## wpiea2021158-print-pdf

## Source details

**Canonical URL:** [wpiea2021158-print-pdf](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021158-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021158-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021158-print-pdf.pdf.json)

---

### I. Introduction — research question and approach
- Research question: effect of passive institutional ownership (index funds, passive institutional investors = PIO) on portfolio firms’ technological innovation.
- Innovation characterized as long-term, risky, and subject to agency problems; hypothesis space:
  - Monitoring view: institutional investors strengthen monitoring, raising innovation.
  - Myopia/incentives view: passive ownership induces risk aversion/myopia, reducing exploration.
- Empirical identification: exploit plausibly exogenous variation from annual Russell index reconstitutions (discontinuity at Russell 1000/2000 cutoff) to isolate causal impacts of PIO on innovation.

### Institutional background and identification strategy
- Russell indices:
  - Russell 3000: 3,000 largest U.S. companies by market cap.
  - Russell 1000: top 1,000 (large-cap); Russell 2000: bottom 2,000 (small-cap).
  - Index membership determined by market capitalization on last trading day in May; index weights floata djusted end of June.
- Mechanism for exogenous variation:
  - Firms crossing the rank threshold face forced buying/selling by index-tracking funds; top Russell 2000 firms receive higher index weights than bottom Russell 1000 firms.
- Sample and design choices:
  - Sample period: "1984 through 2006" (observations after 2006 dropped due to Russell “banding” policy).
  - Bandwidth: "300" firms on either side of the cutoff.
  - Resulting sample size: "7,035" firm-year observations.
  - McCrary (2008) manipulation test: average annual t statistic = "0.44" for bandwidth 300 (not significant), suggesting no precise manipulation of market cap around cutoff.
- Empirical framework: two-stage least squares with R2000_i,t as instrument for PIO_i,t; outcomes Y_i,t+n for n = 1, 2, 3 years after reconstitution; controls include firm and year fixed effects; standard errors clustered by firm.

### Data sources and key variable construction
- Data sources: FTSE Russell, Thomson Reuters Institutional Holdings (13F), CRSP, Compustat, NBER Patent Citation database, USPTO BDSS, Capital IQ (product releases from 1995), ISS/RiskMetrics, Execucomp.
- Institutional ownership definitions:
  - IO = proportion institutional holdings to total common shares outstanding.
  - PIO = percentage of quasi-indexers (passive institutions per Bushee (2001)) in total shares outstanding.
  - TIO = transient institutional ownership; DIO = dedicated institutional ownership.
- Innovation measures and constructions:
  - LnPat = log(1 + # patents); LnCite = log(1 + # citations).
  - ClaimNo, ClaimLength: log transforms of claim counts and claim word counts.
  - Impact labeling: top 10 percent of citations = "Breakthrough"; no citation = "Failure"; otherwise = "Incremental".
  - NewField, Originality, Generality, Product (new products) defined and sourced as in appendix.
- Controls and transformations: ln(Assets), ln(Age), R&D, Capex, PPE, Lev, ROA, HHI, KZ, Tobin’s Q; variables winsorized at top and bottom "1 percent percentiles".

### Main empirical findings — innovation activity (causal effects)
- First-stage: Russel 2000 assignment produces a discontinuity in PIO around cutoff; reported R2000 coefficients for bandwidths ±100, ±200, ±300, ±500: "0.046**", "0.041***", "0.028***", "0.030***". Reported increase in PIO due to inclusion in Russell 2000: depending on spec, a "3–4 percent" increase.
- Second-stage 2SLS (PIO → innovation outcomes):
  - Reported elasticities three years after reconstitution (Year+3):
    - R&D: "5.374**" interpreted as a 1 percent exogenous increase in passive institutional holdings → "5.4 percent" increase in R&D expenses.
    - LnPat: "4.832**" interpreted as a 1 percent PIO increase → "4.8 percent" increase in patent counts.
    - LnCite: "3.997*" interpreted as a 1 percent PIO increase → "4.0 percent" increase in patent citation counts.
  - Selected Year+1 coefficients: R&D Year+1 = "6.838***"; LnPat Year+1 = "4.336**"; LnCite Year+1 = "5.003**".
- Timing: changes in innovation observable as early as one year after PIO changes; discontinuity plots show significant jumps in patent counts and citation counts per patent three years after reconstitution.

### Main empirical findings — innovation strategy, composition, and quality
- Patent claim measures (Table 4 Panel A):
  - ClaimNo: PIO coefficient = "6.678***".
  - ClaimLength: PIO coefficient = "9.332***".
  - Interpretation: more and longer patent claims after PIO increases; consistent with marginal, incremental contributions in crowded fields.
- Citation-distribution categories:
  - Incremental: PIO coefficient = "5.621*" (positive and significant).
  - Breakthrough: PIO effect insignificant.
  - Failure (uncited): PIO effect insignificant.
  - Interpretation: PIO increases associate with more moderately cited (incremental) patents, not more breakthroughs.
- Technological scope and commercialization (Table 4 Panel B):
  - NewField: PIO = "3.926" (insignificant).
  - Originality: PIO = "-1.655**" (negative and significant).
  - Generality: PIO = "-1.462*" (negative and significant).
  - Product (new products): PIO = "-0.440*" (negative and significant).
  - Interpretation: PIO increases reduce patent originality and generality and reduce conversion of patents into new products — firms shift toward exploitation of current knowledge rather than exploration of new fields.

### Channels: monitoring and incentive mechanisms
- Monitoring channel evidence:
  - Vote participation and voting patterns (ISS/RiskMetrics voting analytics; voting sample starts 2003):
    - PIO → Vote Participation: "1.422***" (0.473).
    - PIO → Vote For Percentage: "1.136**" (0.576).
    - PIO → Vote Against Percentage: "0.287" (0.324) (insignificant).
    - PIO → Disagree with Management: "-1.203**" (0.580).
    - Interpretation: passive investors increase vote participation, cast more “For” votes, and exhibit less disagreement with management; increased monitoring plausibly induces greater managerial effort and observable innovation inputs/outputs.
  - Quantified statement in text: a "1.4 percent" increase in vote participation associated with PIO increase.
- Incentive channel evidence:
  - CEO turnover and performance sensitivity (Table 6 Panel A):
    - PIO coefficients on Forced turnover across columns: "7.941**" (3.880); "8.247**" (3.891); "9.090**" (4.094); "8.107**" (3.889); "8.089**" (3.897).
    - Interaction terms significant and positive: PIO*∆ROA = "0.585***" (0.158); PIO*IndROA = "0.405**" (0.189); PIO*∆Ret = "0.500***" (0.168); PIO*IndRet = "0.640***" (0.156).
    - Vega (risk-taking incentive): PIO = "2.337" (2.623) (insignificant).
  - Managerial entrenchment (Table 6 Panel B):
    - Independent Board: PIO = "13.443***" (4.720). Economic translation: increasing PIO from the 25th to the 75th percentile increases probability of an independent board by "20 percent".
    - Severance Agreement: PIO = "-11.255*" (6.494). Economic translation: increasing PIO from 25th to 75th percentile decreases probability of severance contracts by "10 percent".
  - Interpretation: PIO increases raise forced-turnover probability and performance-turnover sensitivity, increase board independence, and reduce severance protections; no evidence that PIO raises Vega — managers face higher replacement risk but not higher risk-taking compensation, encouraging risk-averse, exploitative innovation choices.

