University of Amsterdam Faculty of Economics and Business A thesis presented for the degree of Master of Science in Finance

(1)

University of Amsterdam

Faculty of Economics and Business

A thesis presented for the degree of Master of Science in Finance

Stock market concentration and the real economy. The role of superstar

technology firms and their impact on the real economy.

Supervisor:

dr. Esther Eiling

Candidate:

Calvin Cheng Peng Li

August 14, 2022

Academic year 2021/2022

1

(2)

Statement of Originality

This document is written by Student Calvin Li who declares to take full responsibil- ity for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

Technology firms have a noticeable impact on today’s society. They facil- itate education, research, entertainment and much more. Technology firms’

market capitalization are much greater compared to non-technology firms in the US. Using data from 18 countries over the last thirty years, this paper shows that market concentration of technology firms negatively impacts future annual per capita GDP growth, whereas market concentration of non-technology firms positively impact future annual per capita GDP growth. However, these findings depend on the highly concentrated countries in the sample. Market concentration of non-technology/all firms, on the other hand, remain significantly positive, opposing current financial literature that stock market concentration obscures funding, competition, innovation, and economic growth.

JEL classifications: G15, O14, O33

Keywords: Stock market concentration, Technological progress, Economic growth.

3

(4)

List of Tables

1 Summary statistics for market capitalization by GDP, per capita GDP growth per country and macroeconomic variables, 1979-2021 annually 15 2 Summary statistics for market concentration variables . . . 18 3 Pearson correlation matrix . . . 22

(5)

4 Panel regression of per capita GDP growth on stock market concentration of technology firms. . . 26 5 Panel regression of per capita GDP growth on stock market concentra-

tion of technology firms excluding Finland, Great Britain, and Portugal. 28 6 SIC Codes . . . 32

List of Figures

1 Average stock market concentration . . . 19 2 Market concentration top 5 firms . . . 35 3 Market concentration of top 10 firms . . . 36

5

(6)

1 INTRODUCTION

1 Introduction

In the recent decades, technology firms soared through the stock markets. Starting with the dot-com bubble in the early 2000s, technology firms have gained tremen- dous interest, initially with Microsoft, followed by companies such as Intel, IBM and Apple. At the beginning of 2022, Apple was the first company in the world to reach a market cap of$3 trillion USD. As of the moment of writing, technology firms in the US account for the vast majority in the top ten, Berkshire Hathaway, United Health, and Johnson&Johnson being the only non-technology companies in the top ten ¹. Moreover, it is quite fascinating that a company’s yearly revenue is greater than the total output of several developed countries. Given the market capitalization, it can be easy to say that these large companies account for a relative large amount of the current workforce, however, this fails to be true. Autor, Dorn, Katz, Patterson, and Van Reenen (2020) find that high market value firms have relatively low share of employment.

How do large technology firms affect stock market concentration and the real economy? This question can be very beneficial for policymakers as the rise in market power happens at a rapid pace which exceeds their ability to regulate new devel- opments. I hypothesize that a highly concentrated stock market, even more in the presence of technology firms, will lead to a negative impact on future per capita GDP growth as efficient (human) capital allocation will be disrupted by large (technology) firms.

This paper contributes to the current financial literature by analyzing the relation-

1Largest capitalizations as of August 2022: 1- Apple ($2.72T), 2- Microsoft ($2.16T), 3- Alphabet ($1.56T), 4- Amazon ($1.45T), 5- Tesla ($0.92T), 6- Berkshire Hathaway ($0.65T), 7- United Health ($0.50T), 8- Meta ($0.48T), 9- NVIDIA ($0.45T), 10- Johnson&Johnson ($0.45T)

(7)

1 INTRODUCTION

ship between stock market concentration and economic growth, which relatively few studies have investigated and adds value by extending the previously used models to examine the differential effect of technology and non-technology firms.

I test the effect of stock market concentration across 18 countries over the period 1990 to 2021. Over this period I find that stock market concentration positively impacts future per capita GDP growth, in contrast to existing studies. Existing studies clearly show a negative relationship between future per capita GDP growth. The contrast may be due to a different sample selection and different data. Stock market concentration is defined as the sum of the largest five (ten) market capitalization firms divided by the total market capitalization. In the light of King and Levine (1993) , I regress per capita GDP growth in year t on the market concentration of technology firms in year t-5 and the market concentration of non-technology firms in year t-5. Technology firms are defined using the Standard Industrial Classification (SIC) codes in accordance to the United States Census Bureau. I find that the stock market concentration for technology firms is significantly negative and estimate a coefficient of -2.34, indicating a 2.34% decrease in annual per capita GDP growth in the future. Moreover, I find a positive relationship between stock market concentration of non-technology firms with an estimate of 1.65, indicating an increase of 1.65% in future annual per capita GDP growth. However, the power of the estimates does not hold when highly concentrated countries are excluded. The estimates of technology firms decreases in significance, leading to unreliable measures. On the other hand, the estimates of non-technology firms and all firms remain significant, indicating a positive relationship between stock market concentration and future per capita GDP growth.

This paper will proceed in the following way. Section 2 examines the current literature on stock market concentration and this impact on economic growth. Sections 3

7

(8)

1 INTRODUCTION

and 4 describe the regression model and data used in examining the effect of market concentration on growth. Section 5 shows the results and discusses the implications and section 6 concludes this paper.

(9)

2 LITERATURE REVIEW

2 Literature Review

Stock markets should, in theory, fund new firms, increase competition, innovation and economic growth. However, Bae, Bailey, and Kang (2021) find that high stock market concentration, defined as the top 5 market capitalization firms to total market capitalization ratio, per country, decreases economic growth. In these countries, stock markets are dominated by a small number of highly successful firms, which lead to less efficient capital allocation, stagnant IPOs, and innovation. First, the authors regress real per capita GDP growth rates in year t on the stock market concentration in year t-5. The authors find that a 1 standard deviation increase in stock market concentration which implies a 0.75 percentage point decrease in GDP growth rate in 5 years. Second, the authors regress investment in an industry for each country on the growth rate of value added in that industry to determine an estimate of the degree of capital allocation efficiency. Using cross-sectional regression methods, Bae et al. (2021) find that stock market concentration and capital allocation efficiency are negatively correlated, suggesting that a concentrated stock market is less likely to allocate capital to firms that make more efficient use of capital. Thirdly, panel regressions indicate that high stock market concentration is negatively associated with IPOs and innovation proxies.

