Dont Hide Your Light under a Bushel Innovative

Don’t Hide Your Light under a Bushel: Innovative Originality and Stock Returns David Hirshleifer UC Irvine Po-Hsuan Hsu University of Hong Kong Dongmei Li University of South Carolina

Motivation – Psychological Evidence • Limited attention: – Individuals pay less attention to information that is hard to process • View toward complex signals: – People are more suspicious about more complex signals – They view the subject riskier than it actually is • e. g. , Alter and Oppenheimer (2006), Song and Schwarz (2008, 2009, 2010)

Motivation – Valuation of Innovation • Hard – technical uncertainty – long road from concept to actual profits – market risk and competitor risk • Evidence suggesting market misvaluation of innovation proxies – Chan, Lakonishok, and Sougiannis (2001), Eberhart, Maxwell, and Siddique (2004), Hirshleifer, Hsu, and Li (2012), Cohen, Diether, and Malloy (2012), among others

Question and Hypotheses • Does market fully impound originality of a firm’s innovative activities (IO)? • If IO is + indicator of future fundamentals, limited attention (in spirit of models of Hirshleifer & Teoh 2003, Hirshleifer, Lim & Teoh 2011) – Higher IO predicts higher abnormal returns – Predictability is stronger among firms with • More valuation uncertainty • Lower investor attention • Higher sensitivity of fundamentals to IO

Empirical Findings • IO and fundamentals – High-IO firms have lower contemporaneous ROA/ROE – High-IO predicts significantly higher future ROA/ROE • IO and future abnormal returns – High IO predicts significantly higher future abnormal returns – Predictability much stronger among younger, more opaque, lower investor attention, larger, and growth firms

Measure of Innovative Originality Popular view of invention recombinant search • E. g. , Gilfillan 1935; Schumpeter 1939; Basalla 1988; Weitzman 1996; Henderson and Clark 1990. • Invention comes from • Combining technological components in novel manner • Reconfiguring existing combinations • Example/analogy: • Steamship — combination of boat/steam engine • Behavioral finance

Empirical Proxy of Innovative Originality Breadth/Diversity of knowledge drawn upon by an invention • Patent level • # tech classes contained in a patent’s reference list (N) • Firm level • Average N across a firm’s recently granted patents

Empirical Proxy of IO (cont’d) • Common in innovation/corporate finance literature • E. g. , Trajtenberg, Henderson & Jaffe (1997), Hall, Jaffe & Trajtenberg (2001), Lerner, Sorensen & Strömberg (2011), and Custodio, Ferreira & Matos (2013). • Additional supporting evidence • High IO predicts higher future citations received per patent and better future fundamentals

Proxies of Conditioning Variables • Valuation uncertainty (VU) • Firm age (Kumar 2009): an inverse proxy • Opacity of financial reports (Hutton, Marcus & Tehranian 2009) • VU Index = Standardized opacity – Standardized age • Investor attention • Analyst-to-shareholder ratio (ATS) • PEAD/Earnings surprise (Hirshleifer, Lim & Teoh 2009): an inverse proxy • Attention Index = Standardized ATS – Standardized PEAD/Earnings surprise

Proxies of Conditioning Variables (cont’d) • Sensitivity of fundamentals to IO • Size • Large firms benefit more from high IO • E. g. , Schumpeter (1950); Acs and Audretsch (1987); Cohen, Levin & Mowery (1987) • Book-to-Market (BTM) • Values of growth firms (low BTM) derive more from IO • Sensitivity Index = Standardized size – Standardized BTM • Verification of the validity of these proxies: • Regressing future ROA/ROE on IO • Significantly higher slopes on IO among large firms, low BTM firms.

Empirical Test Design • Portfolio sorts – Industry- and characteristics-adjusted returns – Alphas relative to standard risk factor models • Fama-Mac. Beth regressions – Industry effects – Other return predictors such as: • Size, book-to-market, momentum, innovative efficiency (IE), patents, R&D intensity, institutional ownership, stock illiquidity, short-term return reversal, asset growth, net stock issuance, capital investment, ROA, idiosyncratic volatility (IVOL) and skewness.

