Royal Economic Society Royal Economic Society Royal Economic - - PowerPoint PPT Presentation

Royal Economic Society

Royal Economic Society

Royal Economic Society

The RES Prize

Presented by Rachel Griffith

Royal Economic Society

The RES Prize Michele PELLIZZARI and Giacomo DE GIORGIO

Royal Economic Society

Royal Economic Society

The Economic Journal Lecture Philippe Aghion

Innovation and Top Income Inequality Chair: Rachel Griffith

Royal Economic Society

Royal Economic Society

Economic Journal Lecture: Innovation and Top Income Inequality

Philippe Aghion RES - 31 March 2015

Introduction

Past decades have witnessed a sharp increase in top income inequality worldwide and particularly in developed countries

US MALE WAGE INEQUALITY, 1937-2005

Source: Goldin and Katz (2008)

5 10 15 20 25 30 1918 1920 1922 1924 1926 1928 1930 1932 1934 1936 1938 1940 1942 1944 1946 1948 1950 1952 1954 1956 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 Percentile Share U.S. Top 1% U.S. Top 0.1%

US Top 1% US Top 0.1%

Income shares at the very top over last 100 years: US top 1% increases from 9% in 1978 to 22% in 2012

Source: Atkinson, Piketty & Saez; High Income Database

5 10 15 20 25 1918 1920 1922 1924 1926 1928 1930 1932 1934 1936 1938 1940 1942 1944 1946 1948 1950 1952 1954 1956 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 Percentile Share U.K. Top 1% U.K. Top 0.1%

UK Top 1% UK Top 0.1%

Income shares at the very top: UK top 1% increases from 6% in 1978 to 14% in 2009

Source: Atkinson, Piketty & Saez; High Income Database

Introduction

However no consensus has been reached as to the main underlying factors behind this surge in top income inequality In this lecture we shall argue that innovation is certainly one such factor and that it also affects social mobility.

Introduction

Three parts to the lecture:

Part 1: Model − → we develop a Schumpeterian model of innovation, top income inequality and social mobility

Part 2: Empirical analysis using US aggregate data − → we use cross-state panel data over the period 1995-2010 to look at the effect of innovativeness on top income inequality. − → we use cross Commuting-zone data from Chetty et al (2015) to look at the effect of innovativeness on social mobility.

Part 3: Empirical analysis using using individual data − → we combine individual patenting with individual fiscal data to look at the social mobility of inventors versus non-inventors.

Introduction

Part 1 and Part 2 are drawn from Aghion-Akcigit-Bergeaud-Blundell-Hemous (2015) Part 3 is drawn from ongoing work by Aghion-Akcigit-Toivanen (2015)

Innovation and Top Income Inequality

Philippe Aghion (Harvard) Ufuk Akcigit (UPenn) Antonin Bergeaud (Bank of France) Richard Blundell (UCL) David Hemous (INSEAD) RES - 31 March 2015

Summary of Part 1

We develop a simple Schumpeterian growth model where:

growth results from quality-improving innovations by incumbents or from potential entrants.

facilitating innovation

increases top income shares as top incomes are earned by innovators

Summary of Part 1

We develop a simple Schumpeterian growth model where:

growth results from quality-improving innovations by incumbents or from potential entrants.

facilitating innovation

increases top income shares as top incomes are earned by innovators spurs social mobility as innovation entails creative destruction

Summary of Part 1

The model predicts:

Innovation by entrants and/or incumbents increases top income inequality;

Innovation by entrants increases social mobility;

Entry barriers (e.g. from lobbying), lower the positive effects of entrants’ innovations on top income inequality and social mobility.

Summary of Part 2

Our main empirical findings from cross-state panel regressions:

The top 1% income share is positively and significantly correlated with the state’s degree of "innovativeness"

This at least partly reflects a causal effect of innovation on top incomes

Innovativeness is less positively correlated with broader measures of inequality.