Moreover, the current superstar firms contribute relatively less to the real economy compared past superstar firms. Guti´errez and Philippon (2019a) find that labor productivity contribution to the US has decreased by about 40% since the turn of the millennium. The reason as to why is still unclear. Bloom, Jones, Van Reenen, and Webb (2020) explain that ideas are harder to come up with. Another reason is that declining competition and rising barriers to entry allowed incumbent firms to reduce innovation and investment as proposed by Guti´errez and Philippon (2017).

9

(10)

2 LITERATURE REVIEW

Furthermore, Guti´errez and Philippon (2019b) find that free entry starts to break down around 2000. Davis (2017) argues that entry barriers stem from complex reg- ulations. This is supported by Guti´errez and Philippon (2019a) as they argue that large firms lobby to defend themselves from competition. Reduced competition in- centivizes reduced investment and innovation. Schlingemann and Stulz (2020) find supporting evidence. The authors show that there is a strong downward trend in the extent to which a firm’s market capitalization reflect its contribution to employment.

The authors use employment rates as it is a reliable benchmark for the state of the economy. It does not depend on accounting rules and is collected by an independent party, irrespective of listing. The authors find that employment share of listed firms sharply falls in the 80s.

The current high market capitalization of technology firms may be related to technological revolutions. P´astor and Veronesi (2009) develop a general equilibrium model to explain stock price bubbles during technological revolutions. This model compares two sectors: “new economy” and “old economy”. The old economy implements existing technologies in large-scale production whose output determines wealth. The new economy implements new technology in small-scale production that does not affect wealth. The authors find that stock price bubbles are stronger in the new economy than in the old economy. P´astor and Veronesi (2009) use the NASDAQ index as a proxy for the new economy and the NYSE/AMEX index as a proxy for the old economy. The authors find that for the internet bubble, NASDAQ’s beta doubled and volatility tripled for the 1997-2002 period, while NYSE/AMEX’s volatility doubled during the same period. Greenwood and Jovanovic (1999) find that NASDAQ firms’

(tech firms) market capitalization sharply rise post 1968 and incumbent firms fared badly.

(11)

2 LITERATURE REVIEW

nomic growth and high-tech industry. The authors define high-technology sectors using the SIC codes 28xx, 35xx, 36xx, and 38xx. Using a lagged overlapping rolling average approach, the authors find that economic growth and the initial size of a country’s high-tech industry have a significant positive relationship. In addition, the authors measure the economic magnitude of high-tech value added, as the change in per capita GDP growth if, ceteris paribus, a country moves from the 25th percentile HT value added to the 75th percentile. Brown et al. (2017) calculate a magnitude of 0.4, indicating a 0.4 percentage point faster GDP growth rate.

Summarizing, past literature shows that highly concentrated markets disrupt efficient stock markets. Combined with the fact that technology has a positive effect on GDP growth and that the presence of technology firms disrupts stock markets more compared to non-technology firms, I expect that a concentrated stock market with firms in the technology sector will lead to a reduction in future per capita GDP growth.

11

(12)

3 METHODOLOGY

3 Methodology

In order to examine the relationship between the market concentration of technology firms and economic growth in 18 different countries, I regress GDP per capita in country c and year t on the market concentration of technology firms in the top 5(10) technology firms are classified using SIC codes, in accordance to the United States Census Bureau.² The Bureau determines high-technology firms in order to analyze labour markets correctly and consequently forecast high-technology industry’s contribution to the US economy. Simply regressing GDP per capita on market concentration of technology firms cannot yield a causal relationship. Therefore, I add several financial and macroeconomic indicators as control variables that the literature shows affect economic growth. These variables include market capitalization, economic openness, which is the sum of imports and exports of goods and services, domestic credit provided by the government, investment, turnover, which is the total value of domestic shares traded divided by total market capitalization, government expenditure, and inflation. More importantly, credit, turnover, market concentration, and market capitalization are lagged to address reverse causality bias as higher levels of per capita GDP growth might result in higher market activity which corre- lates to market capitalization and market concentration. Furthermore, using lagged values, as done by Bekaert, Harvey, and Lundblad (2011), allows us to determine the long-term effects of stock market concentration on economic growth. Using this knowledge, I specify the following regression equation:

2Specific SIC codes can be found in the appendix. Walcott (2000) provides a clear overview of different SIC codes used by various sources.

(13)

3 METHODOLOGY

GDP per capita_c,t= β₀+ β₁M arket Concentration(top5(10)(N on)T echf irms_c,t−5 + β₂M arket Capitalization/GDP_c,t−5

+ β₃Credit/GDP_c,t−5+ β₄Inf lation_c,t

+ β₅Openness/GDP_c,t+ β₆T urnover/GDP_c,t−5 + β₇Investment/GDP_c,t

+ β₈Government Expenditure/GDP_c,t+ ϵ_c,t.

In estimating the regression equation, I hypothesize that market concentration of technological firms will have a significant negative effect on GDP per capita as a highly concentrated stock market, specifically, large successful technology firms disturb the efficient allocation of capital throughout the economy as proposed by Bae et al. (2021) and Autor et al. (2020).

13

(14)

4 DATA

4 Data

Estimating the model described previously, requires data on economic growth, market concentration and financial and macroeconomic control variables. In this section, I describe economic growth, market concentration and control variables, in addition to their summary statistics and correlation amongst each other.

4.1 Economic growth

Economic growth data is not readily available as it is difficult to determine. To proxy economic growth, I use annual per capita GDP growth in real terms. The data is readily available from Jord`a, Knoll, Kuvshinov, Schularick, and Taylor (2019).

