## 1. Federal Contracts

## Source details

**Canonical URL:** [1. Federal Contracts](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019172-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019172-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019172-print-pdf.pdf.json)

---

### Context and research question
- Research question: Do firms lobby to obtain firm-specific returns (rent-seeking) or to convey information to policymakers?
- Setting: Sequestration of federal budget accounts that took effect on March 1, 2013.
- Rationale for using procurement contracts:
  - Contracts are measurable at the firm level, avoiding some aggregation/free-rider problems.
  - The sequester was largely unanticipated and distributed by a predetermined formula across government accounts, creating plausibly exogenous variation in firms’ expected returns to lobbying.
- Event window focus: the quarter before and the quarter after sequestration to minimize private information channels (pre-event quarter = October 1, 2012–December 31, 2012; post-event quarter = April 1, 2013–June 30, 2013).

### Identification and empirical strategy
- Treatment definitions:
  - Sequester Flag (binary): = 1 if firm obtained pre-event contracts from federal accounts subject to cuts; 0 otherwise.
  - Continuous exposure measures: Sequester Exposure (total dollar amount of a firm’s contracts in sequestered accounts scaled by total contracts), Average Sequester Ratio (%), Weighted Average Sequester Ratio (%).
- Key identifying assumption: the sequester was exogenous to firms’ prior political activities; the predetermined formula redistributed funds independently of firms’ political efforts.
- Parallel trends: confirmed in the data (Figure 2), supporting a difference-in-differences design.
- Robustness: alternative specifications, matching exercise (3-nearest neighbor propensity score matching), and placebo test.

### Event evidence and market reaction
- Market evidence that the sequester was unanticipated:
  - Cumulative abnormal returns of government-dependent firms declined by 2.3 percent during a 3-day window (from 1 day before to 1 day after) around the event.
  - Government-dependent firms defined using NIPA-based exposure measure; firms above the 75th percentile classified as government-dependent.

### Main findings on lobbying behavior (high-level)
- Divergent post-sequester lobbying responses by exposure:
  - Firms with low exposure (less than 25th percentile) cut lobbying spending by 3.1 percent.
  - Firms with high exposure (more than 75th percentile) increased lobbying spending by 3.3 percent.
- Baseline DID pattern:
  - Non-sequestered firms cut lobbying spending by 21 percent after sequestration; sequestered firms on average increased spending by 0.1 percent (example from Column (1), Panel A).
- Matched-sample DID:
  - Difference-in-difference on Lobby (log) in matched sample: 0.208* (t = 1.90) (Table 2, Panel B).
- Placebo tests:
  - Placebo event date March 1, 2014 and random assignment of sequester status produce no significant coefficients.

### Additional evidence supporting a rent-seeking interpretation
- Competition intensity:
  - Effects stronger when competition is more intense (industry-, firm-, and agency-level measures).
  - Firm-level high-competition subgroup: After sequestration = -0.509*** (t = -3.122); After sequestration*Sequester Flag = 0.470*** (t = 2.816) (Table 6, Panel A).
- Government dependence:
  - Matched sample shows stronger evidence that more government-dependent sequestered firms increased lobbying more (Table 8, Panel B).
- Targeting of lobbying:
  - Sequestered firms continued to target government agencies in their lobbying after sequestration; non-sequestered firms split lobbying efforts more evenly between government agencies and Congress.
  - Agency vs Congress ratio: After sequestration coefficients negative (e.g., -0.265* (t = -1.667)) but After sequestration*Avg. Sequester = 4.676** (t = 2.156) (Table 9).
- Firms vs associations:
  - In selected industries (air transportation, banking, chemicals, manufacturing, oil and gas, telecommunications) firm lobbying relative to industry association lobbying increased after sequestration (Figure 3; top 50 industry lobbyists data).
- Post-sequestration contracts and performance:
  - Sequestered firms that lobbied intensively obtained more contracts afterwards:
    - Contracts (t,t+2) Full Sample: Q1 9.8, Q2 10.9, Q3 12.3, Q4 12.9; Q4-Q1 = 3.1** (t = 2.14) (Table 10).
  - Intensively lobbying sequestered firms did not exhibit better operating performance after sequestration:
    - ROA and ROE (industry-adjusted) show no improvement for lobbying-intensive firms (Table 11; selected coefficients not significant).