Summary of Part 2

From cross-section regressions performed at the CZ level:

Innovativeness is positively correlated with upward social mobility

The positive effects of innovativeness on social mobility, is driven mainly by entrant innovators and less so by incumbent innovators

The positive effects of innovation on the top 1% income share and on social mobility are both dampened in states with higher lobbying intensity

Relationship with existing literature

The analysis in this paper relates to several strands of literature on income inequality and growth

Empirical literature on inequality and growth: Forbes (2000), Banerjee and Duflo (2003), Frank (2009)

Literature on skill-biased technical change: Katz and Murphy (1992), Krusell, Ohanian, Ríos-Rull and Violante (2000), Goldin and Katz (2008), Acemoglu, (1998, 2002 and 2007)

Literature on evolution of income and wealth inequality: Piketty and Saez (2003), Gabaix and Landier (2008), Piketty (2014)

Ongoing work on innovation and social mobility using individual data: Toivanen and Vaananen (2014), Bell et al (2015)

Outline

Introduction Part 1: Model Part 2: Empirical analysis using US aggregate data Part 3: Empirical analysis using individual data Conclusion

Model

Population

Discrete time; continuum of individuals of measure 2: − → half are capital (firm) owners and the rest works as production workers Each individual lives only for one period Every period, a new generation of individuals is born and individuals that are born to current firm owners inherit the firm from their parents The rest of the population works in production unless they successfully innovate and replace incumbents’ children.

Model

Production

A final good is produced according to: ln Yt =

1

0 ln yitdi

Each intermediate is produced with a linear production function yit = qitlit

Model

Innovation

When there is a new innovation in any sector i : qi,t+1 = ηHqi,t. If there is no new innovation in sector i in period t + 1, the incumbent’s technological lead shrinks to ηL where ηL < ηH. If there is a new innovation in sector i, the previous technology becomes fully available to every firm in the economy, therefore the technological lead remains ηH. An incumbent can use lobbying to prevent entry by an innovator → Lobbying is successful with probability z, in which case, the innovation is not implemented.

Model

R&D technology

By spending CJ,t (x) = θJ x2 2 Yt an incumbent (J = I) or entrant (J = E) can innovate with probability x.

Model

Timing of events within each period

In each line i, a potential entrant spends Ct (xi) and the offspring of the incumbent in sector i spends Ct (˜ xi) .

With probability (1 − z) xi the entrant succeeds, replaces the incumbent and obtains a technological lead ηH; with probability ˜ xi the incumbent succeeds and improves its technological lead from ηL to ηH, with probability 1 − (1 − z) xi − ˜ xi, there is no successful innovation and the incumbent stays the leader with a technological lead of ηL

Production and consumption takes place and the period ends.

Model

Equilibrium profits and wages

Marginal cost of production of intermediate producer i at time t : MCit = wt qi,t . Hence the price charged at time t by intermediate producer i is: pi,t = wtηit, qi,t where ηi,t ∈ {ηH, ηL} depending on when the last innovation

• ccurred (recall that recent technologies have higher markups).

Model

Equilibrium labor demand and profits

Use the fact that in equilibrium pi,tyit ≡ Yt. Equilibrium profits in sector i at time t: πit = (pit − MCit)yit = ηit − 1 ηit Yt,

Model

Equilibrium profits

Hence profits are higher if the incumbent has recently innovated, namely: πH,t = ηH − 1 ηH

≡πH

Yt > πL,t = ηL − 1 ηL

≡πL

Yt.

Model

Income inequality

Let µt denote the fraction of high-mark-up sectors Entrepreneur share is: entrepreneur_sharet = Yt − wt Yt = 1 − µt ηH − 1 − µt ηL Thus the entrepreneur share is increasing in the fraction of high-mark-up sectors µt. − → µt in turn depends upon innovation intensities by entrants and incumbents (x and x).

Model

Equilibrium innovation investments

The offspring of a previous period’s incumbent solves: max

˜ x

˜ xπHYt + (1 − ˜ x − (1 − z) x∗) πLYt + (1 − z) x∗wt −θI ˜

x 2 2 Yt

• .

A potential entrant solves: max

x

• (1 − z) xπHYt + (1 − x (1 − z)) wt − θE

x2 2 Yt

• ()

Model

Equilibrium innovation investments

Nash equilibrium (x∗, ˜ x∗) where x∗ and ˜ x∗ are decreasing functions of (θE , θI ) Higher entry barriers (higher z) discourage entrant innovation.

Model

More formally: ˜ x∗ = πH − πL θI = 1 ηL − 1 ηH 1 θI and x∗ =

• πH − 1

ηL +

• 1

ηL − 1 ηH

• ˜

x∗ (1 − z) θE − (1 − z)2

1 ηL − 1 ηH

• .