### Robustness and falsification checks
- Robustness dimensions (Table 7):
  - Alternative bandwidths (±100, ±200, ±300, ±500, Imbens–Kalyanaraman optimal) yield positive and significant PIO effects on LnPat and LnCite; selected LnPat PIO estimates: "5.160**" (2.498) for ±100; "5.198***" (1.853) for optimal; "4.307***" (1.648) for ±500; "7.087***" (2.269) for another spec.
  - Alternative polynomial orders κ = 1,2,3 produce LnPat PIO: "4.336**" (2.121) for κ=1; "3.787**" (1.714) for κ=2; "4.679**" (1.856) for κ=3.
  - Alternative Russell rankings and PIO definitions: end-of-June ranking and including quasi-indexers plus transient investors produce positive and significant effects (examples: PIO "1.407*" (0.839) for LnPat; PIO "2.706**" (1.324) for alternate PIO).
  - Placebo cutoffs at 500, 750, 1500, 1750 produce statistically insignificant PIO coefficients on LnPat (examples: "-1.724" (3.017); "-0.713" (8.183); "-9.004" (17.759); "43.525" (46.202)), supporting validity of identification.
- Manipulation check: McCrary test statistic "0.44" not significant.
- Conclusion: baseline results robust to bandwidths, polynomial orders, alternative definitions, and placebo tests.

### Key summary statistics (selected)
- Sample period: "1984–2006".
- Sample size: "7,035" firm-year observations.
- Means, medians, standard deviations (selected):
  - LnPat — N "7,035" — Mean "0.732" — Median "0.000" — StDev "1.199".
  - LnCite — N "7,035" — Mean "0.883" — Median "0.000" — StDev "1.346".
  - ClaimNo — N "7,035" — Mean "0.984" — Median "0.000" — StDev "1.379".
  - ClaimLength — N "7,035" — Mean "1.445" — Median "0.000" — StDev "1.982".
  - Breakthrough — N "2,463" — Mean "0.685" — Median "0.693" — StDev "0.836".
  - Incremental — N "2,463" — Mean "1.828" — Median "1.609" — StDev "1.087".
  - Failure — N "2,463" — Mean "0.388" — Median "0.000" — StDev "0.733".
  - NewField — N "2,463" — Mean "1.906" — Median "1.792" — StDev "1.203".
  - Originality — N "2,085" — Mean "0.533" — Median "0.559" — StDev "0.232".
  - Generality — N "1,896" — Mean "0.549" — Median "0.580" — StDev "0.243".
  - Product — N "3,808" — Mean "0.016" — Median "0.000" — StDev "0.110".
  - IO — N "7,035" — Mean "0.505" — Median "0.536" — StDev "0.293".
  - PIO — N "7,035" — Mean "0.320" — Median "0.330" — StDev "0.197".
  - TIO — N "7,035" — Mean "0.136" — Median "0.110" — StDev "0.117".
  - DIO — N "7,035" — Mean "0.033" — Median "0.010" — StDev "0.058".
  - LnAssets — N "7,035" — Mean "6.342" — Median "6.326" — StDev "1.132".
  - LnAge — N "7,035" — Mean "3.131" — Median "3.219" — StDev "0.727".
  - R&D — N "7,035" — Mean "0.030" — Median "0.000" — StDev "0.064".
  - Capex — N "7,035" — Mean "0.075" — Median "0.056" — StDev "0.074".
  - PPE — N "7,035" — Mean "0.556" — Median "0.486" — StDev "0.378".
  - Lev — N "7,035" — Mean "0.460" — Median "0.464" — StDev "0.206".
  - ROA — N "7,035" — Mean "0.155" — Median "0.149" — StDev "0.108".
  - HHI — N "7,035" — Mean "0.018" — Median "0.009" — StDev "0.025".
  - KZ — N "7,035" — Mean "-5.738" — Median "-0.686" — StDev "17.514".
  - TobinsQ — N "7,035" — Mean "2.225" — Median "1.582" — StDev "2.188".
- Additional averages:
  - On average, a firm applies for and is eventually granted "5.7" patents a year.
  - Each patent receives "7.3" forward citations on average.
  - An average patent has "6.1" claims (text truncated in source).

### Descriptive patterns and interpretation
- Distribution skewness: mean > median for patent and citation counts, indicating innovation concentrated in most productive firms.
- In the final sample, approximately "35 percent" ("2,463") of firms applied for and were eventually granted at least one patent.
- For patenting firms averages: "14.1" patents in new classes; "2.2" patents in top 10 percent; "1.4" patents never cited; "11.3" appear in middle of citation distribution.
- Institutional ownership composition (average firm): institutional investors hold "50.5 percent" of shares; within institutional holdings: "6.5 percent" dedicated, "26.9 percent" transient, "63.4 percent" quasi-indexers.
- Typical firm characteristics: listed for "29.1 years"; book value of assets = "$1.1 billion"; R&D to asset ratio = "3.0 percent"; Capex to asset ratio = "7.5 percent"; PPE to asset ratio = "55.6 percent"; leverage = "46.0 percent"; ROA = "15.5 percent"; KZ = "-5.7"; Tobin’s Q = "2.2".

### Policy-relevant implications and open questions posed by authors
- Empirical summary:
  - Increased PIO associated with higher R&D, patent counts, and citation counts, but a shift toward exploitation (incremental, patentable projects) rather than exploration (breakthroughs, new fields) and reduced commercialization into new products.
  - Mechanisms: intensified monitoring (increased voting participation and shareholder oversight) and altered managerial incentives (higher forced-turnover risk, greater board independence, reduced severance) that increase effort but promote risk-averse choices; no evidence PIO raises Vega (risk-taking pay).
- Policy questions raised verbatim in source:
  - Is the observed shift to exploitation good or bad for overall firm performance?
  - Will the shift affect aggregate productivity?
  - Should corporate governance guidelines be revisited to reflect the ongoing rise of passive investors?
  - Will a new framework be required that combines better corporate law, financial regulation and structural reforms as the passive investing industry becomes increasingly dominated by giant institutional investors including Blackrock, Vanguard and State Street?
- Authors indicate these are important topics for future research and policy-making.

