Demographics and Innovation€¦ · Demographics and Innovation François Derrien Ambrus Kecskés Phuong-Anh Nguyen Workshop on Technology and Aging Workforce Seoul, May 2018

Demographics and Innovation

François Derrien

Ambrus Kecskés

Phuong-Anh Nguyen

Workshop on Technology and Aging Workforce

Seoul, May 2018

Motivation

• Population is aging consequences for the economy?

• Younger people have characteristics that are crucial for innovation

– Risk taking

– Longer investment horizons

– Creativity

– Interactivity

• Young labor force Innovation Growth

• Question: Are younger labor forces more innovative?

2

This paper

• Link between age structure of the local labor force and innovation

• Mechanism: Labor supply

– In younger areas, firms hire from a larger pool of young employees

More innovative inventors and non-inventors

More innovative firms

• To disentangle this from other channels: Three levels of analysis

– Commuting zone

• Most general, to establish and quantify the link

• Where people live and work (commonly used in labor economics)

– Firm

• Allows to test for alternative channels

– Inventor

• Allows to explore interaction effects

3

Our approach

• Actual population is endogenous to contemporaneous economic

activity

• People migrate because of contemporaneous economic conditions

• Actual population = Native-born population + (Immigration –

Emigration)

• Native-born population

– Projected population using historical births

– Births 20-64 years before are plausibly exogenous to economic activity

(e.g., innovation) decades later

Use this to estimate the causal link between population age structure and

innovation

4

Measurement of age structure

• For example, in 1990

– Look back 20-64 years to 1926-1970

• Collect births each year, adjust for survival

• This is the native-born labor force in 1990

• Calculate measures of age structure in 1990

– mean age

– young share

• = population aged 20 to 39 / Population aged 20 to 64

• Repeat i 99 , 99 , …, 5

– Stop in 2005 because patent data end in 2010

– Really just 1990, 1995, 2000, and 2005

5

Evolution of the mean age of the projected and

actual labor forces by location, 1990-2005

6

Native born labor force

• Similar age structures for actual and projected labor forces

– Correlation of about 0.5

• Different location, different time trends

• But: Remove time trends, not much time-series variation left

We identify off cross-sectional variation

– Want to absorb sources of common variation in innovation and age

structure Use many FEs and control variables

7

Our methodology and main finding

Illustrating figure – year 2000

8

What if we use actual instead of projected

age of the labor force?

9

Commuting-zone-level tests

• Specification

Innovationc,s,t+1 = α.age structurec,t + β.controlsc,t + αs,t

• Controls

– Possible source of endogeneity: Omitted variables

– Correlated with innovation today and births 20-64 years ago

• Variables related to economic conditions

– Population (scale effects in innovation, urban vs. rural areas)

– Income per capita (wealth)

– Growth rate of total income (growth)

• Long-term consequences of past investments (public or private)

– Local government expenditures

– Edu atio % o er 5 ith a helor’s degree or higher

– Patents of local university

– Variables measured at date t

• Results robust to averaging them over the last 20 years

• Clustering at the state-year level

10

Commuting-zone-level tests – Results

11

A one-standard deviation change in mean age (2.3 years)

leads to a 25-30% increase in patents

Firm-level tests

• Specification

Innovationi,j,a,c,s,t+1 = α.age structurec,t + β.controlsi,t + αs,t + αj,t + αa

• Age structure: in The CZ where the firm is headquartered

• Allows us to study

– Industry composition effects (innovative industries locate in younger areas): αj,t

– Firm life cycle effects (younger firms locate in younger areas): αa

– Alternative channels

• Financing supply

• Consumer demand

• Controls

– Standard corporate finance controls Listed firms only

• Robust to including CZ-level controls

• Firm location

– Using headquarter location from Compustat

– Robust to using historical HQ location or location at IPO date

12

Firm-level tests - Results

13

Alternative mechanisms

• Financing supply channel

– Younger areas have younger investors, who prefer equity

Easier to obtain financing for innovative firms

– Look at the link between innovation and age structure

• At the fir ’s HQ, here the fir o tai s is fi a i g

• At the fir ’s R&D hu s, here it produ es its i o atio

– Stronger results at R&D hubs

• Consumer demand channel

– Younger areas have younger consumers, who prefer innovative products

Local firms more innovative in younger areas

– Compare firms in non-tradable industries (in which production and

consumption happen at the same place) vs. tradable industries (in which they

do ’t

– Results only for tradable industries

14

Inventor-level tests

• Specification

Innovationi,j,k,c,s,t+1 = α.age structurec,t + β.controlsi,k,t + αs,t + αj,t + αi

• Focus on stars (top 5% based on patent counts in prior 10 years)

– Dominate production of innovation (majority of patent counts and citations)

– Data sparse for non-stars

• Controls

– Inventor level

• I e tor’s age si e first pate t , pate t sto k

– Firm level

• Firm age (since first patent), patent stock, # of inventors at the same R&D hub, total # of

inventors, total # of R&D hubs

• Less precise than in firm-level tests because we use data from private firms

• But we use firm FEs (αi)

• Allows us to study separately

– The effe t of i e tor’s age s. fir age

– The effect of i e tor’s age s. the age structure of the environment

15

Inventor-level tests - Results

16

Value implications

• Younger labor forces help firms create valuable growth

opportunities

• Market prices should reflect these growth opportunities

• Age structure Firm valuations

– Use the Market-to-Book of (public) firms as a measure of valuation

– Same specification as in firm-level tests

17

Value implications

18

Summary

• Local age structure affects innovation

– At the CZ level

– At the firm level

– At the inventor level

• Labor supply channel

– Firms in younger locations are able to hire younger, more

innovative employees, which enables them to produce more

innovation

• This affects the long-term growth opportunities of firms

and therefore their value

19