Model

Equilibrium share of high mark up sectors

We have: µt = µ∗ = (1 − z) x∗ + ˜ x∗

Model

Equilibrium income shares

The entrepreneur and labor income shares in equilibrium are: entrepreneur_sharet = 1 − 1 ηL + 1 ηL − 1 ηH

• ((1 − z) x∗ + ˜

x∗). and wage_sharet = wt Yt = 1 ηL − 1 ηL − 1 ηH

• ((1 − z) x∗ + ˜

x∗) Thus any change (e.g lower R&D costs) which fosters innovation by incumbents or entrants also increases the entrepreneur share of income. This effect is lower when barriers to entry (z) are larger.

Model

Social mobility

Probability that worker’ offspring is also a worker: Ψ = 1 − x∗ (1 − z) . Hence we define social mobility as M = 1 − Ψ = x∗ (1 − z) , which is increasing in the innovation rate x∗ but less so the higher entry barriers (i.e the higher z). Note that a reduction in the incumbent’s R&D costs will also foster social mobility (general equilibrium effect).

Model

Predictions

Entrant and incumbent innovation increase top income inequality; Entrant innovation increases social mobility; Entry barriers lower the positive effects of entrant innovation on top income inequality and social mobility.

Outline

Introduction Part 1: Model Part 2: Empirical analysis using US aggregate data Part 3: Empirical analysis using individual data Conclusion

Data and measurement

Our core empirical analysis is carried out at US state level. Our dataset covers the period 1975-2010, a time range imposed upon us by the availability of patent data.

Data and measurement

Inequality

Data on share of income owned by the top 1% and the top 10% of income distribution are drawn from the US State-Level Income Inequality Database (Frank, 2009). − → from that data source, we also gather information on Atkinson Index, Theil Index and the Gini Index. In every US state, the top 1% income share has increased between 1975 and 2010 − → the unweighted mean value was around 8% in 1975 and reached 21% in 2007 before slowly decreasing to 16.3% in 2010.

Data and measurement

Innovation

When looking at cross state or more local levels, the US patent office (USPTO) provides complete statistics for patents granted between the years 1975 and 2010. For each patent, it provides information on the state of residence of the patent assignees, the date of application of the patent and a link to every citing patents granted before 2010. For patents with multiple assignees, we assume that they are split evenly among assignees and thus we attribute only a fraction of the patent to each assignee. We follow Jaffe, Hall and Trajtenberg (2001) to address the issue of truncation bias in both the number of patents and the number of citations.

Data and measurement

Innovation

The USPTO classification considers three types of patents according to the official documentation:

Utility patents that are used to protect a new and useful invention, or an improvement to an existing process.

Design patents that are used to protect a new design of a manufactured object.

Plant patents that protect some new varieties of plants.

The first type accounts for more than 90% of all patents at the USPTO and it is the only type of patents for which we have complete data. − → We thus focus on utility patents, in line with the patenting literature.

Data and measurement

Innovation

There is a substantial amount of variation in innovativeness both across states and over time.

Between 1975 and 1990: Delaware, Connecticut, New Jersey and Massachusetts were the most innovating states, whereas Arkansas, Mississippi and Hawaii were the least innovative states with less than 0.05 patents per thousands inhabitants

Between 1990 and 2009, the most innovative states were Idaho, Vermont, Massachusetts, Minnesota and California, whereas Arkansas, West Virginia and Mississippi all had less than 0.06 patents per 1000 inhabitants.

Data and measurement

Quality of innovation

Four measures of innovation quality, aggregated at the state level:

3, 4 and 5 year windows citations counter − → the number of citations received within no more than 3, 4 or 5 years after the application date

Is the patent among the 5% most cited in the year by 2010 − → dummy variable equal to one if the patent applied for in a given year belong to the top 5% most cited patents.

Total corrected citation counter − → the number of times a patent has been cited

Has the patent been renewed − → dummy variable equal to one if the patent has been renewed (at least one) before 2014

Data and measurement

Control variables

Output gap to control for the business cycle Share of state GDP accounted for by the financial sector Size of the government sector GDP per capita Growth of total population

Regression equation

Regressing top income inequality on innovativeness: log(yit) = A + Bi + Bt + β1 log(innovi(t−1)) + β2Xit + εit.