Growth is then computed as the difference in logs:

∆ ln P er capita GDP_c,t= ln P er capita GDP_c,t− ln P er capita GDP_c,t−1

Table 1 shows that the average annual GDP growth per capita for 18 different countries. We can see that the majority of countries have GDP growth rates of approximately 1.50%, but there are a few outliers, such as Ireland, Italy, Spain and Switzerland, with annual per capita GDP growth rates of 4.07%, 0.82%, 2.09%, and 0.89%, respectively. Compared to Bae et al. (2021), the values are approximately the same, the small difference might be caused by a difference in indexation, i.e., Jord`a et al. (2019) measures GDP using constant 1990 INTL Dollars, rather than constant 2005 US Dollars, used by the World Bank’s World Development Indicators (WB WDI).

(15)

Table 1: Summary statistics for market capitalization by GDP, per capita GDP growth per country and macroeconomic variables, 1979-2021 annually

Mkt. cap./GDP_t−5 Per cap GDP growth

Inv./GDP Cred./GDP_t−5 Open./GDP Exp./GDP Turn./Cap_t−5 Inflation

Australia 77.42 1.78 25.06 87.65 31.55 25.53 49.77 4.46

Belgium 44.69 1.53 22.00 81.70 142.39 31.24 22.88 3.32

Canada 107.33 1.33 21.97 74.65 54.66 18.29 41.23 3.50

Denmark 29.21 1.17 19.92 145.38 56.22 38.36 28.35 4.32

Finland 24.56 1.55 22.10 68.45 52.63 28.07 37.32 4.69

France 44.84 1.55 21.82 79.52 39.94 24.44 59.68 4.19

Germany 30.59 1.29 21.67 90.27 55.50 12.69 94.87 2.54

Ireland 53.32 4.07 22.59 86.51 152.29 40.10 19.86 4.52

Italy 22.11 0.82 20.06 73.24 47.99 32.95 155.69 5.41

Japan 66.44 1.37 25.89 90.07 21.09 16.86 96.27 2.82

Netherlands 64.98 1.62 20.63 111.63 98.83 29.01 62.17 3.29

Norway 37.99 1.63 25.33 94.76 49.63 36.79 59.65 4.34

Portugal 37.19 1.78 23.16 91.60 65.94 26.24 64.11 7.58

Spain 47.28 2.09 23.59 101.94 50.56 21.37 77.01 6.47

Sweden 48.19 1.65 21.80 96.18 55.70 31.71 40.81 4.22

Switzerland 144.81 0.89 25.62 142.13 54.88 10.22 122.47 2.29

United Kingdom 91.02 1.89 19.34 90.98 38.16 36.59 53.81 4.93

United States 90.25 1.51 20.52 55.24 23.89 20.16 131.09 3.60

Notes: This table shows average values for per capita GDP growth, market capitalization/GDP and all macroeconomic variables. Per capita GDP growth, investment/GDP, credit/GDP, openness/GDP and expenditures/GDP are obtained from Jord`a et al. (2019) and ranges from 1985-2017. Market capitalization/GDP and turnover/GDP data is obtained from WB WDI ranging from 1979-2021. Inflation data is retrieved from OECD database for the period 1975-2021. All values are in percentages (%).

(16)

4.2 Market concentration 4 DATA

4.2 Market concentration

Following Bae et al. (2021), I collect end-year market capitalization, stock price times number of shares outstanding from the Compustat database for all countries but the United States. United States stock market data is obtained from the CRSP database.

Market capitalization values are sorted by year and country to obtain the five (ten) largest firms per year and country. Market concentration is then calculated as the sum of market capitalization of the largest five (ten) market capitalization firms divided by the total market capitalization of the entire domestic stock market exchange. The period for stock market concentration ranges from 1984 to 2021. Table 1 shows that there are some differences across countries. For example, the vast majority of countries have a significant number of large firms in the top 5(10). The UK, Finland, and Ireland report a highly concentrated stock market as 86.45%(89.71%), 77.33%(87.68%), and 67.49%(83.09%) of the entire stock market capitalization is in the top 5(10) largest firms, respectively. The US, Japan, and Canada, on the other hand, note a fairly diluted stock market, as only 5.57%(13.54%), 10.24%(17.34%), and 18.16%(30.28%) of the stock market is in the top 5(10), respectively.

To examine the effect of technology firms on the stock market, I construct a similar market concentration variable. Instead of using all firms in the top five(ten), I calculate the market concentration conditional on a firm being in the technology industry, defined by the SIC codes found in Appendix A. Figure 1 shows the average stock market concentration in the sample over the years for (non)technology and all firms in top 5(10). Surprisingly, technology stock market concentration for ten firms is not much higher compared to 5 firms. However, one can observe that during the dotcom bubble, concentration increased fairly and oscillates around 10% during the period 2006-2021. In total, stock market concentration tends to decrease over the sample

(17)

4.2 Market concentration 4 DATA

period.

Table 3 shows the correlation values among market concentration and other variables used in this paper. Market concentration of technology firms in top five is relatively uncorrelated with the control variables, implying that multicollinearity is not an is- sue. On the other hand, market concentration of technology firms in top ten has slightly higher correlations but the correlations are not alarmingly high. Obviously, market concentration of technology firms is highly correlated with other market concentration variables as they are both dependent on the total market concentration.

An increase in the market concentration of technology firms in the top 5(10) leads to a decrease in market concentration of non-technology firms in the top 5(10) as the number of firms is fixed. Furthermore, market concentration of technology firms is only weakly correlated with both market capitalization/GDP and Turnover/Cap., -0.02(0.00) and -0.05(-0.05), respectively. Moreover, one can see that per capita GDP growth bears no correlation with market concentration of technology firms, but a positive correlation with market concentration of non-technology firms, albeit insignificant.