IO of Selected Industries Industry (FF 48) Mean Healthcare 7. 01 Medical equipment 7. 38 Pharmaceutical products 6. 59 Chemicals 6. 53 Machinery 5. 95 Electrical equipment 5. 78 Automobiles and trucks 5. 10 Aircraft 4. 93 Shipbuilding, railroad equipment 5. 54 Telecommunications 6. 32 Personal services 4. 73 Business services 7. 88 Computers 6. 26 Electronic equipment 5. 92 Measuring and control equipment 6. 25 Transportation 7. 00 Stdev 4. 48 4. 91 4. 41 3. 83 3. 38 3. 92 2. 45 2. 08 2. 63 4. 16 4. 22 5. 43 3. 75 3. 82 4. 42 Min 1. 00 1. 00 2. 00 1. 00 P 30 4. 60 4. 47 4. 33 4. 61 4. 00 3. 71 3. 83 3. 74 4. 00 4. 48 2. 50 5. 00 4. 08 4. 00 P 50 6. 04 5. 50 5. 80 5. 12 5. 00 4. 92 4. 75 4. 66 5. 03 4. 00 6. 60 5. 50 5. 00 5. 33 6. 00 P 70 8. 00 8. 35 7. 00 7. 25 6. 93 6. 50 5. 75 5. 77 5. 90 6. 38 5. 00 8. 83 7. 25 6. 50 7. 26 9. 00 Max 27. 00 43. 92 42. 47 38. 00 27. 25 49. 06 20. 50 12. 65 14. 40 31. 00 24. 00 66. 00 36. 61 46. 50 31. 23 22. 00

Summary Statistics of IO Portfolios IO Rank Obs No 3270 Low 419 Middle High IO Size (mn) BTM ROA (%) ROE (%) 642 0. 83 1. 68 1. 00 3. 02 1154 0. 74 1. 32 1. 50 5. 37 4334 0. 70 3. 05 4. 40 409 9. 78 2033 0. 64 0. 29 0. 80

IO and Next Year’s ROE IO ROE ΔROE MTB Adv. Ex Cap. Ex R&D IE R 2 Model 1 Model 2 Estimate (%) t-stat 1. 42 (9. 03) 0. 94 (7. 96) 25. 01 (24. 32) 24. 89 (24. 36) -4. 48 (-11. 03) -4. 41 (-10. 93) -1. 60 (-2. 02) -1. 56 (-1. 95) 0. 72 (4. 75) 0. 71 (4. 65) 0. 58 (1. 59) 0. 62 (1. 67) -2. 38 (-7. 04) -2. 52 (-7. 32) 1. 10 (5. 35) 0. 37 Results even stronger for next five-year average ROE and are robust for ROA.

Return Predictive Power of IO IO Rank Exret No Low Middle 0. 60 Indadjret Charadjret Alpha MKT SMB HML UMD -0. 04 -0. 03 -0. 13 0. 98 0. 18 0. 07 0. 03 (2. 54) (1. 37) (2. 38) (-1. 10) (-0. 13) (-2. 07) (59. 50) (7. 03) 0. 48 -0. 12 (1. 74) (-1. 73) -0. 13 (-0. 53) -0. 13 1. 00 0. 17 -0. 12 -0. 03 (-1. 49) (40. 60) (4. 96) (-2. 54) (-1. 02) 0. 73 0. 07 0. 16 0. 97 -0. 15 -0. 17 -0. 02 (2. 83) (0. 76) (0. 31) (2. 49) (53. 59) (-6. 07) (-5. 75) (-1. 10) 0. 80 0. 09 (3. 09) (1. 86) 0. 19 (0. 76) 0. 24 0. 95 0. 01 -0. 13 -0. 04 (2. 74) (44. 26) (0. 21) (-3. 08) (-1. 64) High-Low 0. 32 0. 22 (2. 62) (2. 43) 0. 32 (2. 67) 0. 37 -0. 05 -0. 16 -0. 02 -0. 01 (3. 12) (-1. 58) (-3. 39) (-0. 26) (-0. 34) High Results robust to controlling for other risk factors.

Valuation Uncertainty (VU) and Return Predictive Power of IO Age Exret Ind-adjret Char-adjret 4 F Young 0. 75 0. 68 0. 80 0. 81 0. 82 0. 89 0. 83 0. 79 (2. 83) (2. 79) (3. 05) (2. 66) (2. 60) (2. 90) (2. 69) (2. 66) 0. 20 (1. 50) 0. 08 (0. 87) 0. 21 (1. 58) 0. 25 (1. 72) 0. 25 (1. 70) 0. 28 (1. 92) 0. 16 (1. 10) Old 4 F + IMC 4 F + INV Opacity Exret Ind-adjret Char-adjret 4 F High 0. 57 0. 33 0. 62 0. 80 0. 79 0. 68 0. 66 (1. 81) (1. 19) (1. 99) (2. 15) (2. 13) (1. 84) (1. 86) 0. 27 (1. 25) 0. 14 (0. 90) 0. 41 (1. 84) 0. 29 (1. 07) 0. 30 (1. 09) 0. 26 (0. 97) 0. 33 (1. 20) 0. 29 (1. 11) Ind-adjret Char-adjret 4 F 0. 69 0. 61 0. 72 0. 70 0. 79 0. 71 0. 64 (2. 39) (2. 32) (2. 54) (2. 01) (2. 23) (2. 04) (1. 93) 0. 19 (1. 47) 0. 11 (1. 17) 0. 21 (1. 60) 0. 27 (1. 88) 0. 26 (1. 83) 0. 26 (1. 78) 0. 30 (2. 07) 0. 19 (1. 37) Low VU Index Exret High Low 4 F + IMC 4 F + INV 4 F + LIQ 4 F + EMI Returns, alphas for high-low IO portfolios