OLS regressions of patents per capita on top 1%

OLS regressions of various measure of innovations on top 1%

OLS regressions of innovation on various measure of inequality

Instrumentation

First instrument

Following Aghion et al (2004), we consider the time-varying State composition of the appropriation committees of the Senate and the House of Representatives. A Committee member often push towards subsidizing research education in her State, in order to increase her chances of reelection in that State. − → a state with one of its congressmen seating on the committee is likely to receive more funding for research education, which should increase its innovativeness in following years.

Instrumentation

Second instrument

Second instrument based on knowledge spillovers. − → The idea is to instrument innovation in a state by the sum of innovation intensities in other states weighted by the relative innovation spillovers from these other states.

Instrumentation

Second instrument

More formally, if m(i, j, T) is the number of citations from a patent in state i, to a patent of state j over period 1975-1984, and if innov(j, t) denotes our measure of innovativeness in state j at time t, then we posit: wi,j = m(i, j, T)

∑

k=i

m(i, k, T) and Yi,t = ∑

j=i

wi,j ∗ innov(j, t − 1).

IV regressions: First instrument (appropriation committee)

IV regressions of number of patents per capita (various lags) on top 1%

IV regressions: Second Instrument (spillover)

IV regressions of innovation on top 1% with additional controls for financial sector and oil. First instrument (appropriation committee). … Col 2: remove NY, DE, CT and SD (highest shares of financial sector). Col 3: remove all patents from financial-related IPC classes. Col 6: remove all patents from oil- related IPC classes.

Magnitude of the effects

When measured by the number of patent per capita, innovativeness accounts on average for about 17% of the total increase in the top 1% income share between 1975 and 2010 according to either IV regression

Extensions

The effect of innovativeness on social mobility Entrant versus incumbent innovation Lobbying as a dampening factor

35 40 45 50 55

• 0.5

0.5 1.0 1.5 2.0

Upward Mobility (Y 25 ) Absolute Upward Mobility vs. Income Growth in CZ Annualized Real Income Growth From 1990 to 2008 ρ = 0.369 (0.092)

35 40 45 50 55 $40,000 $50,000 $60,000 $70,000 $80,000

Upward Mobility (Y 25 )

Interquartile Range (p25-75) in Mean Household Parent Income 1996-2000

Upward Mobility vs. Inequality in CZ The “Great Gatsby” Curve Within the U.S. ρ = -0.475 (0.089)

35 40 45 50 55 5 10 15 20 Income Share of the Top 1% Based on Parents 1996-2000

Upward Mobility (Y 25 ) Upward Mobility vs. Top 1% Income Share in CZ Controlling for Interquartile Range (p25-75) ρ = 0.178 (0.068)

CZ level: Effect of innovation on social mobility. OLS regressions

CZ level: New Entrants VS Incumbent innovation, effect on social mobility. OLS regressions.

State level: New Entrants VS Incumbent innovation, effect on top 1%. IV regressions (First instrument: appropriation committee).

Effect of lobbying on new entrant and incumbent innovation on top 1% and social mobility. IV regressions for col 3, OLS for others.

Summarizing Part 2

We have analyzed the effect of innovation-led growth on top incomes and on social mobility. We found positive and significant correlations between (entrant) innovation, top income shares and social mobility. Our instrumentation at cross-state level suggested a causality from innovativeness to top income shares. When measured by the number of patent per capita, innovativeness accounts on average across US states for about 17% of the total increase in the top 1% income share between 1975 and 2010.

Outline

Introduction Part 1: Model Part 2: Empirical analysis using US aggregate data Part 3: Empirical analysis using individual data Conclusion

Living ”American Dream” in Finland: The Social Mobility of Innovators

Philippe Aghion Ufuk Akcigit Otto Toivanen Harvard UPenn KU Leuven RES - 31 March 2015

Data

The data used now includes

all inventors in our data (i.e., individuals who obtained a USPTO patent 1990 - 1999) that work in firms that participate in the R&D survey.

The original inventor sample consists of some 75% of all Finnish inventors of USPTO patents that could be matched to the Finnish employer-employee data.

The 884 inventors in the current data are circa 38% of the 2328 inventors in the full data.

a random sample of (almost) 100K control individuals from those same firms.

