1. Vivian Fang, Carlson School of Management, University of Minnesota Xuan Tian, Kelly School of Business, Indiana University Sheri Tice, A.B. Freeman School of Business, Tulane University 1

2. Motivation  There is a longstanding academic and policy debate about the effect of liquidity on managerial myopia.  Managerial myopia is defined as the underinvestment in long-term, intangible projects for the purpose of meeting short-term goals. 2

3. Motivation (Cont’d) First view  High stock liquidity exacerbates managerial myopia.  Stein (1988, 1989) and Shleifer and Summers (1988)  In the presence of information asymmetry between managers and investors, takeover pressure could induce managerial myopia.  Takeover pressure also gives managers less power over shareholders.  Liquidity increases the probability of a hostile takeover (Kyle and Vila 1991).  Porter (1992) and Bhide (1993)  Long-term, intangible investments tend to depress short-term earnings.  “Non-dedicated owners” may exit in response to depressed earnings.  High liquidity encourages “Wall Street Walk”, resulting in a drop in price. 3

4. Motivation (Cont’d) Second view  High stock liquidity mitigates managerial myopia.  Maug (1998)  Predicts more monitoring activities by blockholders in highly liquid firms.  Admati and Pfleiderer (2009), Edmans (2009) and Edmans and Manso (2011)  A blockholder actively collects information about fundamental value and impounds it into stock price through trading in a liquid market.  Efficient stock price reflects fundamental information rather than short- term earnings. 4

5. Research Question  “Managerial myopia is difficult to test because it results in underinvestment in activities that are difficult to observe.” Stein (2003)  Lack of innovation is a critical symptom of managerial myopia.  We use observable innovation outputs (patents and citations) to capture managerial myopia (or lack thereof). Does stock liquidity enhance or impede firm innovation? 5

6. Main Findings  We show a negative causal effect of stock liquidity on innovation productivity, using a diff-in-diff approach.  Surrounding decimalization of the minimum tick size in 2000-2001  Surrounding the shift in the minimum tick size in 1997  More pronounced for pilot firms that converted to decimals in 2000  Possible mechanisms: firms with a larger, exogenous increase in liquidity surrounding decimalization have  A high probability of facing a hostile takeover  An increase in the holdings of non-dedicated institutional investors 6

7. Contribution  We shed light on the longstanding academic and policy debate.  This is the first paper in the literature that provides causal evidence that stock liquidity impedes firm innovation.  We uncover a previously under-identified adverse consequence of regulatory effort to enhance stock liquidity. 7

8. Contribution (Cont’d)  Add to the literature on managerial myopia and innovation  Aghion, Van Reenen, and Zingales (2013); Lerner, Sorensen, and Stromberg (2011); Ferreira, Manso, and Silva (2012); Atanassov (2012); Chemmanur and Tian (2012)  Add to the literature linking liquidity to firm outcomes  Fang, Noe, and Tice (2009); Bharath, Jayaraman, and Nagar (2012); Norli, Ostergaard, and Schindele (2010); Edmans, Fang, and Zur (2013) 8

9. Sample Selection  Data sources  NBER patent database: Patent and citation data  TAQ database: Intraday trades and quotes  Thomson’s 13f database and Brian Bushee’s website: Institutional ownership and classification  Compustat: Financial data  We end up with  39,469 firm-year observations between 1994 and 2005 9

10. Key Variables  Measuring innovation productivity (Hall, Jaffe, and Trajtenberg 2001, 2005)  Number of filed patents (and eventually granted), adjusted for application-grant lag: INNOV_PAT t+n  Number of non-self citations each patent receives, adjusted for citation-lag: INNOV_CITE t+n  Measuring stock liquidity (Fang, Noe and Tice 2009)  A stock illiquidity measure: ILLIQ t, log of the annual RESPRD 10

11. Control Variables  LN_MV, the natural logarithm of firm market capitalization  ROA, return-on-assets ratio  RDTA, R&D expenditure over total assets  PPETA, net PPE scaled by total assets  LEV, total debt to total assets ratio  CAPEXTA, capital expenditure scaled by total assets  HINDEX (HINDEX 2 ), Herfindahl index based on annual sales  Q, Tobin’s Q  KZINDEX, Kaplan and Zingales (1997) five-variable KZ index  LN_AGE, the natural logarithm of one plus the number of years the firm is listed on Compustat. 11