Investor Attention and Return Predictive Power of IO Attention proxied by analyst-to-shareholder ratio (ATS) ATS Exret Ind-adjret Char-adjret 4 F 4 F + IMC 4 F + INV 4 F + LIQ 4 F + EMI Low High 0. 34 0. 23 0. 39 0. 37 0. 42 0. 32 (2. 61) (2. 46) (2. 57) (2. 96) (2. 86) (2. 79) (3. 11) (2. 36) 0. 12 (0. 72) 0. 07 (0. 43) 0. 18 (1. 11) 0. 12 (0. 67) 0. 14 (0. 76) 0. 18 (1. 01) 0. 14 (0. 77) 0. 12 (0. 65) Attention proxied by Post-Earnings-Announcement-Drift / Earning Surprise (PEAD) PEAD Exret High 0. 83 0. 64 0. 68 0. 75 0. 73 0. 72 0. 67 (3. 96) (3. 47) (3. 17) (3. 30) (3. 25) (3. 14) (3. 09) (2. 94) 0. 14 (0. 53) 0. 04 (0. 19) 0. 28 (1. 04) 0. 34 (1. 12) 0. 32 (1. 03) 0. 40 (1. 28) 0. 20 (0. 66) Low Ind-adjret Char-adjret 4 F 4 F + IMC 4 F + INV 4 F + LIQ 4 F + EMI Attention proxied by the composite index Index Exret Low 0. 60 0. 39 0. 54 0. 70 0. 69 0. 71 0. 72 0. 56 (3. 61) (3. 26) (3. 24) (4. 25) (4. 17) (4. 16) (4. 32) (3. 54) 0. 06 (0. 36) 0. 06 (0. 46) 0. 12 (0. 75) 0. 13 (0. 75) 0. 12 (0. 70) 0. 15 (0. 87) 0. 13 (0. 76) High Ind-adjret Char-adjret 4 F 4 F + IMC 4 F + INV 4 F + LIQ 4 F + EMI

Sensitivity of Future Profitability to IO and Return Predictive Power of IO 4 F 4 F + INV Size Exret Big 0. 46 0. 35 0. 42 0. 53 0. 52 0. 56 0. 41 (2. 86) (3. 11) (2. 70) (3. 19) (3. 08) (3. 12) (3. 32) (2. 50) 0. 00 (-0. 03) -0. 06 (-0. 61) 0. 03 (0. 34) -0. 08 (-0. 81) -0. 08 (-0. 76) -0. 07 (-0. 65) -0. 07 (-0. 68) -0. 06 (-0. 60) 4 F 4 F + IMC 4 F + INV Small Ind-adjret Char-adjret 4 F + IMC BTM Exret Low 0. 35 0. 25 0. 34 0. 42 0. 41 0. 42 0. 31 (2. 44) (2. 35) (2. 32) (2. 87) (2. 79) (2. 84) (2. 79) (2. 14) 0. 25 (1. 38) 0. 11 (0. 82) 0. 09 (0. 51) 0. 16 (0. 68) 0. 16 (0. 70) 0. 11 (0. 49) 0. 21 (0. 92) 0. 21 (0. 94) 4 F 4 F + IMC 4 F + INV High Ind-adjret Char-adjret 4 F + LIQ 4 F + EMI Sensitivity Index Exret High 0. 35 0. 27 0. 36 0. 41 0. 40 0. 44 0. 31 (2. 45) (2. 56) (2. 52) (2. 76) (2. 67) (2. 69) (2. 90) (2. 06) 0. 05 (0. 36) 0. 03 (0. 23) -0. 01 (-0. 04) 0. 07 (0. 47) 0. 06 (0. 42) 0. 07 (0. 48) 0. 06 (0. 41) 0. 11 (0. 75) Low Ind-adjret Char-adjret 4 F + LIQ 4 F + EMI