17

(18)

Table 2: Summary statistics for market concentration variables

Market con.

tech(5)_t−5

Market con.

non-tech(5)_t−5

Market con.

all(5)t−5

Market con.

tech(10)_t−5

Market con.

non-tech(10)_t−5

Market con.

all(10)_t−5

Australia 1.07 35.76 36.83 1.59 45.68 47.27

Belgium 1.03 55.42 56.45 2.40 69.72 72.12

Canada 1.82 18.16 19.98 2.43 27.85 30.28

Switzerland 14.36 42.18 56.55 16.71 55.32 72.04

Germany 2.62 39.58 42.20 3.26 53.02 56.28

Denmark 19.38 29.24 48.62 22.28 43.86 66.14

Spain 0.19 49.98 50.18 0.60 65.36 65.96

Finland 53.47 23.87 77.33 54.21 33.46 87.68

France 4.97 32.45 37.42 5.60 45.08 50.68

Ireland 23.20 44.49 67.69 26.27 56.82 83.09

Italy 0.00 47.99 47.99 0.00 62.21 62.21

Japan 0.77 10.24 11.01 1.65 15.68 17.34

Netherlands 6.74 55.62 62.36 7.93 65.44 73.37

Norway 1.08 53.46 54.55 2.32 65.96 68.28

Portugal 0.00 61.97 61.97 0.00 81.52 81.52

Sweden 6.90 29.40 36.29 8.67 43.96 52.63

United Kingdom 1.19 85.26 86.45 1.78 87.94 89.71

United States 3.03 5.57 8.60 4.32 9.22 13.54

Notes: This table shows the average lagged market concentration for (non)technology firms in the top 5(10) per country. Market

(19)

4.3 Financial and macroeconomic variables 4 DATA

Figure 1: Average stock market concentration

Notes: This figure plots the average stock market concentration and if the majority of top 5(10) firms is in the technology industry over the sample period 1992-2021.

4.3 Financial and macroeconomic variables

Lastly, I construct control variables that the literature shows affect economic growth.

Firstly, from the World Bank’s World Development Indicators, I use domestic market capitalization of all firms by domestic GDP (Mkt. cap./GDP_t−5) and the value of domestic shares traded by domestic market capitalization (Turnover/Cap_t−5) for the

19

(20)

period 1975-2021. Secondly, I use the investment to GDP ratio (Investment/GDP), domestic credit provided to non-financial private sector by GDP (Credit/GDPt−5), overall government expenditures by GDP (Expenditures/GDP) and the sum of exports and imports by GDP (Openness/GDP), provided by Jord`a et al. (2019) for the period 1985-2017. Lastly, I use the OECD database to obtain annual domestic inflation rates (Inflation) for the period 1975-2021.

Table 1 shows summary statistics for the financial and macroeconomic control variables. One can see that the size of financial markets differ among the countries as Canada and Switzerland’s market capitalization/GDP are 103.33% and 144.81%, respectively, while Italy and Finland’s market capitalization/GDP are 22.11% and 24.56%, respectively, indicating that the latter countries have relatively smaller companies compared to the former. Moreover, the amount of activity on stock markets is also quite different. Italy, Switzerland, and the United States respectively report a Turnover/Cap of 155.69%, 122.47% and 131.09%, whereas Belgium, Denmark, and Ireland respectively report a Turnover/Cap of 22.88%, 28.35%, and 19.86%, indicating relatively large stock market activity in the former three countries compared to the latter three countries.

Not only are the financial markets different, but the economies are quite different as well. For example, Openness/GDP, indicating the degree to which non-domestic transactions take place, ranges from 21.09% to 152.29% for Japan and Ireland, respectively. This tells us that, Japan’s economy is quite restricted to domestically produced products and services, whereas Ireland’s economy thrives on foreign goods and services. Moreover, there is also a degree to which governments spend relative to the country’s GDP. Ireland’s economy has a relatively high share of government expenditures, 40.10%, whereas Switzerland has a relatively low share of government

(21)

Table 3 presents the Pearson correlation matrix for all variables used. One can see that the control variables, labeled 8 to 14, have relatively low correlation amongst each other. The highest absolute correlation coefficient, statistically significant at the 5% level, is 0.37 between turnover/cap and government expenditures. In the next section, I perform regression analyses to examine the statistical relationship between market concentration of technology firms and per capita GDP growth.

21

(22)

Table 3: Pearson correlation matrix

[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

[1] Mkt. Con. Tech(5)t−5 1.00

[2] Mkt. Con. Non-tech(5)_t−5 -0.36^∗ 1.00

[3] Mkt. Con. All(5)_t−5 0.34^∗ 0.76^∗ 1.00

[4] Mkt. Con. Tech(10)_t−5 0.99^∗ -0.37^∗ 0.33^∗ 1.00

[5] Mkt. Con. Non-tech(10)_t−5 -0.40^∗ 0.97^∗ 0.70^∗ -0.40^∗ 1.00

[6] Mkt. Con. All(10)t−5 0.30^∗ 0.74^∗ 0.96^∗ 0.30^∗ 0.75^∗ 1.00

[7] Per cap. GDP growth -0.00 0.07 0.07 -0.00 0.08 0.08 1.00

[8] Investment/GDP 0.05 -0.17^∗ -0.14^∗ 0.06 -0.17^∗ -0.14^∗ 0.03 1.00

[9] Domestic creditt−5 0.09^∗ 0.05 0.11^∗ 0.12^∗ 0.08 0.17^∗ -0.12^∗ -0.07 1.00

[10] Openness/GDP 0.14^∗ 0.19^∗ 0.29^∗ 0.17^∗ 0.25^∗ 0.39^∗ 0.03 -0.04 0.16^∗ 1.00

[11] Government expenditures/GDP 0.07 0.33^∗ 0.38^∗ 0.08 0.35^∗ 0.43^∗ -0.04 -0.34^∗ 0.03 0.33^∗ 1.00

[12] Market Capitalization/GDP_t−5 -0.02 -0.09^∗ -0.11^∗ 0.00 -0.17^∗ -0.17^∗ 0.09^∗ 0.07 0.28^∗ -0.06 -0.33^∗ 1.00

[13] Turnover/Cap_t−5 -0.05 -0.19^∗ -0.22^∗ -0.05 -0.22^∗ -0.25^∗ 0.02 0.05 0.09^∗ -0.17^∗ -0.37^∗ 0.15^∗ 1.00 [14] Inflation -0.12^∗ 0.17^∗ 0.08^∗ -0.13^∗ 0.20^∗ 0.11^∗ -0.31^∗ 0.21^∗ -0.36^∗ -0.10^∗ 0.14^∗ -0.35^∗ -0.23^∗ 1.00

Notes: This table reports the Pearson correlation coefficients among the variables. * denotes statistical significance at the 5% level.