12. Summary Statistics 12 Variable5%25%MedianMean75%95%SDN INNOV_PAT0.000 0.7920.0005.5491.95039,469 INNOV_CITE0.000 0.6200.0003.3861.19239,469 ILLIQ-6.573-5.363-4.377-4.482-3.557-2.6441.20839,469 LN_MV2.4574.1035.4685.6046.9639.2542.03639,469 RDTA0.000 0.0550.0610.2500.12939,469 ROA-0.2950.0440.1140.0780.1720.2840.18539,469 PPETA0.0160.0890.2090.2850.4300.7890.24239,469 LEV0.0000.0180.1710.2090.3390.5910.20239,469 CAPEXTA0.0030.0210.0430.0620.0790.1940.06339,469 HINDEX0.0000.0160.0670.1420.1920.5480.19239,469 Q0.7841.0751.4582.1122.3315.8221.86239,469 KZINDEX-42.38-5.353-0.682-9.0000.8942.62731.1239,469 LN_AGE0.6931.6092.3032.2923.0913.8070.97939,469

13. OLD Regressions 13 The baseline (with # of patents as D.V.) has the following specification: INNOV_PAT i,t+n = a + bILLIQ i,t + cLN_MV i,t + dRDTA i,t + eROA i,t + fPPETA i,t + gLEV i,t + hCAPEXTA i,t + iHINDEX i,t +jHINDEX 2 i,t + kQ i,t + lKZINDEX i,t + mLN_AGE i,t +YR t +FIRM i + error i,t (1) Innovation measured by INNOV_PAT Dependent Variables INNOV_PAT t+1 INNOV_PAT t+2 INNOV_PAT t+3 ILLIQ t 0.141 *** 0.168 *** 0.170 *** (0.020)(0.023)(0.026) Year and Firm FE Included Number of Obs. Used39,46933,09827,363 Adjusted R 2 0.8390.8440.849

14. OLS Regressions (Cont’d) 14 The baseline (with # of citations as D.V.) has the following specification: INNOV_CITE i,t+n = a + bILLIQ i,t + cLN_MV i,t + dRDTA i,t + eROA i,t + fPPETA i,t + gLEV i,t + hCAPEXTA i,t + iHINDEX i,t +jHINDEX 2 i,t + kQ i,t + lKZINDEX i,t + mLN_AGE i,t +YR t +FIRM i + error i,t (2) Innovation measured by INNOV_CITE Dependent Variables INNOV_CITE t+1 INNOV_CITE t+2 INNOV_CITE t+3 ILLIQ t 0.104 *** 0.106 *** (0.015)(0.016)(0.019) Year and Firm FE Included Number of Obs. Used39,46933,09827,363 Adjusted R 2 0.6520.653

15. Identification Diff-in-Diff analysis using decimalization 15  We make use of an exogenous shock to stock liquidity: decimalization  Based on ΔRESPRD t-1 to t+1, we sort 3,375 sample firms into terciles.  We match firm from the top and from the bottom tercile, using a one-to- one nearest neighbor propensity score matching, without replacement.  PAT (CITE): sum of # of patents (citations per patent) in the 3-year window before or after decimalization. Mean treatment difference (after - before) Mean control difference (after - before) Mean DiD estimator (treat - control) T-statistics for DiD estimator PAT-5.169-1.682-3.487 ** -2.265 (1.103)(1.074)(1.540) CITE-11.14-8.522-2.616 ** -1.976 (0.986)(0.884)(1.324)

16. Identification (Cont’d) Diff-in-Diff analysis using decimalization 16  The figure on the left (right) shows the average innovation captured by the mean number of patents (citations per patent) for treatment and control firms, from 3 years before decimalization to 3 years after decimalization. Decimalization year is denoted as year 0.

17. Identification (Cont’d) Diff-in-Diff analysis using the 1997 shock 17  We make use of another exogenous shock to stock liquidity: the shift in minimum tick size from 1/8 th to 1/16 th in 1997  Similarly, based on ΔRESPRD t-1 to t+1, we sort sample firms into terciles and apply a one-to-one nearest neighbor propensity score matching.  PAT (CITE): sum of # of patents (citations per patent) in the 3-year window before or after decimalization. Mean treatment difference (after - before) Mean control difference (after - before) Mean DiD estimator (treat - control) T-statistics for DiD estimator PAT-1.9732.621-4.595 ** -1.976 (0.797)(2.185)(2.326) CITE-9.065-4.360-4.706 ** -2.177 (1.806)(1.189)(2.162)

18. Identification (Cont’d) Diff-in-Diff analysis comparing pilot and non-pilot stocks 18  The conversion on the NYSE was completed in 4 phases.  Phase 1: 7 issues in July 2000  Phase 2: 52 firms (representing 57 issues), starting September 25, 2000  Phase 3: Additional 94 securities on December 4, 2000  Phase 4: the rest of the non-pilot securities listed on the NYSE in January 2001

