .01.02.03.04.05 -.05-.04-.03-.02 -.02 -.03 -.04 -.05 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 relative income to the U.S., percentage relative income to the U.S., percentage South Asia Sub-Saharan Af rica -.05-.04-.03-.02-.01 0 .01 .02 .03 .04 .05 relative growth in income relative growth in income -.05-.03-.01.01 .03 .05 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 relative income to the U.S., percentage relative income to the U.S., percentage Note: All regions are defined following World Bank country groups 17. 17 See http://databank.worldbank.org/data/download/site-content/CLASS.xls for the current classification. 28 Table A2. List of countries Albania Denmark Latvia Qatar Dominican Algeria Republic Lesotho Romania Argentina Ecuador Lithuania Russia Armenia Egypt Luxembourg Saudi Arabia Australia El Salvador Macedonia, FYR Senegal Austria Estonia Madagascar Serbia Azerbaijan Ethiopia Malaysia Singapore Bahrain Finland Mali Slovak Republic Bangladesh France Mauritania Slovenia Barbados Gambia, The Mauritius South Africa Belgium Georgia Mexico Spain Benin Germany Mongolia Sri Lanka Bolivia Greece Montenegro, Rep. of Sweden Bosnia and Herzegovina Guatemala Morocco Switzerland Botswana Honduras Mozambique Syria Brazil Hong Kong SAR Namibia Tajikistan Bulgaria Hungary Nepal Tanzania Burkina Faso Iceland Netherlands Thailand Burundi India New Zealand Trinidad and Tobago Cambodia Indonesia Nicaragua Tunisia Cameroon Ireland Nigeria Turkey Canada Italy Norway Uganda Chad Jamaica Oman Ukraine Chile Japan Pakistan United Arab Emirates China Jordan Panama United Kingdom Colombia Kazakhstan Paraguay United States Costa Rica Kenya Peru Uruguay Croatia Korea Philippines Venezuela Cyprus Kuwait Poland Vietnam Czech Republic Kyrgyz Republic Portugal Zambia Zimbabwe Note: The table presents countries included in Table 1. 29 Table A3. First stage regression of Column (4) in Table 2. (1) (2) Average technology Relative income x Average readiness (-) technology readiness (-) Power Distance 0.0279 0.0618*** (0.0381) (0.0108) Individualism 0.783*** -0.0757*** (0.0322) (0.00912) Masculinity -0.563*** -0.0604*** (0.0361) (0.0102) Uncertainty Avoidance 0.161*** 0.00136 (0.0251) (0.00712) Long-term Orientation 0.733*** -0.00253 (0.0240) (0.00680) Indulgence 0.568*** 0.0195*** (0.0250) (0.00709) Power Distance x Relative income -0.301*** -0.343*** (0.0847) (0.0240) Individualism x Relative income -1.112*** 0.337*** (0.0623) (0.0177) Masculinity x Relative income 0.567*** -0.0726*** (0.0626) (0.0177) Uncertainty Avoidance x Relative income -0.653*** -0.264*** (0.0562) (0.0159) Long-term Orientation x Relative income -0.695*** 0.384*** (0.0498) (0.0141) Indulgence x Relative income -1.347*** -0.238*** (0.0732) (0.0208) Relative income x Relative income 247.5*** 2.194 (9.711) (2.754) Observations 2794 2794 Note: Standard errors are given in parentheses. * p<0.1, ** p<0.05, *** p<0.01. Coefficients of constants are not reported to save space. 30 Table A4. Growth and technology, OLS and IV regression (2 dimensions) Relative decadal growth (1) (2) (3) (4) (5) (6) OLS IV OLS IV OLS IV Relative income -0.0605*** -0.0727*** -0.0609*** -0.0639*** -0.0245*** -0.109*** (0.00344) (0.00537) (0.00309) (0.00474) (0.00828) (0.0144) Relative income x Average technology readiness (-) 0.000720*** 0.000619*** 0.000603*** 0.000562*** 0.00325*** 0.000900** (0.0000768) (0.000147) (0.0000689) (0.000130) (0.000192) (0.000381) Average technology readiness (-) 0.000492*** 0.000602*** 0.000498*** 0.000525*** (0.0000255) (0.0000457) (0.0000229) (0.0000402) Observations 3726 3726 3726 3726 3726 3726 Year fixed effect - - yes yes yes yes Country fixed effect - - - - yes yes R-square 0.188 0.184 0.361 0.361 0.496 0.475 First-stage F-stat 222.8 221.2 613.3 First-stage Sargan-test 9.674 10.42 29.15 Notes: Coefficient estimates from ordinary least squares regressions at the country-year level. Standard errors are given in parentheses. * p<0.1, ** p<0.05, *** p<0.01. The dependent variable is the relative annualized decadal growth of real GDP per capita, compared to the growth in the US. Relative income is the relative real GDP per capita from the initial year of the decade (US’s real GDP per