Transcript Document
Patents as an Appropriation Mechanism URL: http://socrates.berkeley.edu/~scotch/cohenetal.pdf Growth matters • First half of the 20th century: real GDP per unit of labor grew at 2.75%. • Reminder: the rule of 70. At growth rate a%, value doubles in about 70/a years. • After the first 25 years, output per person was twice as high, and after 50 years four times as high. • In the last quarter of the 20th century, growth has slowed down, but never below about 1.75%. 300 250 200 Industry R&D Federal R&D Total R&D 150 100 50 1995 1988 1981 1974 1967 1960 0 1953 billions of 1996 dollars R&D Spending: Total, Federal and Industry 250000 200000 Industry R&D 150000 Federal R&D Total R&D 100000 Domestic Patent Grants 50000 1995 1988 1981 1974 1967 1960 0 1953 millions of $1996, domestic patent grants 300000 s R&D Spending and Domestic Patent Grants Q F ( L, C , K , T ) Each of L,C,K,T grows. Change per unit time: Growth Accounting L, C , K , T Q FL L FC C FK K FT T Wages satisfy: where FL, FC are the marginal product of L,C Q wL L wC C FK K FT T Rearrange: wL FL , wC FC Q wL L L wC C C 1 [ FK K FT T ] Q Q L Q C Q =wage share of GDP Q L C 1 F K F T K T Q L C Q =capital share of GDP Subtract the wage and capital growth shares from growth in GDP to get the portion of growth not accounted for by L,C. About half!! R&D, Patents and Productivity • The growth residual (after accounting for labor and capital growth) is about half. But is it linked to the growth in “knowledge”? Yes, if knowledge is “accumulated R&D.” Private R&D spending has a bigger effect on GDP than public spending. Why? • Rate of return on R&D spending is greater than rate of return on capital (stock market). Is this what you expect? Yes, because…. • Suppose the rates of return were the same. Is R&D spending too high? Too low? R&D and Patenting • R&D spending is an input. Patents are an output. What is it that we really want to measure? Can we measure innovation? (no) Welfare (GDP?) is what we care about in any case. • Nevertheless, it might be of interest to record: 1. 2. 3. 4. • Total patent grants (not domestic US grants) have grown more or less commensurately with industrial spending in late 20th century. (Foreign patent grants grew from 18% to half in postwar.) Cost of patents is $1m-$2m (1996 dollars) Rate of return on R&D spending is greater than rate of return on capital (stock market). Is this what you expect? Suppose the rates of return were the same. Is R&D spending too high? Too low? The value side: Several Methodologies Patent Values are skewed. Does skewness matter? What matters for incentives? Scherer and Harhoff http://socrates.berkeley.edu/~scotch/skew.pdf • First two rows: survey evidence • Next rows: royalty data from technology licensing offices. • Penultimate rows: Stock value • Drugs: Sales Evidence from Renewal Data: R(t) C(t) time t What do we learn from renewals? • Gives a distribution of value of patent rights. (Distinguish from value of innovation.) • Most of the value is in the top few patents. • Half of patents “end” at about year 10. • Only about 10-15% of R&D spending is recovered from patent value (24% of private spending) Final Methodology: Estimate the effect of patents on firm’s stock value • Three regressors that affect firm value: patent counts, R&D spending, patent citations, • Should help us distinguish between the value of the innovation and the value of a patent right. • Early studies used patent counts and R&D. The R&D variable reduces the value of the patent right to ¼ its previous value. • Patent citations are a better predictor of value than patent counts, but still less good than R&D spending. when g K A is small Empirical question: For which types of K is the coefficient g larger?