(23)

5 RESULTS AND DISCUSSION

5 Results and discussion

Table 4 shows the regression output for the main regression equation mentioned in section 3, i.e., regressing per capita GDP growth on the five-year lags of market concentration of (non)technology firms in the top 5(10). All standard errors are robust and clustered by country. T-statistics are presented in parentheses.

Column (1) reports the regression estimates for market concentration of technology firms in the top 5, market capitalization/GDP, turnover/cap, and other control variables. The signs of market capitalization/GDP, turnover/cap, and control variables are in line with Bae et al. (2021), however, different from the authors, market capitalization and turnover remain statistically significant when adding market concentration. Government spending, credit/GDP are negatively associated with future per capita GDP growth and trade openness is positively associated with future per capita GDP growth, as expected from the AD-AS model. The coefficient of market concentration of technology firms in top 5 is statistically significant negative at the 1% level. A unit (percentage) increase in market concentration of technology firms in the top 5 leads to a significant 2.34 percent point decrease in future annual per capita GDP growth. In the long run, it will simply accumulate over time and the effect of a strongly concentrated stock market of technology firms will decrease GDP growth.

In column (4), I replace market concentration of technology firms in top 5 firms with technology firms in top 10. The signs of all variables remain the same, but the effect of market concentration is more pronounced. Instead, a unit increase in market concentration of technology firms in the top 10 will decrease future annual per capita GDP growth by 2.88%.

Column (2) shows the regression estimates for market concentration of non-technology

23

(24)

firms in the top and the control variables. Similar to column (1), the control variables show signs in line with current literature. However, the coefficient market concentration is statistically significant positive at the 5% level. A unit increase in market concentration would imply an increase of 1.65% in future annual per capita GDP growth. This result is not in line with Bae et al. (2021), who show a clear decrease in future annual per capita GDP growth. This difference may be due to a lack of observations, as the authors’ sample size is vastly greater and more diverse in countries/economies. For example, the authors include both developing and developed countries, whereas this sample only includes developed countries, and have been developed during the time period.

In column (5), I replace market concentration of non-technology firms in the top 5 with non-technology firms in the top 10. The coefficient not only increases in significance but also in magnitude. A unit increase in market concentration of non- technology firms in the top 10 leads to an increase of 2.43% in future annual per capita GDP growth. This coefficient estimate is statistically significant at the 1%

level.

Columns (3) and (6) show the regression output for the overall market concentration.

In column (3), we can see that the coefficient of market concentration for all firms in the top 5 is 0.96, however, this estimate is not statistically significant.

Similarly, column (6) shows that the coefficient of market concentration for all firms in the top 10 is 1.73, lacking statistical significance as well. Noteworthy are the signs of the control variables. The signs remain the same throughout the regression models. Signifying that the variables are proper control variables, and the insignificance is a proper result. However, this result is not in line with previous papers. The results indicate that stock market concentration of technology firms in the top 5(10)

(25)

in general is positively related to future economic growth. This evidence is puzzling given the function of the stock market and previous literature.

25

(26)

Table 4: Panel regression of per capita GDP growth on stock market concentration of technology firms.

(1) (2) (3) (4) (5) (6)

Market concentration tech(5)t−5 -2.34^∗∗∗

(-3.32)

Market concentration non-tech(5)t−5 1.65^∗∗

(2.17)

Market concentration all(5)t−5 0.96

(1.01)

Market concentration tech(10)t−5 -2.88^∗∗∗

(-3.94)

Market concentration non-tech(10)t−5 2.43^∗∗∗

(-2.60)

Market concentration all(10)t−5 1.73

(1.51) Market capitalization/GDPt−5 -1.31^∗∗∗ -1.32^∗∗∗ -1.29^∗∗∗ -1.30^∗∗∗ -1.30^∗∗∗ -1.29^∗∗∗

(-3.32) (-3.57) (-3.37) (-3.35) (-3.72) (-3.48) Turnover/Capt−5 -0.56^∗∗∗ -0.55^∗∗∗ -0.62^∗∗∗ -0.54^∗∗∗ -0.52^∗∗∗ -0.61^∗∗∗

(-4.08) (-3.92) (-3.85) (-3.96) (-3.64) (-3.83)

Domestic credit/GDPt−5 -1.56 -1.48 -1.47 -1.58 -1.52^∗ -1.47

(-1.50) (-1.57) (-1.49) (-1.53) (-1.60) (-1.51) Government spending/GDP -25.85^∗∗ -25.03^∗∗∗ -23.00^∗∗ -26.15^∗∗∗ -26.34^∗∗ -23.47^∗∗

(-2.10) (-2.09) (-2.03) (-2.12) (-2.16) (-2.06) Inflation -27.7^∗∗ -25.62^∗∗ -25.03^∗∗ 27.82^∗∗ -26.97^∗∗ -25.55^∗∗

(-2.27) (-2.00) (-2.00) (-2.12) (-2.06) (-2.06)

Openness/GDP 4.24^∗ 4.70^∗ 4.55^∗ 4.36^∗ 4.82^∗ 4.62^∗

(1.78) (1.91) (1.96) (1.82) (1.91) (1.94)

Investment/GDP -2.66 -4.03 -3.45 -2.75 -4.07 -3.60

(-0.30) (-0.45) (-0.41) (-0.31) (-0.45) (-0.42)

Country FE Yes Yes Yes Yes Yes Yes

No. of observations 428 428 428 428 428 428

R² 18.53 18.81 17.98 18.92 19.59 18.27

Notes: This table presents the results of panel regression in which GDP per capita is regressed on market concentration of technology firms at t-5 for top 5 and top 10 firms. The sample includes yearly observations for the period

(27)

5.1 Robustness check 5 RESULTS AND DISCUSSION

5.1 Robustness check

In order to test the robustness of the previous findings, I exclude highly concentrated markets. More specifically, I exclude Finland, Great Britain, and Portugal. These countries show exceptionally high market concentration levels compared to literature.