*Source: wpiea2021158-print-pdf — https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021158-print-pdf.pdf*

### References _______________________________________________________________ 39

### I. INTRODUCTION

### Background and research question
- Rise of passive institutional ownership (index funds) is a major trend with potential implications for market efficiency and financial stability.
- Paper examines effect of passive institutional investors on portfolio firms’ technological innovation.
- Innovation is characterized as long-term, highly risky, and subject to agency problems; institutional investors may either strengthen monitoring (raising innovation) or induce risk aversion/myopia (reducing exploration).

### Key innovation concepts and measures
- Innovation activity: observable, countable inputs and outputs such as R&D expenditures, patent counts, and patent citations.
- Innovation strategy: scope, breadth, depth, and fundamental nature of the firm’s innovation effort (exploration vs. exploitation).
- Patent-based indicators used include number and length of patent claims, citation distribution (uncited, middle, highly cited), originality, generality, and conversion of patents into new products.
- Executive incentives metrics include CEO turnover, Vega of executive compensation, CEO power, board independence, and severance pay.

### Identification challenges
- Omitted variable concern: passive ownership and innovation may be driven by unobserved common characteristics.
- Reverse causality: expected changes in innovation may attract passive investors.
- Standard OLS regressions insufficient to establish causality.

### Empirical strategy (overview)
- Use plausibly exogenous variation from annual Russell index reconstitutions to establish causality between passive institutional ownership and firm innovation activity and strategy.
- Exploit discontinuity around Russell 1000/2000 threshold: firms just inside top of Russell 2000 have discontinuously higher passive ownership than firms at bottom of Russell 1000.
- Analyses focus on firms near Russell index threshold to isolate effects driven by varying passive institutional ownership.

### Institutional background: the Russell indices
- Russell 3000 Index: market-value-weighted index of the 3,000 largest companies listed in the United States.
- Russell 1000: top 1,000 firms (large-cap); Russell 2000: bottom 2,000 firms (small-cap).
- Index membership determined by firms’ market capitalization on the last trading day in May each year.
- Russell considers firms headquartered in the United States with stock price at or above $1.
- Market capitalization computed as closing share price times total number of all common shares outstanding (excludes preferred stock and non–freely floating stock); in multiple share-class cases uses “primary trading vehicle” market cap.
- Index assignments occur at end of May; index weights determined at end of June using float-adjusted market capitalization.

### Mechanism for exogenous variation
- If a firm moves from rank 1001 in May of year t-1 to rank 1000 in May of year t, it switches indices and faces forced buying or selling by index-tracking funds.
- Top Russell 2000 firms have large index weights; bottom Russell 1000 firms have trivial weights—leading to sizeable differences in passive ownership for firms near the threshold.
- Russell 2000 is the most popular Russell index with regard to assets benchmarked, implying more funds benchmarked to Russell 2000 than Russell 1000.

### Main empirical findings: innovation activity (causal effects)
- An exogenous increase in passive institutional ownership leads to a significant increase in:
  - R&D expenditures (innovation input).
  - Patent counts (innovation output).
  - Patent citation counts (innovation output).
- Results robust to alternative measures of institutional ownership, alternative bandwidths, alternative kernels, and placebo tests.

### Main empirical findings: innovation strategy (exploration vs. exploitation)
- Patent claim measures:
  - Both the number and the length of patent claims increase after an increase in passive institutional ownership, consistent with greater incremental research and marginal contributions to crowded fields.
- Citation distribution:
  - Citations to patents filed by firms with increased passive ownership tend to concentrate in the middle of the citation distribution with thinning tails.
  - Passive institutions have an insignificant effect on uncited and highly cited patents (breakthrough innovation) but a significantly positive effect on moderately cited (incremental) innovations.
- New-area patents:
  - Number of patents filed in new areas (fields where the firm has not patented before) does not change significantly.
- Patent quality and commercialization:
  - Originality and generality of patents shrink with higher passive ownership.
  - Fewer patents are turned into new products after passive institutional ownership increases.
- Net interpretation: firms shift toward exploitation of current knowledge rather than exploration into new fields when passive institutional ownership rises.

### Channels explored
- Monitoring channel:
  - Passive institutional investors intensify monitoring: increased participation in shareholder meetings and voting behavior that is favorable to management.
  - Increased oversight leads managers to expend more effort on innovation activities, raising observable inputs and outputs (R&D, patents, citations).
- Incentive channel:
  - Passive ownership increases probability of forced CEO turnover in case of poor financial performance or adverse market response.
  - Vega of executive compensation does not increase with passive institutional ownership (no increase in risk-taking incentives).
  - Passive investors reduce CEO power by increasing board independence and reducing severance pay.
  - Under greater career concerns and weaker risk-taking incentives, managers avoid risky exploration and favor exploitation that yields quicker and more certain returns.

### Contribution to literature
- Extends literature on institutional investors and corporate innovation by focusing specifically on passive institutional investors and distinguishing effects on innovation activities versus innovation strategies.
- Contrasts with previous studies on institutional ownership (including active and foreign institutional ownership, and hedge fund activism) by highlighting that passive ownership increases observable innovation activity but shifts strategy toward incremental, exploitive innovation rather than breakthrough, explorative innovation.

*Source: wpiea2021158-print-pdf — https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021158-print-pdf.pdf*

### 1000. Russell 1000 has to compete with S&P 500 for large-cap stocks, whereas Russell 2000

### wpiea2021158-print-pdf - 1000. Russell 1000 has to compete with S&P 500 for large-cap stocks, whereas Russell 2000

### Index context and identification strategy
- Russell 2000 had more assets benchmarked than Russell 1000 in 2018: "$1.2 trillion" tracked Russell 2000 while "only $500 billion" tracked Russell 1000.
- Institutional investors therefore hold a greater share of firms at the top of Russell 2000 compared to those at the bottom of Russell 1000; this difference arises from index composition rather than firm characteristics.
- Russell index reconstitution provides plausibly exogenous variation in passive institutional ownership (PIO) around the Russell 1000/2000 cutoff, enabling a regression discontinuity (RD) identification.

### Reconstructing end-of-May market capitalization and data processing
- Russell keeps end-of-May market capitalization proprietary and applies float adjustments in June; the authors reconstruct the end-of-May ranking used to assign index membership.
- Procedure to restore end-of-May market capitalization:
  - Use all 3,000 firms in the Russell 3000 universe.
  - Obtain end-of-May share prices from CRSP.
  - Obtain number of shares outstanding from Compustat quarterly shares outstanding item.
  - Use Compustat quarterly earnings report date to determine fiscal quarter closest to last trading day of May.
  - Use monthly CRSP factor to adjust the share item for corporate distributions after fiscal quarter-end and before end of May.
  - For missing price or share data, extract manually from Bloomberg.
  - When defining shares outstanding, choose the larger of the shares obtained from this procedure and the CRSP shares.

### Test of local randomness (manipulation check)
- Applied McCrary (2008) test on the distribution of the assignment variable (Russell market capitalization) on an annual basis.
- Result: Average annual t statistics for the McCrary test for a bandwidth of 300 firms is "0.44", not statistically significantly different from zero.
- Interpretation: Unable to reject the null that the density is continuous at the threshold; firms cannot precisely manipulate market capitalization to influence index assignment.