19. Identification (Cont’d) Diff-in-Diff analysis comparing pilot and non-pilot stocks 19  PILOT: a dummy to indicate pilot firms (phase 1-3)  YR_2000: equals 1 for year 2000 and 0 for year 1999  PILOT×YR_2000 is an interaction term of the two. INNOV_PAT i,t+1 (INNOV_CITE i,t+1 ) = a + bPILOT i × YR_2000 + c PILOT i + d YR_2000 + e’CONTROLS i,t + IND j + error i,t (3) (1)(2) Dependent Variables INNOV_PAT t+1 INNOV_CITE t+1 PILOT i ×YR_2000 -0.485 ** -0.309 * (0.213)(0.164) PILOT i 0.2890.313 * (0.243)(0.166) YR_2000 -0.0140.091 (0.165)(0.097) Controls and Industry FE Included Number of Obs. Used 2,160 Adjusted R 2 0.5500.481

20. Underlying mechanism Takeover pressure 20  Stein (1988, 1989): In the presence of information asymmetry, shareholders tend to undervalue the stocks of companies investing in innovative projects. This leads to a higher probability of the firm facing a hostile takeover. In view of this, managers tend to put more effort in short-term projects that offer quicker returns instead of investing in long-term innovative projects.  Shleifer and Summers (1988): managers have less power over shareholders when takeover threats are high, which leads to fewer incentives to invest in activities with only long-run payoffs.  Kyle and Vila (1991) argue that high liquidity increases a firm’s exposure to takeovers.  Thus, takeover exposures could be an underlying economic mechanism through which stock liquidity impedes firm innovation.

21. Underlying mechanism (Cont’d) Takeover pressure 21  Hostile Takeover is the average probability of being a target in a hostile takeover, in the three-year window before or after decimalization, with the probability being the predicted value of TARGET based on the coefficients estimated in the logit regression of TARGET i,t+1 = a + bQ i,t + cPPETA i,t + dLN_CASH i,t + eBLOCK i,t + fLN_MV i,t + gINDMA_DUM i,t + hLEV i,t + iROA i,t + YR t + IND j + error i,t, where TARGET is a dummy variable equal to one if the company is target of an attempted or completed hostile acquisition and zero otherwise. Similarly, All Takeover is the average probability of being a target in any takeover, in the three-year window before or after decimalization. Mean treatment difference (after - before) Mean control difference (after -before) Mean DiD estimator (treat - control) T-statistics for DiD estimator Hostile Takeovers0.2120.0350.177 *** 5.036 (0.021)(0.024)(0.032) All Takeovers0.0400.0190.022 * 1.828 (0.008)(0.009)(0.012)

22. Underlying mechanism (Cont’d) Non-dedicated institutional investors 22  Porter (1992) argues that investment in long-term, intangible assets tends to depress short-term earnings. He stresses that transient shareholders may exit in response to a low quarterly earnings report and quasi-indexers have little or no incentives to monitor. If managers have incentives to keep the stock price high, they may cut investment in long-term projects to boost short-term profits.  This effect should be more pronounced when liquidity is high because high liquidity makes it easier to exit (Bhide (1993)).  Bushee (1998) highlights the possibility of cutting R&D expenditures as a way to reverse earnings decline, especially when transient institutional ownership is high.  Thus, pressure exerted by non-dedicated institutional investors could be another channel through which liquidity impedes firm innovation.

23. Underlying mechanism (Cont’d) Non-dedicated institutional investors 23  Bushee (1998, 2001) classification: Transient: relatively high diversification, high turnover, frequent momentum trading Quasi-indexer: high diversification, low turnover, index type buy-and-hold strategies Dedicated: high concentration, low turnover, no trading sensitivity to current earnings  TRAPCT, QUAPCT, and DEDPCT is the average institutional holdings (%) held by transient institutional investors, quasi-indexers, and dedicated institutional investors in the 3-year window before or after decimalization Mean treatment difference (after - before) Mean control difference (after -before) Mean DiD estimator (treat - control) T-statistics for DiD estimator TRAPCT0.040-0.0120.052 *** 11.42 (0.004)(0.002)(0.005) QUAPCT0.0640.0090.055 *** 9.127 (0.005)(0.004)(0.006) DEDPCT0.0130.0070.0051.586 (0.002)(0.003)

24. Other Robustness Checks  We estimate Eq. (1) within each of the Fama-French 12 industry  Positive coefficient on ILLIQ in 11 industries and 6 are significant  The OLS results are robust to  Using different liquidity proxies including relative quoted spread, Amihud (2002) illiquidity ratio, and PIN measure of Easley, Kiefer, and O'Hara (1997)  Controlling for M&A deal size or removing firms involved in M&As  Partitioning sample into size quartiles  Interacting liquidity with time effects 24

25. Conclusion  Using a diff-in-diff approach and exploiting the variation in stock liquidity generated by two exogenous shocks (the decimalization of the minimum tick size in 2001 and the shift in minimum tick size from $1/8th to $1/16th in 1997), we show stock liquidity has a causal negative effect on firm innovation.  There are least two possible underlying mechanisms.  High liquidity makes firms more prone to hostile takeover pressure.  High liquidity attracts transient investors who trade frequently to chase current profits or quasi-indexers who follow passive indexing strategies and fail to govern. 25