Table 5 shows the regression output for the robustness check. From the table, one can see that the estimates change significantly. Most notably, the coefficients of market concentration of technology firms in top 5 and top 10 and market concentration of all firms. In columns (1) and (4), market concentration of technology firms in both top 5 and 10 lose statistical significance at any level compared to table 4. This loss is most likely due to the exclusion of Finland, as the country reports very high concentrations, shown in figures 2 and 3 panel (a). On the other hand, in columns (3) and (6), market concentration of all firms in the top 5 and top 10 gain statistical significance at the 10% level. An unit increase in market concentration of 5(10) firms indicate 1.66%(2.29%) annual per capita GDP growth rates in five years. However, since per capita GDP growth rates are highly cyclical, due to the nature of business cycles, we can merely say that there is a significant positive effect of market concentration on per capita GDP growth.

To summarize, using the complete sample of 18 countries, I find that market concentration of technology firms in the top 5(10) is statistically significantly related to future per capita GDP growth. However, this finding disappears once Finland is excluded from the sample, most likely due to the highly concentrated technology firms in the top 5. Market concentration of non-technology firms remains significant, albeit at the 10% level. The estimates’ signs are positive, which is unexpected given previous literature.

27

(28)

Table 5: Panel regression of per capita GDP growth on stock market concentration of technology firms excluding Finland, Great Britain, and Portugal.

(1) (2) (3) (4) (5) (6)

Market concentration tech(5)t−5 -1.45 (-0.60)

Market concentration non-tech(5)t−5 1.51^∗ (1.76)

Market concentration all(5)t−5 1.66^∗

(1.66)

Market concentration tech(10)t−5 -2.12

(-0.78)

Market concentration non-tech(10)t−5 2.05^∗

(1.87)

Market concentration all(10)t−5 2.29^∗

(1.76) Market capitalization/GDPt−5 -1.33^∗∗ -1.36^∗∗∗ -1.36^∗∗∗ -1.32^∗∗∗ -1.35^∗∗∗ -1.35^∗∗∗

(-2.49) (-2.67) (-2.64) (-2.51) (-2.76) (-2.69) Turnover/Capt−5 -0.56^∗∗∗ -0.53^∗∗∗ -0.56^∗∗∗ -0.55^∗∗∗ -0.52^∗∗∗ -0.55^∗∗∗

(-3.65) (-3.67) (-3.64) (-3.57) (-3.56) (-3.62)

Domestic credit/GDPt−5 -1.47 -1.41 -1.41 -1.47 -1.41 -1.41

(-1.43) (-1.47) (-1.46) (-1.45) (-1.48) (-1.44) Government spending/GDP -24.47^∗ -25.22^∗ -24.20^∗∗ -24.70^∗ -25.73^∗ -24.62^∗ (-1.67) (-1.73) (-1.77) (-1.67) (-1.75) (-1.79)

Inflation -21.96 -20.57 -20.20 -21.87 -21.46 -21.29

(-1.58) (-1.36) (-1.37) (-1.54) (-1.40) (-1.42)

Openness/GDP 3.39^∗ 3.84^∗ 3.97^∗ 3.47^∗ 3.89^∗ 3.94^∗

(1.66) (1.79) (1.79) (1.72) (1.81) (1.78)

Investment/GDP 2.62 1.18 1.21 2.50 1.06 1.28

(0.29) (0.12) (0.13) (0.26) (0.11) (0.13)

Country FE Yes Yes Yes Yes Yes Yes

No. of observations 364 364 364 3648 364 364

R² 17.35 18.29 18.26 17.46 18.53 18.44

Notes: This table presents the results of panel regression in which GDP per capita is regressed on market concentration of technology firms at t-5 for top 5 and top 10 firms excluding Finland, Great Britain, and Portugal.

(29)

6 CONCLUSION

6 Conclusion

The role of stock markets is to improve capital allocation throughout the economy in order to increase productivity and, in turn, increase general human welfare. I investigate the effect of stock market concentration, but more specifically, the effect of technology firms’ stock market concentration for 18 different developed countries with the use of panel regressions. This paper shows that a concentrated stock market of technology firms decrease per capita GDP growth significantly in the future, whereas stock market concentration of non-technology and all firms increase per capita GDP growth in the future. These findings, however, are not robust when excluding Finland, which has a highly concentrated technology stock market. The effect of technology firms is insignificant. Moreover, market concentration of non- technology firms remain significantly positive, although at the 10% level. Market concentration of all firms retain its positive effect on future per capita GDP growth and increase in significance, to 10%.

The paper’s findings are unexpected and stand on the opposing side in the financial literature. Current economic and financial theory agree that, highly concentrated markets produce monopolies, increasing prices, but most importantly, obscures in- centives to innovate, leading to a decrease in long run macroeconomic progress.

29

(30)

References References

References

Autor, D., Dorn, D., Katz, L. F., Patterson, C., & Van Reenen, J. (2020). The fall of the labor share and the rise of superstar firms. The Quarterly Journal of Economics, 135 (2), 645–709.

Bae, K.-H., Bailey, W., & Kang, J. (2021). Why is stock market concentration bad for the economy? Journal of Financial Economics, 140 (2), 436–459.

Bekaert, G., Harvey, C. R., & Lundblad, C. (2011). Financial openness and productivity. World Development , 39 (1), 1–19.

Bloom, N., Jones, C. I., Van Reenen, J., & Webb, M. (2020). Are ideas getting harder to find? American Economic Review , 110 (4), 1104–44.

Brown, J. R., Martinsson, G., & Petersen, B. C. (2017). Stock markets, credit markets, and technology-led growth. Journal of Financial Intermediation, 32 , 45–59.

Davis, S. J. (2017). Regulatory complexity and policy uncertainty: headwinds of our own making. Becker Friedman Institute for Research in economics working paper (2723980).

Greenwood, J., & Jovanovic, B. (1999). The information-technology revolution and the stock market. American Economic Review , 89 (2), 116–122.