### Banding policy and sample restriction
- After 2006, Russell implemented a "banding" policy: a stock will remain in its original index if its market capitalization does not deviate far enough to warrant an index change.
- Because assignment after 2006 is no longer based exclusively on market capitalization, observations after 2006 are dropped to avoid introducing noise to the quasi-natural experiment.
- Sample period used: 1984 through 2006.

### Empirical framework and estimands
- First-stage specification (sharp RD interpretation):
  - R2000_i,t is a dummy equal to one if firm i is placed in the Russell 2000 index in year t, zero otherwise.
  - R_i,t signifies the market capitalization rank of firm i in year t minus 1,000, determined by end-of-May market capitalization.
  - F_i,t is the proxy for Russell index float adjustment, computed as the difference between the end-of-May market capitalization rank and the actual rank assigned by FTSE Russell in June.
  - PIO_i,t is passive institutional ownership of firm i in the next available quarter after the index reconstitution in year t.
- Two-stage least squares structure:
  - Stage 1: PIO_i,t = α + τ R2000_i,t + f(R_i,t) + γ F_i,t + δ X_i,t−1 + u_i + ν_t + ε_i,t
  - Stage 2: Y_i,t+n = β0 + β1 PIÔ_i,t + g(R_i,t) + β2 F_i,t + β3 X_i,t−1 + u_i + ν_t + ξ_i,t
  - Y_i,t+n denotes innovation outcomes in year n after reconstitution, where n = 1, 2, or 3.
  - f(·) and g(·) are parameterized as κ-th order polynomials, allowing different functional forms on either side of the cutoff.
  - Standard errors are clustered by firm.
- Outcome measures include R&D, patents, citations, claim counts and lengths, originality and generality scores, NewField, and new product releases.

### Data sources
- FTSE Russell: Russell index constituent information, index weights, and market capitalizations on the day of each reconstitution (available since 1984).
- Thomson Reuters Institutional Holdings (SEC 13F filings): institutional ownership; passive institutions identified using Bushee (2001) classifications.
- CRSP and Compustat: share prices and shares outstanding; merged quarterly matching for shares.
- NBER Patent Citation database and USPTO BDSS: patent- and citation-based innovation measures.
- Capital IQ Key Developments database: new product releases (data from 1995).
- ISS/RiskMetrics: shareholder voting and board independence.
- Execucomp: CEO turnover and compensation.
- Bushee (2001) classification used to label fund managers as "quasi-indexers", "transient investors", or "dedicated investors"; passive institutions indicator equals one if classified as "quasi-indexer".
- Institutional ownership (IO) = proportion of institutional holdings to total common shares outstanding.
- Passive institutional ownership (PIO) = percentage of quasi-indexers in total shares outstanding.

### Variable construction for innovation measures
- Patent count: total number of patents applied (and eventually granted) by a firm in a year.
- Citation count: total forward citations received by a firm scaled by number of patents.
- Log transforms used: LnPat = log(1 + # patents), LnCite = log(1 + # citations).
- Claim-level measures from parsed patent full texts: ClaimNo (average number of claims per patent per firm-year), ClaimLength (average word count of claims).
- Impact labeling: top 10 percent of citations in technology class = "Breakthrough"; no citation = "Failure"; otherwise = "Incremental".
- NewField: count of patents in technology classes where the firm had never filed before.
- Originality and Generality indices as defined:
  - Originality_i = 1 − Σ_j s_ij^2 / n_i
  - Generality_k = 1 − Σ_l s_kl^2 / n_k
- New product announcements hand-collected from Capital IQ (starts from 1995).

### Controls and transformations
- Controls: ln(Assets), ln(Age), R&D, Capex, PPE, Lev, ROA, HHI, KZ (Kaplan and Zingales, 1997), Tobin’s Q.
- Firm and year fixed effects included.
- All variables winsorized at the top and bottom "1 percent percentiles".

### Sample design and bandwidth
- Chosen bandwidth: "300" firms on either side of the index cutoff.
- Resulting sample size: "7,035" firm-year observations.
- Robustness: estimates shown robust to alternative bandwidths and Imbens and Kalyanaraman (2012) algorithm-selected optimal bandwidth.

### Key summary statistics (Table 1)
- Sample period: 1984–2006.
- Sample size: "7,035" firm-year observations.
- Selected variable summaries (Variable — N — Mean — Median — StDev):
  - LnPat — "7,035" — "0.732" — "0.000" — "1.199"
  - LnCite — "7,035" — "0.883" — "0.000" — "1.346"
  - ClaimNo — "7,035" — "0.984" — "0.000" — "1.379"
  - ClaimLength — "7,035" — "1.445" — "0.000" — "1.982"
  - Breakthrough — "2,463" — "0.685" — "0.693" — "0.836"
  - Incremental — "2,463" — "1.828" — "1.609" — "1.087"
  - Failure — "2,463" — "0.388" — "0.000" — "0.733"
  - NewField — "2,463" — "1.906" — "1.792" — "1.203"
  - Originality — "2,085" — "0.533" — "0.559" — "0.232"
  - Generality — "1,896" — "0.549" — "0.580" — "0.243"
  - Product — "3,808" — "0.016" — "0.000" — "0.110"
  - IO — "7,035" — "0.505" — "0.536" — "0.293"
  - PIO — "7,035" — "0.320" — "0.330" — "0.197"
  - TIO — "7,035" — "0.136" — "0.110" — "0.117"
  - DIO — "7,035" — "0.033" — "0.010" — "0.058"
  - LnAssets — "7,035" — "6.342" — "6.326" — "1.132"
  - LnAge — "7,035" — "3.131" — "3.219" — "0.727"
  - R&D — "7,035" — "0.030" — "0.000" — "0.064"
  - Capex — "7,035" — "0.075" — "0.056" — "0.074"
  - PPE — "7,035" — "0.556" — "0.486" — "0.378"
  - Lev — "7,035" — "0.460" — "0.464" — "0.206"
  - ROA — "7,035" — "0.155" — "0.149" — "0.108"
  - HHI — "7,035" — "0.018" — "0.009" — "0.025"
  - KZ — "7,035" — "-5.738" — "-0.686" — "17.514"
  - TobinsQ — "7,035" — "2.225" — "1.582" — "2.188"
- Additional reported averages (interpretation in text):
  - On average, a firm applies for and is eventually granted "5.7" patents a year.
  - Each patent receives "7.3" forward citations on average.
  - An average patent has "6.1" claims and each claim has (text truncated in source).