Average IO IO Rank Low Middle High Firm Age Young Old 2. 97 3. 05 5. 39 5. 36 10. 33 9. 32 IO Rank Low Middle High Analyst-to-Shareholder Low High 3. 08 3. 03 5. 37 9. 26 10. 06 Size IO Rank Low Middle High Big 3. 36 5. 35 8. 48 Small 2. 95 5. 37 10. 13 High 3. 54 6. 37 12. 03 Opacity Low 3. 61 6. 38 11. 40 PEAD/Earnings Surprise High Low 3. 56 3. 43 5. 82 5. 79 10. 08 10. 37 Book-to-Market Low High 3. 06 2. 97 5. 38 5. 36 9. 90 9. 61 VU Index High Low 2. 98 3. 04 5. 39 5. 36 10. 38 9. 33 Attention Index Low High 3. 05 3. 06 5. 37 9. 75 9. 57 Sensitivity Index High Low 3. 11 2. 91 5. 37 5. 36 9. 67 9. 99 Return predictive power of IO not driven by difference in IO spreads.

Average Size (in Millions) IO Rank Low Middle High Firm Age Young Old 532 1754 872 6287 581 3356 IO Rank Low Middle High Analyst-to-Shareholder Low High 2073 861 7873 1568 3777 1025 Size IO Rank Low Middle High Big 10413 20405 15578 Small 131 146 136 High 1233 4224 2066 Opacity Low 2290 9182 4403 PEAD/Earnings Surprise High Low 3320 1994 8554 5448 4549 2210 Book-to-Market Low High 1550 707 6101 1885 2699 1007 VU Index High Low 565 1679 1048 6062 704 3100 Attention Index Low High 1255 1589 4546 5280 2161 2589 Sensitivity Index High Low 2016 125 6542 155 3228 122 Return predictive power of IO not driven by small firms.

IE vs. IO IE Rank L IO Exret Indadjret High-Low 0. 14 0. 12 Charadjret 0. 05 4 F 4 F + IMC 4 F + INV 4 F + LIQ 4 F + EMI 0. 16 0. 21 0. 13 (0. 66) (0. 71) (0. 24) (0. 78) (0. 79) (0. 99) (0. 61) M High-Low 0. 14 (0. 61) 0. 08 0. 18 0. 21 0. 22 (0. 42) (0. 77) (0. 95) (1. 00) 0. 20 0. 28 (0. 87) (1. 22) 0. 15 (0. 68) H High-Low 0. 75 (2. 43) 0. 63 0. 67 0. 58 0. 55 (2. 20) (2. 12) (2. 02) (1. 90) 0. 64 0. 53 (2. 21) (1. 76) 0. 69 (2. 31) Average IO IO Rank IE Rank L M H H-L L 2. 94 5. 36 9. 69 6. 74 M 3. 32 5. 34 9. 02 5. 70 H 3. 26 5. 43 9. 96 6. 71

Fama-Mac. Beth Regressions (I) IO Control Variables Model 1 0. 16 BTM, size, momentum, institutional ownership, (7. 22) illiquidity, short-term return reversal Above variables plus innovative efficiency, patents-to. Model 2 0. 08 assets, R&D/ME, ROA, investment, net stock issuance, (2. 93) idiosyncratic volatility, skewness, sales diversity

Fama-Mac. Beth Regressions (II) Firm Age IO Young Slope 0. 12 t-stat (2. 56) Old 0. 06 (1. 90) Analyst-to-Shareholder IO Low Slope 0. 08 t-stat (1. 99) High 0. 02 (0. 46) Size IO Big Slope 0. 17 t-stat (3. 15) Small 0. 11 (3. 03) Opacity High 0. 17 (3. 16) Low -0. 01 (-0. 21) PEAD/Earnings Surprise High 0. 13 (1. 74) Low 0. 06 (0. 64) Book-to-Market Low 0. 11 (2. 90) High 0. 04 (0. 89) VU Index High 0. 17 (4. 62) Low 0. 04 (1. 85) High. Low 0. 13 (3. 15) Attention Index Low High 0. 11 -0. 02 0. 13 (2. 81) (-0. 44) (2. 41) Sensitivity Index High Low 0. 22 0. 05 0. 17 (3. 47) (1. 30) (2. 29) IO effect much stronger among firms with more valuation uncertainty, lower investor attention, and higher sensitivity of profitability to IO.

Additional Robustness Tests • Fama-French five-factor model (market, size, value, investment, profitability factors) • Subperiod analysis • Transaction costs • Alternative IO measure based on generalized Herfindahl index

Alternative Explanations Risk • Information asymmetry • Financing constraints • Obsolescence • Investment-specific technological change

Summary and Conclusion • We document a positive relation between firms’ IO and future fundamentals and abnormal returns. • Relation stronger among firms with – Lower investor attention – Higher valuation uncertainty – Higher sensitivity of future profitability to IO. • Overall, evidence suggests that IO is a useful input for firm valuation, and evidence is consistent with limited investor attention and skepticism of complexity.