capita at the same year equals 1). Average technology readiness in the regression represents the average ranking for technology readiness. A higher number means a better ranking, and higher technology readiness. This variable is instrumented in column (2) (4) (6), by 2-dimensions of country specific attitudes, namely long-term orientation, and indulgence. First stage F-stat and Sargan test for over- identification are both reported. Regressions in all columns have the same sample to ease comparison. The coefficient estimates on constant are not reported to save space. Column (1) and (2) has no fixed effects, while column (3) and (4) is added with year fixed effects. In column (5) and (6), we added country fixed effect to replace the linear term of average technology readiness, in order to capture country specific characteristics in addition to technology readiness. The main variable of interest in all columns are the technology readiness, interacted with relative income from the initial year. This coefficient has been significant through all columns. Countries involved this regression are listed below. 31 Table A5. Technology and growth, OLS and IV (first principle component) Relative decadal growth (1) (2) (3) (4) (5) (6) OLS IV OLS IV OLS IV Relative income -0.0525*** -0.0315*** -0.0558*** -0.0534*** -0.0654*** 0.0776*** (0.00297) (0.0104) (0.00266) (0.0101) (0.00656) (0.0261) Relative income x Average technology readiness (-) 0.000691*** 0.000380 0.000662*** 0.000865*** 0.00275*** 0.00776*** (0.0000756) (0.000248) (0.0000672) (0.000236) (0.000174) (0.000900) Average technology readiness (-) 0.000383*** 0.000150 0.000407*** 0.000386*** (0.0000262) (0.000114) (0.0000234) (0.000111) Observations 2794 2794 2794 2794 2794 2794 Country fixed effect - - - - yes yes Year fixed effect - - yes yes yes yes R-square 0.201 0.151 0.382 0.380 0.555 0.416 First-stage F-stat 31.60 23.44 131.2 Notes: Coefficient estimates from ordinary least squares regressions at the country-year level. Standard errors are given in parentheses. * p<0.1, ** p<0.05, *** p<0.01. The dependent variable is the relative annualized decadal growth of real GDP per capita, compared to the growth in the US. Relative income is the relative real GDP per capita from the initial year of the decade (US’s real GDP per capita at the same year equals 1). Average technology readiness in the regression represents the average ranking for technology readiness. A higher number means a better ranking, and higher technology readiness. This variable is instrumented in column (2) (4) (6), by the first principle component of the 6 dimensions of attitude. First stage F-stat is reported. Regressions in all columns have the same sample to ease comparison. The coefficient estimates on constant are not reported to save space. Column (1) and (2) has no fixed effects, while column (3) and (4) is added with year fixed effects. In column (5) and (6), we added country fixed effect to replace the linear term of average technology readiness, in order to capture country specific characteristics in addition to technology readiness. The main variable of interest in all columns are the technology readiness, interacted with relative income from the initial year. First stage regression of column (4) is provided in Appendix Table A6, and the weights in the first component is reported in Appendix Table A7. 32 Table A6. First stage regression of column (4) in Table A5 (1) (2) Average technology Relative income x Average readiness (-) technology readiness (-) Attitude -9.434*** 2.189*** (0.703) (0.237) Attitude x Relative income 16.19*** -6.931*** (1.197) (0.404) Relative income 91.99*** -8.251*** (1.947) (0.657) Constant -73.92*** -8.618*** (3.235) (1.092) Observations 2794 2794 Table A7. Principle Component of Attitude Principle Component Power Distance 0.6 Individualism -0.6 Masculinity 0.1 Uncertainty Avoidance 0.3 Long-term Orientation 0.2 Indulgence -0.4 33 Table A8: List of countries United States Guatemala Vietnam Sudan United Kingdom Haiti Algeria Swaziland Austria Honduras Angola Tanzania Belgium Mexico Botswana Togo Denmark Nicaragua Burundi Tunisia France Panama Cameroon Uganda Germany Paraguay Central African Republic