*Source: wpiea2021158-print-pdf*

### 22.6 words. The fact that the mean value of patent and citation counts is larger than the

### wpiea2021158-print-pdf - 22.6 words. The fact that the mean value of patent and citation counts is larger than the

### Descriptive statistics and sample composition
- Mean greater than median for patent and citation counts, indicating a highly skewed distribution of innovation activities concentrated in most productive firms.
- Approximately 35 percent (2463) of the firms in the final sample have applied for and been eventually granted at least one patent.
- For patenting firms, averages: 14.1 patents filed in new technology classes; 2.2 patents fall into the top 10 percent category; 1.4 patents are never cited; 11.3 appear in the middle of the citation distribution.
- Mean scores: patent originality = 0.53; patent generality = 0.55.
- Average number of new products introduced = 0.02.
- Institutional ownership composition (average firm): 50.5 percent institutional investors; of institutional investors, 6.5 percent dedicated institutions, 26.9 percent transient investors, 63.4 percent quasi-indexers (passive institutional investors by authors’ definition).
- Typical firm characteristics: listed for 29.1 years; book value of assets = $1.1 billion; R&D to asset ratio = 3.0 percent; capital expenditure to asset ratio = 7.5 percent; PPE to asset ratio = 55.6 percent; leverage ratio = 46.0 percent; return on assets ratio = 15.5 percent; KZ index = -5.7; Tobin’s Q = 2.2.

### A. Discontinuities in Passive Institutional Ownership
- Visual and regression discontinuity evidence shows a significant discontinuity in passive institutional ownership around the Russell 1000/2000 cutoff.
- Firms at the top of Russell 2000 have significantly higher passive institutional ownership than firms at the bottom of Russell 1000.
- First-stage 2SLS regression results (Table 2) for bandwidths ±100, ±200, ±300, ±500 report R2000 coefficients: 0.046**, 0.041***, 0.028***, 0.030*** respectively (standard errors reported in table).
- Reported increase in passive institutional ownership due to inclusion in Russell 2000: depending on specification, a 3–4 percent increase.

### B. Innovation Activity — causal effects of Passive Institutional Ownership (PIO)
- Second-stage 2SLS estimates (Table 3) show PIO coefficients positive and significant for innovation input and output.
- Reported elasticities (three years after reconstitution): a 1 percent exogenous increase in passive institutional holdings leads to:
  - 5.4 percent increase in R&D expenses (PIO coefficient Year+3 on R&D = 5.374**).
  - 4.8 percent increase in patent counts (PIO coefficient Year+3 on LnPat = 4.832**).
  - 4.0 percent increase in patent citation counts (PIO coefficient Year+3 on LnCite = 3.997*).
- Table 3 coefficient estimates (selected Year+1 and Year+3 PIO values): R&D Year+1 = 6.838***; R&D Year+3 = 5.374**; LnPat Year+1 = 4.336**; LnPat Year+3 = 4.832**; LnCite Year+1 = 5.003**; LnCite Year+3 = 3.997*.
- Figures 2 and 3 (discontinuity plots) show significant jumps in patent counts and citation counts per patent three years after index reconstitution for firms crossing from Russell 1000 to Russell 2000.

### C. Innovation Strategy — composition and quality of innovation
- Patent claim measures:
  - PIO positive and significant on ClaimNo = 6.678*** and ClaimLength = 9.332*** (Table 4 Panel A), indicating more and longer claims (narrower scope) after PIO increases.
- Citation-distribution categories:
  - PIO coefficients: Breakthrough (insignificant), Failure (insignificant), Incremental = 5.621* (positive and significant). Interpretation: PIO increases associate with more incremental innovations, not more breakthroughs.
- Technological scope and value (Table 4 Panel B):
  - NewField (patents in new technology classes): PIO = 3.926 (insignificant).
  - Originality score: PIO = -1.655** (negative and significant), indicating lower patent originality after PIO increases.
  - Generality score: PIO = -1.462* (negative and significant), indicating reduced generality (less cross-field influence).
  - Product introductions (new products): PIO = -0.440* (negative and significant), indicating fewer new products despite more patents.
- Overall interpretation:
  - PIO increases raise R&D, patent counts, and citation counts, but shift firms toward exploitation of existing technologies and incremental, patentable projects rather than exploration of new fields or pursuit of breakthrough innovation.
  - Changes in innovation are observable as early as one year after PIO changes.

### V. Plausible underlying channels
- Two channels proposed to explain results:
  - Monitoring channel: intensified monitoring by institutional investors increases managerial effort, raising innovation input and observable outputs (R&D, patents, citations).
  - Incentive channel: passive institutional investors alter managers’ incentives and career concerns, promoting risk-averse behavior that favors patentable, incremental projects over risky breakthrough research.
- Both channels tested in subsequent analyses (monitoring tests discussed next).

### A. Monitoring (empirical test preview)
- Shareholder voting participation used as an observable component of monitoring; ISS/RiskMetrics Voting Analytics database employed for mutual fund votes.
- Sample period for voting analysis begins in 2003 due to voting data availability.
- Instrumental-variable estimates reported in Table 5 (column(1) preview): PIO coefficient on vote participation is positive and significant at the 1 percent level; magnitude interpretation begins to quantify the association between a 1 percent increase in passive institutional holdings and vote participation (full coefficient and standard error in Table 5).

*Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021158-print-pdf.pdf*

### 1.4 percent increase in their vote participation. Fewer passive investors abstain, withhold, or

### wpiea2021158-print-pdf - 1.4 percent increase in their vote participation. Fewer passive investors abstain, withhold, or

### Monitoring Channel — passive institutional investor (PIO) voting and monitoring effects
- Main regression evidence (Table 5) on passive institutional investors’ voting behavior (standard errors clustered at firm level):
  - PIO coefficient on Vote Participation: 1.422*** (0.473).
  - PIO coefficient on Vote For Percentage: 1.136** (0.576).
  - PIO coefficient on Vote Against Percentage: 0.287 (0.324).
  - PIO coefficient on Disagree with Management: -1.203** (0.580).
  - Controls with notable coefficients: Lev 0.094*** (0.034); ROA -0.251** (0.105); Tobin’s Q 0.011** (0.005).
  - Observations: 1,259 for all columns; Adj. R2: 0.017, 0.015, 0.029, 0.017 respectively.
- Interpretations and mechanisms:
  - Passive investors increase vote participation (quantified above) and cast significantly more “For” votes across a range of issues.
  - Percentage of opposing votes does not change significantly (insignificant PIO on Vote Against Percentage).
  - Increase in PIO is associated with fewer shareholder-manager conflicts (negative and significant PIO on Disagree with Management).
  - Proposed explanations: passive funds vote based on proxy-voting recommendations (leading to higher support for proposals aligned with proxy protocols) and larger shareholdings reduce dissent; increased monitoring by institutional shareholders yields more managerial effort and observable innovation outcomes (e.g., more registered patents).