Guti´errez, G., & Philippon, T. (2017). Declining competition and investment in the us (Tech. Rep.). National Bureau of Economic Research.

Guti´errez, G., & Philippon, T. (2019a). Fading stars. In Aea papers and proceedings (Vol. 109, pp. 312–16).

Guti´errez, G., & Philippon, T. (2019b). The failure of free entry (Tech. Rep.).

National Bureau of Economic Research.

Jord`a, `O., Knoll, K., Kuvshinov, D., Schularick, M., & Taylor, A. M. (2019). The

(31)

References References

rate of return on everything, 1870–2015. The Quarterly Journal of Economics, 134 (3), 1225–1298.

King, R. G., & Levine, R. (1993). Finance and growth: Schumpeter might be right.

The quarterly journal of economics, 108 (3), 717–737.

P´astor, L., & Veronesi, P. (2009). Technological revolutions and stock prices. Amer- ican Economic Review , 99 (4), 1451–83.

Schlingemann, F. P., & Stulz, R. M. (2020). Have exchange-listed firms become less important for the economy? (Tech. Rep.). National Bureau of Economic Research.

Walcott, S. (2000, 01). Defining and measuring high technology in georgia.

31

(32)

Appendix A

Table 6: SIC Codes

SIC Code Categories

2833 Medicinals, botanicals 2834 Pharmaceutical preparations

2835 Diagnostic substances

2836 Other biological products

3571 Electronic computers

3572 Computer storage devices

3575 Computer terminals

3577 Computer peripherals

3578 Calculating, accounting equipment 3579 Office machines, nec.

3661 Telephone, telegraph apparatus 3663 Radio, TV communications equip.

3669 Communications equipment nec.

3671 Electron tubes

3672 Printed circuit boards 3674 Semiconductors, related devices 3675 Electronic capacitors

3676 Electronic resistors

3677 Electronic coils, transformers 3678 Electronic connectors

(33)

Table 6: – continued from previous page

SIC Codes Categories

3679 Electronic components, nec.

3761 Guided missiles, space vehicles 3764 Space propulsion units, parts 3769 Space vehicle equipment, nec.

3812 Search, navigation equipment 3821 Laboratory apparatus, furniture 3822 Environmental controls 3823 Process control instruments 3824 Fluid meters, counting devices 3825 Instruments measuring electricity 3826 Analytical instruments 3827 Optical instruments, lenses 3829 Measuring, controlling devices 3841 Surgical and medical instruments 3842 Surgical appliances and supplies 3843 Dental equipment and supplies 3844 X-ray apparatus, tubes 3845 Electromedical equipment 7371 Computer programming services

7372 Prepackaged software

7373 Computer integrated systems design

Continued on next page

(34)

Table 6: – continued from previous page

SIC Codes Categories

7374 Data processing, preparation 7375 Information retrieval services 7376 Computer facilities management 7377 Computer rental, leasing 7378 Computer maintenance, repair 7379 Computer-related services, nec.

8711 Engineering services

8712 Architectural services

8713 Surveying services

8731 Commercial physical research 8732 Commercial nonphysical research 8733 Noncommercial research org.

8734 Testing laboratories

Notes: This table provides the SIC used by this paper to define a technology company along with the respective category.

(35)

Figure 2: Market concentration top 5 firms

(a) Market concentration of top 5 technology firms over the sample period

(b) Market concentration of top 5 non-technology firms over the sample period

Notes: This figure plots the stock market concentration of top 5 firms per country over the sample period 1992-2021.

(36)

Figure 3: Market concentration of top 10 firms

(a) Market concentration of top 10 technology firms over the sample period

(b) Market concentration of top 10 non-technology firms over the sample period

(37)

Master's Thesis

This notebook is for my thesis

presented for the Master's degree in Quantitative Finance. The notebook will be the source code of my paper's data preparation, analysis, and figures.

Please contact me at calvin.li@student.uva.nl (mailto:calvin.li@student.uva.nl) for any questions.

Stock market concentration and the real economy. The role of superstar firms and their impact on the real economy.

Data preparation

Prerequisites

Import relevant libraries

In [ ]:

Mount drive

In [ ]:

Prepare macro data using JST

In [ ]:

Create master country data

import pandas as pd import numpy as np import datetime as dt

import matplotlib.pyplot as plt import os

import statsmodels.formula.api as smf

%cd "C:\Users\Calvin\iCloudDrive\Documents\Finance\Master's Thesis\DATA"

# Read data

jst = pd.read_excel("JSTdatasetR5.xlsx", sheet_name="Data")

# Filter variables

jst = jst[["year", "iso","rgdpmad", "cpi", "iy", "tloans", "expenditure", "revenue", "gdp", "exports", "imports"]]

# Restrict data to 1985 and later jst = jst[jst['year']>1984]

# Calculate log returns

jst['gdpgrowth'] = jst.groupby("iso")['rgdpmad'].apply(lambda x: np.log(x) - np.log(x.shift()))*100

# Calculate economic openness according to Bae et al. (2021) jst['openness'] = (jst['exports'] + jst['imports'])/jst['gdp']

# Calculate government expenditures by gdp jst['exp_gdp'] = jst['expenditure']/jst['gdp']

# Calculate domestic credit by gdp

jst['credit_gdp'] = jst['tloans']/jst['gdp']

# Replace missing values with NaN jst = jst.replace('..', np.nan)

# Change variable name jst['loc']=jst['iso']

# Drop unnecessary variables

jst = jst[['year','loc','rgdpmad', 'iy', 'credit_gdp', 'exp_gdp', 'openness', 'gdpgrowth']]

#jst = jst[jst['loc']!='GBR']

# Set multiindex

jst = jst.set_index(['loc','year'])

(38)

In [ ]:

Concentration data

In [ ]:

US stock data

def end_year_data(df):

"""

