IFS Real Options, Patents, Productivity and Market Value November 2002 Nicholas Bloom (Institute for Fiscal Studies) John van Reenen (Institute for Fiscal Studies & UCL) Summary Part 1: Patents Data IFS • There is a consensus that technological advance is crucial in the “new economy” • Patents provide a powerful indicator of this technology • We hand-match patents from over 12,000 assignees to 450 UK parent firms. • Using this dataset we show a strong and significant effect of patents on – Productivity – Market Value • Patent citations are also shown to informative Summary Part 2: Real Options IFS • We use this data to test new “Real Options” theories • Embodying new technology requires heavy investment, training and marketing. • When firms patent technologies they have the option to see how market conditions develop • This generates patenting real options • Hence, higher uncertainty will lead to a more gradual technology take up • This turns out to be empirically significant Previous Patenting Work IFS • Toivanen, Stoneman and Bosworth (1998) and Bosworth, Wharton and Greenhalgh (2000) find patenting effects on market value in UK firms. • Griliches (1981), Hall (1993), and Hall, Jaffe and Tratjenberg (2001) report effects on market value in US firms. • Greenhalgh, Longland and Bosworth (2000) report a positive employment effect of patenting in UK firms. Patents Data IFS • We constructed the new IFS-Leverhulme dataset using patenting, accounting and financial data. • The patenting data was hand matched from the 12,000 largest US PTO patenting assignees to their UK parent companies. • The remaining 128,000 patenting subsidiaries were then computer matched – which is less accurate. • This provides reliable firm level patenting information from 1968 to 1993 on the UK and Overseas subsidiaries of about 200 UK firms IFS Patents Data frequency of patents by year no. patents from that year 3000 2000 1000 0 1960 1967 1970 1980 application year application year of the patent 1990 1994 1996 IFS Patents Data The distribution of firms by total patents: 1968-96 Firms >1 >10 >25 >100 >250 >1000 236 161 117 75 41 12 The Top 8 UK Patenting Firms ICI 8422 Shell 7200 SmithKline Beecham 3672 BP 3632 BTR 3432 Lucas Industries 3119 GEC 3054 Hanson 2892 IFS Citations Data • Citations provide a proxy of patent values, which appear to be extremely variable. • This allows us to fine tune our raw patent counts Histogram of number of cites per patent .3 frequency .2 .1 0 012345 10 20 30 tot no. cites 40 50 IFS Citations Data The Five Most Cited Patents Patent Topic Shell Grand Metropolitan ICI Synthetic Resins Microwave heating package Herbicide compositions Unilever Anticalculus composition British Oxygen Pharmaceutical Corp. Treatment Grant Year 1972 Cites 1976-96 221 1980 174 1977 130 1977 97 1975 89 IFS Citations Data • But the lag between patenting and citing can lead to truncation biases when using citation weights Lag from patenting to citation citing frequency .1 .05 0 0 1 2 3 4 5 10 20 lag lag in years 35 IFS Citations Data • We correct for these truncation biases in citations data using a Fourier series estimator Actual and Normalizing Mean Total Cites Per Patent 15 10 5 0 1960 1980 application year 2000 IFS The IFS-Leverhulme Dataset • We match patents with Datastream accounting data Median Mean Min. Max. Capital (1985 £m) 143 744 1.6 18,514 Employment (1000s) 8,398 24,374 40 312,000 Sales (1985 £m) 362 1,224 1.15 20,980 Market Value (1985 £m) 153 740 0.29 19,468 Patents 3 12.6 0 409 Patent Stock 10 42.6 0 1218 Cite Stock 49.2 202 0 5157 Uncertainty 1.39 1.47 0.60 6.6 Observations Per Firm 22 20 3 29 Patenting & Productivity IFS • Standard production models (see Griliches, 1990) usually assume Cobb-Douglas production y AGa K bLc where: G is knowledge stock, K is capital, and L is labour • We proxy he knowledge stock using the stock of patents (PAT) built up using the perpetual inventory method. • This allows us to