### Incentive Channel — CEO turnover, managerial entrenchment, and risk-taking incentives
- Probit and regression evidence (Table 6 Panel A — CEO turnover and compensation; standard errors clustered at firm level):
  - PIO coefficients on Forced turnover across columns: 7.941** (3.880); 8.247** (3.891); 9.090** (4.094); 8.107** (3.889); 8.089** (3.897).
  - Interaction terms (PIO × performance measures):
    - PIO*∆ROA: 0.585*** (0.158).
    - PIO*IndROA: 0.405** (0.189).
    - PIO*∆Ret: 0.500*** (0.168).
    - PIO*IndRet: 0.640*** (0.156).
  - Vega (measure of executives’ compensation sensitivity to volatility): PIO coefficient 2.337 (2.623) — not statistically significant.
  - Sample sizes and model specifications: Obs. vary (7,031; 5,940; 7,007; 3,119); Adj. R2 reported (examples: 0.057, 0.062, 0.059, 0.060, 0.063; Vega model Adj. R2 0.320).
- Probit evidence (Table 6 Panel B — managerial entrenchment):
  - Board independence (Independent Board): PIO 13.443*** (4.720).
    - Economic translation: increasing passive institutional holdings from the 25th to the 75th percentile increases probability of having an independent board by 20 percent.
  - Severance Agreement (dummy = 1 if firm has severance contract): PIO -11.255* (6.494).
    - Economic translation: increasing passive institutional holdings from the 25th to the 75th percentile decreases probability of implementing severance contracts by 10 percent.
  - Other notable coefficients: HHI -29.075*** (10.562) in Independent Board regression; Tobin’s Q 0.113*** (0.034).
  - Observations: Independent Board Obs. 2,443 (Adj.R2 0.171); Severance Agreement Obs. 1,919 (Adj.R2 0.080).
- Interpretations and mechanisms:
  - Higher PIO leads to higher probability of forced CEO turnover and greater performance-turnover sensitivity (interaction terms positive and significant).
  - No evidence that PIO increases managers’ risk-taking incentives via compensation Vega (insignificant coefficient).
  - PIO is associated with increased board independence and reduced severance agreements, weakening managerial entrenchment and increasing managers’ replacement risk — driving risk-averse behavior and a shift toward exploitative (incremental) innovation.

### Robustness checks — alternative specifications and falsification tests
- Robustness dimensions (Table 7):
  - Alternative bandwidths around the Russell 1000/2000 threshold (±100, ±200, ±500, and Imbens–Kalyanaraman optimal bandwidth):
    - PIO effects on LnPat remain positive and significant across bandwidths: 5.160** (2.498) for ±100; 5.198*** (1.853) for optimal; 4.307*** (1.648) for ±500; 7.087*** (2.269) for another bandwidth specification.
    - Observations vary: 2,284; 4,623; 11,925; 3,054. Adj. R2 examples: 0.290, 0.309, 0.276.
  - Alternative polynomial orders in two-stage least squares (κ=1,2,3):
    - PIO on LnPat: 4.336** (2.121) for κ=1; 3.787** (1.714) for κ=2; 4.679** (1.856) for κ=3. Obs. 7,035; Adj. R2 0.292 for each.
  - Alternative Russell rankings and alternative definitions of PIO (Panel C):
    - End-of-June ranking and including quasi-indexers plus transient investors produce positive and significant PIO effects on LnPat and LnCite: examples PIO 1.407* (0.839) for LnPat; PIO 2.706** (1.324) when alternate PIO definition used.
    - Obs. examples: 7,093; 7,035. Adj. R2 examples: 0.286, 0.269.
  - Placebo (falsification) thresholds (Panel D) at Cutoff=500, 750, 1500, 1750:
    - PIO coefficients on LnPat statistically insignificant (examples: -1.724 (3.017); -0.713 (8.183); -9.004 (17.759); 43.525 (46.202)), suggesting no discontinuity at placebo cutoffs.
- Summary of robustness:
  - Baseline results are robust to alternative bandwidths, polynomial orders, Russell ranking definitions, and alternative PIO definitions.
  - Falsification tests indicate treatment effects are absent at placebo thresholds, reducing likelihood that findings are spurious.

### Key empirical conclusions and policy-relevant implications
- Empirical findings:
  - Increased passive institutional ownership (PIO) is associated with:
    - Higher vote participation (PIO 1.422*** (0.473)) and more “For” votes.
    - A higher probability of forced CEO turnover (PIO coefficients ~7.9–9.1 with statistical significance).
    - Increased board independence (PIO 13.443*** (4.720)).
    - Decreased likelihood of severance agreements (PIO -11.255* (6.494)).
    - Increased observable innovation inputs and outputs (LnPat estimates positive and significant across robustness checks).
    - A shift in firms’ innovation strategy toward exploitation (incremental innovation) rather than exploration.
  - Mechanisms: enhanced monitoring (active voting) and increased managerial risk aversion due to higher replacement probabilities and weaker severance protection, not increased risk-taking incentives via compensation.
- Policy questions raised (listed verbatim as posed in the source):
  - Is the observed shift to exploitation good or bad for overall firm performance?
  - Will the shift affect aggregate productivity?
  - Should corporate governance guidelines be revisited to reflect the ongoing rise of passive investors?
  - Will a new framework be required that combines better corporate law, financial regulation and structural reforms as the passive investing industry becomes increasingly dominated by giant institutional investors including Blackrock, Vanguard and State Street?
- Suggested direction: these questions are indicated as important topics for future research and policy-making.