Convert DataFrame to end of year data args:

df(DataFrame): DataFrame that needs to be converted returns:

df_price (DataFrame): DataFrame with only end-of-year values """

df = df.dropna()

df['date'] = pd.to_datetime(df['datadate'].astype(str), format= '%Y%m%d') df['year'] = df['date'].dt.year

df['month_year'] = df['date'].dt.to_period('M')

df_price = pd.DataFrame(df.groupby(['gvkey','year']).last()) return df_price

# Loop through files and append to list

for i in ['JPN','AUS','DEU','FRA', 'CHE', 'NLD','BEL', 'NOR', 'ESP','DNK','FIN','ITA','PRT','SWE','IRL','NA']

file_name = "NA{}.csv"

df_list = []

for i in range(1,9):

df_list.append(pd.read_csv(file_name.format(i), low_memory=False))

# Create list of DataFrames df_list2 = []

for i in df_list:

df_list2.append(end_year_data(i))

# Concatenate country returns df_uk = pd.concat(df_list2)

countries = ['USA','Canada','Sweden','Portugal','Ireland','Italy','Finland','Denmark','Belgium','Spain','Switzerland','Aus file_name = "{}.csv"

df_list = []

for i in countries:

df_list.append(pd.read_csv(file_name.format(i), low_memory=False)) countries_data = pd.concat(df_list)

countries_data = countries_data.set_index('gvkey') countries_data.to_csv('Master.csv')

# Lists to use

countries = ['USA','GBR','DEU','FIN','IRL','FRA','CAN','NLD','ESP','ITA','NOR','SWE','CHE','AUS','JPN','BEL','DNK','PRT']

list_tech_oregon = ['3570', '3571', '3572','3575','3576','3577','3578','3579','3600','3612','3613','3620','3621','3630','3 list_tech_census = ['2833', '2834','2835','2836','3570','3571','3572','3575','3577','3578','3579','3661','3663','3669','36

(39)

In [ ]:

Concentration top 20 In [ ]:

In [ ]:

Concentration top 15 In [ ]:

In [ ]:

Concentration 10 In [ ]:

In [ ]:

# Read file

US_stock_data = pd.read_csv('US stock data.csv', low_memory=False)

# Format date

US_stock_data['date'] = pd.to_datetime(US_stock_data['date'], format="%Y%m%d")

# Get year and month

US_stock_data['year'] = US_stock_data['date'].dt.year US_stock_data['month'] = US_stock_data['date'].dt.month

# Drop negative stock values

US_stock_data = US_stock_data.drop(US_stock_data[US_stock_data.PRC<0].index)

# Drop missing data

US_stock_data = US_stock_data.dropna()

# Calculate market cap per firm per index

US_stock_data['Market_cap'] = US_stock_data['SHROUT']*US_stock_data['PRC']

# Keep end year values

US_stock_data = US_stock_data.drop(US_stock_data[US_stock_data.month!=12].index)

# Drop dates before 1962

US_stock_data = US_stock_data.drop(US_stock_data[US_stock_data.year<1985].index)

# Set index to year

US_stock_data= US_stock_data.set_index('year')

# Assign rank to largest 5 firms for the year

US_stock_data['rank'] = US_stock_data.groupby('year')['Market_cap'].rank(method='max', ascending = False) US_stock_data['total_cap'] = US_stock_data.groupby('year')['Market_cap'].sum()

US_stock_data['loc'] = 'USA'

# Filter top 20

market_20 = US_stock_data[US_stock_data['rank']<21]

# Calculate total market concentration

total_cap2 = pd.DataFrame(US_stock_data.groupby(['loc','year'])['Market_cap'].sum()) total_cap2.columns=['total_cap']

market_20['tech'] = market_20.SICCD.apply(lambda x: 1 if x in list_tech_census else 0)

market_20['market_tech_20_sum'] = market_20.groupby('year').apply(lambda x: x[x['tech']==1]['Market_cap'].sum()) market_20['market_non_tech_20_sum'] = market_20.groupby('year').apply(lambda x: x[x['tech']==0]['Market_cap'].sum())

# Calculate total market capitalization of top 20 tech firms

total_cap2['market_tech_20_sum'] = market_20.groupby(['loc','year'])['market_tech_20_sum'].mean()

# Calculate total market capitalization of top 20 non-tech firms

total_cap2['market_non_tech_20_sum'] = market_20.groupby(['loc','year'])['market_non_tech_20_sum'].mean()

# Filter top 15

total_cap2['market_non_tech_15_sum'] = market_15.groupby(['loc','year'])['market_non_tech_15_sum'].mean() total_cap2

total_cap2.rename(columns={'total_cap':'market_cap'}, inplace=True)

# Filter top 10

(40)

In [ ]:

Concentration 5 In [ ]:

In [ ]:

Non-US stock data

In [ ]:

Create regression dataset In [ ]:

Market concentration for 20 firms In [ ]:

In [ ]:

# Filter top 15

master = pd.read_csv("Master.csv")

# Calculate market cap of company each year in each country master['market_cap'] = master['cshoc']*master['prccd']

# Assign ranks per year per country to companies

master['rank'] = master.groupby(['year','loc'])['market_cap'].rank(method='max', ascending = False)

# Domestic firms only

master = master[master['loc'].isin(countries)]

# Exclude 2022

master = master[master['year']!=2022]

# Make dataframe of total market cap by year and country

total_cap=pd.DataFrame(master.groupby(['loc','year'])['market_cap'].sum())

# Make DataFrame of largest five companies

master_20 = master.drop(master[master['rank']>20].index) master_20

master_20['sic']=master_20['sic'].astype(int) master_20['sic']=master_20['sic'].astype(str)

master_20['tech'] = master_20.sic.apply(lambda x: 1 if x in list_tech_census else 0) master_20

list_master = []

for i in countries:

list_master.append(master_20[master_20['loc']==i])

def market_con_maker(df):

df = df.set_index('year')

df['market_tech_20_sum'] = df.groupby('year').apply(lambda x: x[x['tech']==1]['market_cap'].sum()) df['market_non_tech_20_sum'] = df.groupby('year').apply(lambda x: x[x['tech']==0]['market_cap'].sum()) return df

market = []

University of Amsterdam Faculty of Economics and Business A thesis presented for the degree of Master of Science in Finance