estimate “ a ” – the return to patents ln( y) ln( A) a ln( PAT ) b ln( K ) c ln( L) • Using patent citations allow us to fine tune our knowledge stock measure IFS Productivity Equation Results Sales All Firms Patenters Capital 0.333 * 0.436 * 0.438 * 0.468 * Employment 0.650 * 0.558 * 0.554 * 0.502 * Patent Stock 0.024 * Citation Stock 0.030 * No. Firms 2063 211 211 189 No. Obs. 18,068 2219 2219 1896 0.468 * 0.502 * -0.012 0.039* 189 1896 Notes: A full set of firm and time dummies is included. All coefficient marked * are significant at the 1% level All variables are in logs. Estimation covers 1968-1993. Patenting and Market Value IFS • The effect of patents on firm performance can also be measured using forward looking market values • Following Griliches (1981), Bosworth, Wharton and Greenhalgh(2000), and Hall et al (2000) we use a Tobin's Q functional form. V PAT log( ) a ( ) K K where V Tobin' s Q log( ) K IFS Market Value Results Patent Stock/Capital Citation Stock/Capital No. Firms No. Obs. Log Tobin’s Q (log(V/K)) 1.620* -0.352* 205 2053 0.427* 0.491 * 182 1748 182 1748 Notes: A full set of firm and time dummies is included. All coefficient marked * are significant at the 1% level All variables are in logs. Estimation covers 1968-1993. Patents and Real Options IFS • Bertola (1988), Pindyck (1988), Dixit (1989) and Dixit and Pindyck (1994) first noted the importance of real options in generating investment thresholds for individual projects. • Abel and Eberly (1996) and Bloom (2000) extend this theory to show how real options lead firms to be cautious in responding to demand shocks. • This cautionary effect of real options on investment has been shown empirically by Guiso and Parigi (1999) and Bloom, Bond and Van Reenen (2001). Modeling Patents & Real Options IFS • To model this caution effect of real options we define “G” as the firms potential knowledge stock and “Ge” as its embodied knowledge • We can then define the elasticity of embodied to actual knowledge as Ge G l ( ) a G Ge where l ( ) 0 • Higher uncertainty leads to a lower elasticity of embodiment – a slower pass through of patents into production Modeling Patents & Real Options IFS • We prove that the effect of total patents (PAT) will be positive • But the effect of new patents on productivity will be reduced by higher uncertainty - the caution effect • The direct effects of uncertainty will be ambiguous. • Interestingly, while this is true for productivity, market values are forward looking. • To investigate these effects we add in uncertainty levels and interaction effects. Our Uncertainty Measure IFS • Our uncertainty measure is the average daily share returns variance of our firms over the period • Using a firm specific time invariant uncertainty measure matches the underlying theory • This share returns uncertainty measure has been used before by Leahy and Whited (1998) and Bloom, Bond and Van Reenen (2001). IFS Mean Daily Returns Standard Deviation (%) Our Uncertainty Measure 2.5 2 1.5 1 1970 1980 Year 1990 Notes: This is the unweighted mean of our measure of the standard deviation of daily returns over the year. Mean Daily Share Returns – our entire sample Patent Real Options Results IFS Tobin’s Q Real Sales Capital 0.451* 0.446* Employment 0.517 * 0.553* Patent Stock 0.025* 0.038* Uncertainty -0.036* Uncertainty Tobin’s Q Uncertainty Pat. Stock -0.015* 0.297* -0.010* Tobin’s Q 0.913* 1.743* -0.265^ -0.073 Firm Dummies No Yes No Yes No. Firms 211 211 205 205 No. Obs. 2053 2053 2037 2037 Notes: All coefficient marked * and ^ are significant at the 1% and 10% level All variables are in logs. Estimation covers 1968-1993. Conclusion IFS • Patents appear to play an important role in determining productivity and market value • But their impact on productivity is delayed when higher uncertainty reduces the rate of technological embodiment • Hence, micro and macro stability could play a large role in encouraging technological development.