*Italic: Source — wpiea2021158-print-pdf (content excerpt).*

### REFERENCES

### REFERENCES

### References
- Acharya V., Baghai, R., Subramanian, K., 2013. Labor laws and innovation. Journal of Law and Economics 56, 997-1037.
- Aghion, P., Van Reenen J., Zingales L., 2013. Innovation and institutional ownership. American Economic Review 103, 277-304.
- Appel, I., Gormley, T., Keim, D., 2016. Passive investors, not passive owners. Journal of Financial Economics 121(1), 111-141.
- Argente, D., Baslandze, S., Hanley, D., Moreira, S., 2019. Patents to products: Innovation, product creation, and firm growth. Unpublished Working Paper.
- Balsmeier, B., Fleming, L., Manso, G., 2017. Independent boards and innovation. Journal of Financial Economics 123, 536–557.
- Bebchuk, L.A., Cohen, A., Ferrell, A., 2009. What matters in corporate governance?. Review of Financial Studies 22(2), 783-827.
- Benartzi, S., Thaler, R.H., 1995. Myopic loss aversion and the equity premium puzzle. Quarterly Journal of Economics 110(1), 73–92.
- Bertrand, M., Mullainathan, S., 2003. Enjoying the quiet life? Corporate governance and managerial preferences. Journal of Political Economy 111(5), 1043–75.
- Bethel, J., Gillan, S., 2002. The impact of the institutional and regulatory environment on shareholder voting. Financial Management 31(4), 29-54.
- Blanding, M., 2017. Vanguard, Trian and the problem with “passive” index funds.
- Blundell, R., Griffith, R., Van Reenen, J., 1999. Market share, market value and innovation in a panel of British manufacturing firms. Review of Economic Studies 66(3), 529–554.
- Boone, A., White, J., 2015. The effect of institutional ownership on firm transparency and information production. Journal of Financial Economics 117, 508–533.
- Bradley, D., Kim, I., Tian, X., 2017. Do unions affect innovation? Management Science 63(7), 2049-2395.
- Bushee, B. J., 2001. Do institutional investors prefer near-term earnings over long-run value? Contemporary Accounting Research 18(2), 207-246.
- Brav, A., Jiang, W., Ma, S., and Tian, X., 2018. How does hedge fund activism reshape corporate innovation? Journal of Financial Economics, 130(2), 237-264.
- Chemmanur, T., Loutskina, E., and Tian, X., 2014. Corporate venture capital, value creation, and innovation. Review of Financial Studies 27(8), 2434-2473.
- Cornaggia, J., Mao, Y., Tian, X, and Wolfe, B., 2015. Does banking competition affect innovation? Journal of Financial Economics 115, 189-209.
- Crane, A., Michenaud, S., Weston, J., 2016. The effect of institutional ownership on payout policy: Evidence from index thresholds. Review of Financial Studies 29(6), 1377-1408.
- Del Guercio, D., 1996. The distorting effect of the prudent-man laws on institutional equity investments. Journal of Financial Economics 40(1), 31-62.
- Fang, V., Tian, X., and Tice, S., 2014. Does stock liquidity enhance or impede firm innovation? Journal of Finance 69, 2085-2125.
- Fich, E., Harford, J., Tran, A., 2015. Motivated monitors: the importance of institutional investors’ portfolio weights. Journal of Financial Economics 118, 21–48.
- Fisch, J.E., Hamdani, A., Solomon. D., S., 2019. The new titans of wall street: A theoretical framework of passive investors. University of Pennsylvania Law Review 168, 17-72.
- Gillan, S. L., Starks, L. T., 2003. Corporate governance, corporate ownership, and the role of institutional investors: A global perspective. Journal of Applied Finance 13, 4-22.
- Griliches, Z., 1990. Patent statistics as economic indicators: A survey. Journal of Economic Literature 28, 1661-1707.
- Gu, Y., Mao, C., and Tian. X., 2017. Bank interventions and firm innovation: Evidence from debt covenant violations. Journal of Law and Economics 60(4), 637-671.
- Hall, B.H., Jaffe A., Trajtenberg, M., 2001. The NBER patent citation data file: Lessons, insights and methodological tools. NBER Working Paper.
- Hall, B.H., Jaffe, A., Trajtenberg, M., 2005. Market value and patent citations. RAND Journal of Economics 36(1), 16-38.
- Harris, M., Raviv, A., 1978. Some results on incentive contracts with application to education and employment, health insurance, and law enforcement. American Economic Review 68(1), 20-30.
- He, J., Tian, X., 2013. The dark side of analyst coverage: The case of innovation. Journal of Financial Economics 109, 856-878.
- He, J., Tian, X., 2018. Finance and corporate innovation: A survey. Asia-Pacific Journal of Financial Studies 47, 165-212.
- He, J., Tian, X., 2020. Institutions and innovation: A review of recent literature, Annual Review of Financial Economics 12, 377-398.
- Hirshleifer, D., Hsu, P.H., Li, D., 2018. Innovative originality, profitability, and stock returns. Review of Financial Studies 31, 2553-2605.
- Hirshleifer, D., Low, A., Teoh, S.H., 2012. Are overconfident CEOs better innovators? Journal of Finance 72, 1457-1498.
- Holmstrom, B., 1989. Agency costs and innovation. Journal of Economic Behavior and Organization 12, 305-327.
- Holmstrom, B., 1999. Managerial incentive problems: A dynamic perspective. Review of Economic Studies 66(1), 169-182.
- Hsu, P.H., Tian, X., Xu, Y., 2014. Financial development and innovation: Cross-country evidence. Journal of Financial Economics 112(1), 116-135.
- Imbens, G., Kalyanaraman, K., 2012. Optimal bandwidth choice for the regression discontinuity estimator. Review of Economic Studies 79(3), 933-959.
- Jaffe, A., 1989. Real effects of academic research. American Economic Review 79(5), 957-970.
- Jensen, M.C., Meckling, W.H., 1976. Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics 3, 305-360.
- Jenter, D., Kanaan, F., 2015. CEO turnover and relative performance evaluation. Journal of Finance 70(5). 2155-2183.
- Kahan, M., Rock, E.B., 2007. Hedge funds in corporate governance and corporate control. University of Pennsylvania Law Review 155, 1021.
- Kahn, C., Winton A., 1998. Ownership structure, speculation, and shareholder intervention. Journal of Finance 53(1), 99-129.
- Kaplan, S.N., Zingales, L., 1997. Do investment-cash flow sensitivities provide useful measures of financing constraints?. Quarterly Journal of Economics 112(1), 169-215.
- Lanjouw, J.O., Schankerman, M., 2004. Patent quality and research productivity: Measuring innovation with multiple indicators. Economic Journal 114(495), 441-465.
- Lee, D.S., Lemieux, T., 2010. Regression discontinuity designs in economics. Journal of Economic Literature 48(2), 281-355.
- Lerner, J., Sorensen, M., Stromberg, P., 2011. Private equity and long-run investment: The case of innovation. Journal of Finance 66, 445-477.
- Levine, R., Lin, C., Wei, L., 2017. Insider trading and innovation. Journal of Law and Economics 60(4), 749-800.
- Lin, C., Liu, S., Manso, G., 2019. Shareholder litigation and corporate innovation. Working paper, Hong Kong University.
- Liu, Y., Mauer, D.C., 2011. Corporate cash holdings and CEO compensation incentives, Journal of Financial Economics 102(1), 183-198.
- Liu, Y., Shen, Y., Wang, J., Wang, Q., 2020. Passive institutional investors and corporate innovation. Working paper, Southwestern University of Finance and Economics.
- Luong, H., Moshirian, F., Nguyen, L., Tian, X., Zhang, B., 2017. How do foreign institutional investors enhance firm innovation? Journal of Financial and Quantitative Analysis 52, 1449-1490.
- Imbens, G., Kalyanaraman, K., 2012. Optimal bandwidth choice for the regression discontinuity estimator. Review of Economic Studies 79(3), 933-959.
- McCrary, J., 2008. Manipulation of the assignment variable in the regression discontinuity design: A density test. Journal of Econometrics 142, 698-714.
- Manso, G., 2011. Motivating innovation. Journal of Finance 66(5), 1823-1860.
- Matutes, C., Regibeau, P., Rockett, K., 1996. Optimal patent design and the Diffusion of Innovations. RAND Journal of Economics 27(1), 60-83.
- Moshirian, F., Tian, X., Zhang, B., Zhang, W., 2020. Financial liberalization and innovation. Journal of Financial Economics, forthcoming.
- Mullins, W., 2014. The governance impact of index funds: evidence from a regression discontinuity. Unpublished Working Paper, University of Maryland.
- Porter, M. E., 1985. The Competitive Advantage: Creating and Sustaining Superior Performance, NY: Free Press.
- Sariol, A.M., Abebe, M.A., 2017. The influence of CEO power on explorative and exploitative organizational innovation. Journal of Business Research 73, 38-45.
- Schochet, P.Z., 2008. Technical methods report: statistical power for regression discontinuity designs in education evaluations. NCEE 2008-4026.
- Seru, A., 2014. Firm boundaries matter: Evidence from conglomerates and R&D activity. Journal of Financial Economics 111, 381-405.
- Shleifer, A., Vishny, R.W., 1986. Large shareholders and corporate control. Journal of Political Economy 94(3), 461-488.
- Sias, R.W., 1996. Volatility and the institutional investor. Financial Analysts Journal 52(2), 13-20.
- Solow, R., 1957. Technical change and the aggregate production function. Review of Economics and Statistics 39, 312-320.
- Tian, X., Wang T., 2014. Tolerance for failure and corporate innovation. Review of Financial Studies 27, 211-255.
- Trajtenberg, M, Jaffe, A., Henderson, R., 1997. University versus corporate patents: A window on the basicness of inventions. Economics of Innovation and New Technology 5(1), 19-50.