These individuals represent some 5% of the Finnish working age population.

In 1991, we have 82 184 individuals in our sample of whom 843

• btain at least one USPTO patent between 1990 and 1999.

For 1999, we have 94 806 individuals of whom 882 have obtained at least one USPTO patent between 1990 and 1999.

Wage Income Growth (1)

10 20 30 40 50 60 70 80 90 income percentiles

Wage Income Growth (1990‐1999) by Percentiles

non‐inventors inventors

Wage Income Growth (2)

90 92 94 96 98 income percentiles

Wage Income Growth (1990‐1999) by Percentiles

non‐inventors inventors

Capital vs Labor Income in 1999

90 92 94 96 98

Inventor/Non‐inventor Ratio by Type of Income in 1999

capital income ratio wage income ratio

Transition Matrix

Table 1: Transitions 1991 to 1999 non-inventors 1991 / 1999 top-10=0 top-10=1 Conditional Prob. top-10=0 88.05 4.17 4.51 top-10=1 2.34 5.45 69.96 inventors 1991 / 1999 top-10=0 top-10=1 Conditional Prob. top-10=0 41.95 19.61 31.86 top-10=1 7.60 30.84 80.23

Transition Matrix by Father’s Education

Table 2: Transitions 1991 to 1999 conditional on father’s education Father’s education < 12 years non-inventors inventors 91 / 99 top10=0 top10=1 C/Pr 91 / 99 top10=0 top10=1 C/Pr top10=0 86.55 5.13 5.60 top10=0 44.81 19.10 29.88 top10=1 2.41 5.91 71.03 top10=1 6.84 29.25 81.07 Father’s education ≥12 years 91 / 99 1 C/Pr 91 / 99 top-10=0 top-10=1 C/Pr top10=0 88.24 4.05 4.39 top10=0 39.24 20.85 34.70 top10=1 2.36 5.35 69.30 top10=1 8.07 31.84 79.78

Transition Matrix by Gender

Table 3: Transitions 1991 to 1999 conditional on gender Female non-inventors inventors 91 / 99 1 Con Pr 91 / 99 top-10=0 top-10=1 Con Pr top10=0 95.73 2.02 2.07 top-10=0 67.78 11.11 14.08 top10=1 0.87 1.38 61.33 top-10=1 1.11 20.00 94.74 Male 91 / 99 1 Con Pr 91 / 99 top-10=0 top-10=1 Con Pr top10=0 84.37 5.22 5.83 top-10=0 39.37 20.76 34.53 top10=1 3.07 7.34 70.51 top-10=1 8.35 31.52 79.06

Transition Matrix by Age

Table 4: Transitions 1991 to 1999 by age (inventors only) < median age 1991 / 1999 top-10=0 top-10=1 Conditional Prob. top-10=0 47.19 26.53 35.99 top-10=1 5.10 21.17 80.56 > median age 1991 / 1999 top-10=0 top-10=1 Conditional Prob. top-10=0 38.98 14.29 26.83 top-10=1 9.39 37.35 79.93

Transition Matrix by Innovation Quality

Table 5: Transitions 1991 to 1999 by quality of invention < 20 citations 1991 / 1999 top-10=0 top-10=1 Conditional Prob. top-10=0 43.60 17.08 28.15 top-10=1 8.15 31.18 79.29 ≥20 citations 1991/1999 top-10=0 top-10=1 Conditional Prob. top-10=0 35.78 38.53 51.85 top-10=1 2.75 22.94 89.30

Labor Income in 1999

• 0.1132
• 0.0516
• 0.0331

Labor Income in 1999

0.15 0.27 0.42 0.87 1‐9 10‐19 20‐29 30+ citation counts

Percentage Increase in Wage (relative to 0‐cited)

Conclusion

Overall, our findings suggest avenues for further research on (innovation-led) growth, inequality and social mobility.

Analyze how factors such as innate ability, family situation, gender, education, parental education/income, affect the probability for an inventor to make it to top income brackets.

Policy implications: e.g., how do we factor in *innovation* when designing tax policy and combining with entry policy, patent policy,... to achieve more inclusive innovation-driven growth?

Go deeper into how institutions affect the relationship between innovation, top income inequality, and social mobility.

Royal Economic Society