Burkina Faso Italy Peru Chad Zambia Congo, Democratic Netherlands Uruguay Republic of the Armenia Norway Belize Benin Azerbaijan Sweden Suriname Equatorial Guinea Belarus Switzerland Iran Ethiopia Albania Canada Jordan Gabon Georgia Japan Kuwait Ghana Kazakhstan Finland Lebanon Guinea-Bissau Bulgaria Greece Oman Guinea Moldova Iceland Saudi Arabia Kenya Russia Ireland Syria Lesotho Tajikistan Portugal United Arab Emirates Liberia China Spain Egypt Madagascar Turkmenistan Turkey Yemen Malawi Ukraine Australia Bangladesh Mali Uzbekistan New Zealand Cambodia Mauritania Czech Republic South Africa Sri Lanka Mauritius Slovak Republic Argentina India Morocco Estonia Bolivia Indonesia Mozambique Latvia Brazil Korea Niger Hungary Chile Malaysia Nigeria Lithuania Colombia Nepal Zimbabwe Mongolia Costa Rica Pakistan Rwanda Croatia Dominican Republic Philippines Senegal Slovenia Ecuador Singapore Sierra Leone Poland El Salvador Thailand Namibia Romania Note: The table presents countries included in Table 3. 34 Table A9: Estimating Technology Adoption with Regional Dummies (1) (2) (3) (4) (5) (6) (7) Bandwidth Computer Internet Number of Payments Tractor Electricity per internet per mil users (%) ATM per by credit per mil production user capita mil capita and debit capita per mil cards per capita mil capita Relative income -3.618*** 2.220*** 0.387 -1.072*** 0.547** -1.847*** -2.979*** (0.818) (0.215) (0.279) (0.231) (0.266) (0.143) (0.174) EAP -435.4*** 118.4*** -48.61* 8.481 -24.39 -295.0*** -405.9*** (80.72) (20.76) (25.21) (44.27) (29.85) (13.63) (17.35) ECA -792.3*** 113.2*** -44.66* -191.1*** -57.90** -249.8*** -406.8*** (123.2) (20.78) (25.21) (21.64) (25.45) (14.02) (17.70) LAC -374.4*** 120.0*** -51.10** -293.7*** -403.0*** (85.68) (20.76) (25.20) (13.58) (17.37) MNA -444.1*** 122.6*** -47.76* -132.7*** -46.98* -279.9*** -394.4*** (80.24) (20.74) (25.20) (37.24) (25.79) (13.58) (17.40) SAR -452.5*** 121.6*** -62.83** -287.4*** -401.7*** (80.33) (20.79) (25.23) (13.69) (17.39) SSA -449.5*** 121.2*** -60.60** -288.0*** -401.1*** (80.14) (20.77) (25.18) (13.57) (17.38) EAP x Relative income 6.038*** -1.454*** 0.0865 -0.259 -0.311 2.837*** 3.753*** (1.002) (0.218) (0.282) (0.437) (0.309) (0.162) (0.175) ECA x Relative income 14.05*** -1.510*** 0.296 1.628*** -0.0432 2.733*** 3.778*** (2.290) (0.217) (0.281) (0.234) (0.268) (0.155) (0.186) LAC x Relative income 3.616*** -1.840*** 0.243 2.638*** 3.316*** (0.974) (0.218) (0.282) (0.149) (0.175) MNA x Relative income 3.812*** -2.142*** -0.204 0.494 -0.614** 1.836*** 3.041*** (0.819) (0.215) (0.280) (0.376) (0.272) (0.143) (0.174) SAR x Relative income 4.654*** -2.064*** 0.474 2.356*** 3.208*** (0.932) (0.287) (0.304) (0.306) (0.220) SSA x Relative income 4.117*** -2.026*** -0.0954 2.105*** 3.187*** (0.855) (0.220) (0.286) (0.146) (0.177) Observations 2967 1281 4170 368 372 4115 3151 R-square 0.219 0.839 0.709 0.545 0.476 0.611 0.671 Note: This table reports all regional dummies with North America as the default region. Standard errors are given in parentheses. * p<0.1, ** p<0.05, *** p<0.01. Time-fixed effects are included in all regressions. Coefficients of constants are not reported to save space. 35 Table A10: Market Concentration in Telecom and Finance (1) (2) Mobile operators Concentration Bank concentration Relative income -33.65 -1.831*** (20.52) (0.223) EAP 1328.9 -147.9*** (1834.6) (21.49) ECA -1088.4 -146.1*** (1830.8) (21.45) LAC -272.5 -144.6*** (1837.4) (21.48) MNA -211.6 -144.3*** (1835.7) (21.46) SAR -1835.0 -164.0*** (1852.6) (21.76) SSA 13.51 -135.7*** (1830.6) (21.43) EAP x Relative income 14.63 1.890*** (20.71) (0.224) ECA x Relative income 29.33 1.870*** (20.59) (0.224) LAC x Relative income 34.90 1.757*** (21.75) (0.231) MNA x Relative income 43.11** 1.852*** (20.59) (0.223) SAR x Relative income 119.8*** 3.063*** (28.16) (0.450) SSA x Relative income 53.61** 1.929*** (21.08) (0.235) Observations 3303 2962 R-square 0.248 0.111 Note: This table reports all regional dummies with North America as the default region. Standard errors are given in parentheses. * p<0.1, ** p<0.05, *** p<0.01. Time-fixed effects are included in all regressions. Coefficients of constants are not reported to save space. 36