### Appendix — Definition and Source of Variables
- LnPat
  - Definition: Natural logarithm of one plus the total number of patents filed (and eventually granted) by a firm in a year
  - Source: NBER Patent Citation Database
- LnCite
  - Definition: Natural logarithm of one plus the total number of citations made to a firm’s patents scaled by the total number of patents
  - Source: NBER Patent Citation Database
- ClaimNo
  - Definition: Natural logarithm of one plus the total number of claims associated with a firm’s patents scaled by the total number of patents
  - Source: USPTO Patent Claims Research Dataset
- ClaimLength
  - Definition: Natural logarithm of one plus the total word counts associated with a firm’s patent claims scaled by the total number of patents
  - Source: USPTO Patent Claims Research Dataset
- Breakthrough
  - Definition: Natural logarithm of one plus the number of patents that fall in the top 10 percent percentile of the citation distribution within technology class and application year
  - Source: NBER Patent Citation Database
- Failure
  - Definition: Natural logarithm of one plus the number of patents that receive no citations
  - Source: NBER Patent Citation Database
- Incremental
  - Definition: Natural logarithm of one plus the number of patents that are neither breakthroughs nor failures
  - Source: NBER Patent Citation Database
- NewField
  - Definition: Natural logarithm of one plus the number of patents that are filed in technology classes new to the firm
  - Source: NBER Patent Citation Database
- Originality
  - Definition: Dispersion of cited patents over technological classes
  - Source: NBER Patent Citation Database
- Generality
  - Definition: Dispersion of citing patents over technological classes
  - Source: NBER Patent Citation Database
- Product
  - Definition: Natural logarithm of one plus the total number of new products
  - Source: Capital IQ
- IO
  - Definition: Percentage of common shares held by all types of institutional investors
  - Source: Institutional 13(f) Holdings
- PIO
  - Definition: Percentage of common shares held by passive institutional investors or quasi-indexers
  - Source: Brian Bushee website
- TIO
  - Definition: Percentage of common shares held by transient institutional investors
  - Source: Brian Bushee website
- DIO
  - Definition: Percentage of common shares held by dedicated institutional investors
  - Source: Brian Bushee website
- Size
  - Definition: Natural logarithm of total assets of a firm
  - Source: Compustat
- Age
  - Definition: Number of years since the firm’s IPO
  - Source: Compustat
- R&D
  - Definition: Research and development expenses scaled by total assets
  - Source: Compustat
- Capex
  - Definition: Capital expenditure scaled by total assets
  - Source: Compustat
- PPE
  - Definition: Net property, plant, and equipment scaled by total assets
  - Source: Compustat
- Lev
  - Definition: Total debt over total assets
  - Source: Compustat
- ROA
  - Definition: Operating income before depreciation divided by total assets
  - Source: Compustat
- HHI
  - Definition: Herfindal-Hirschman index constructed based on two-digit Standard Industrial Classification (SIC) number
  - Source: Compustat
- TobinsQ
  - Definition: Market value of equity plus book value of assets minus book value of equity minus balance sheet deferred taxes scaled by total assets
  - Source: Compustat
- KZ
  - Definition: −1.002 × cash flow [(income before extraordinary items + depreciation and amortization)/lagged net property, plant, and equipment] + 0.283 × tobinsq [market value of equity + book value of total assets − book value of equity − balance sheet deferred tax] + 3.139 × leverage [total debt/total assets] − 39.368 × dividends [(preferred dividends + common dividends)/lagged net property, plant, and equipment] − 1.315 × cash holdings [(cash and short-term investment)/(lagged net property, plant, and equipment)]
  - Source: Compustat
- Vote Participation
  - Definition: Number of meetings where passive funds cast a vote over total number of meetings
  - Source: ISS
- Vote For Percentage
  - Definition: Number of meetings where passive funds cast a “For” vote over total number of meetings
  - Source: ISS
- Vote Against Percentage
  - Definition: Number of meetings where passive funds cast an “Against” vote over total number of meetings
  - Source: ISS
- Disagree with Management
  - Definition: Number of meetings in which passive funds’ vote is different from management vote recommendation over total number of meetings
  - Source: ISS
- CEO Turnover
  - Definition: A dummy variable that equals one when there is forced CEO change in a year, and zero otherwise
  - Source: Execucomp
- ∆ROA
  - Definition: A dummy variable that equals one if the firm’s return on assets in the current year is lower than that in the previous year, and zero otherwise
  - Source: Compustat
- IndROA
  - Definition: A dummy variable that equals one if the firm’s return on assets is lower than the industry median, and zero otherwise
  - Source: Compustat
- ∆Ret
  - Definition: A dummy variable that equals one if the firm’s stock return in the current year is lower than that in the previous year, and zero otherwise
  - Source: CRSP
- IndRet
  - Definition: A dummy variable that equals one if the firm’s stock return is lower than the industry median, and zero otherwise
  - Source: CRSP
- Vega
  - Definition: Natural logarithm of an option’s price sensitivity to changes in the volatility of the underlying asset
  - Source: Execucomp
- Independent Board
  - Definition: A dummy that equals one if the majority of board members are classified as independent, and zero otherwise
  - Source: ISS
- Severance Agreement
  - Definition: A dummy that equals one if a firm has a contract which ensures executives some income protection in the event of losing their positions, and zero otherwise
  - Source: ISS

*Source: wpiea2021158-print-pdf - REFERENCES*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021158-print-pdf.pdf_