### Data, measurement, and sample construction
- Main data sources: Federal contracts (usaspending.gov), Lobbying Disclosure Act reports (Senate Office of Public Records), OMB account-level sequester report (published March 1, 2013).
- Contract data processing:
  - Keep only new contracts; exclude modified contracts.
  - Exclude contracts below $150,000.
  - Use federal contracts in fiscal years 2011 and 2012 (October 1, 2010 to September 30, 2012) to determine exposure.
  - Merge contract account identifiers with OMB sequester ratios at agency-account level.
- Lobbying data processing:
  - Extract quarterly lobbying reports; compute total lobbying spending per firm for pre- and post-event quarters; use natural logarithm of lobbying spending (Lobby (log)).
  - Classify lobbying targets as Senate/House or Federal Agencies; split lobbying equally across listed entities when computing target shares.
  - Industry association lobbying compiled for suggestive evidence (ignored in main DID due to free-rider contamination).
- Sequester ratio construction (examples preserved):
  - Federal Highway Administration accounts: Emergency relief program (5 percent; $2.02 billion sequester amount $101 million), payment to the Transportation Trust Fund (5.1 percent; $6.2 billion sequester amount $316 million), Federal-aid Highways (5.1 percent; $739 million sequester amount $38 million).
  - Simple average sequester ratio example = 5.07 percent; weighted average example = 5.08 percent.

### Financial variables and controls
- Variables from COMPUSTAT and other sources: Size (log total assets), R&D (XRDQ scaled by beginning assets), Tobin’s Q, cash flow, industry concentration (Herfindahl-Hirschman Index from Hoberg-Phillips TNIC), government dependence (Belo, Gala, and Li, 2013).
- Sequestration exposure measures computed from 2011–2012 federal contracts:
  - Sequester Flag, Sequester Exposure, Average Sequester Ratio (%), Weighted Average Sequester Ratio (%).

### Descriptive statistics (selected exact figures from Table 1, Panel A, full sample of 221 firms)
- Lobby ($millions): Mean 0.818; Median 0.300; Std. Dev. 1.323; p25 0.083; p75 1.049; Before 0.842; After 0.794; Sequestered 0.838; Non-sequestered 0.596.
- Lobby (log): Mean 12.543; Median 12.612; Std. Dev. 1.589; p25 11.326; p75 13.863; Before 12.549; After 12.537; Sequestered 12.524; Non-sequestered 12.754.
- Sequester Flag: Mean 0.919; Median 1.000; Std. Dev. 0.274; p25 1.000; p75 1.000.
- Sequester Exposure: Mean 0.754; Median 0.925; Std. Dev. 0.337; p25 0.623; p75 1.000; Sequestered 0.821; Non-sequestered 0.000.
- Average Sequester Ratio (%): Mean 5.087; Median 5.264; Std. Dev. 2.491; p25 3.760; p75 7.422; Sequestered 5.538; Non-sequestered 0.000.
- Weighted Average Sequester Ratio (%): Mean 5.064; Median 5.119; Std. Dev. 2.487; p25 3.753; p75 7.422; Sequestered 5.513; Non-sequestered 0.000.
- Size (log assets): Mean 9.614; Median 9.661; Std. Dev. 1.751.
- Federal Contracts ($millions) - Beginning of Period: Mean 320.706; Median 68.960; Std. Dev. 1195.820; p25 1.309; p75 105.435.
- Government Dependence: Mean 0.131; Median 0.112; Std. Dev. 0.085; p25 0.074; p75 0.173.

### Panel estimations and key regression coefficients (selected exact estimates)
- Table 3, Panel A (Full Sample, Observations 442):
  - After sequestration coefficients: -0.207** (t = -2.032); -0.137* (t = -1.760); -0.101 (t = -1.322); -0.103 (t = -1.342).
  - After sequestration*Sequester Flag: 0.214** (t = 1.993).
  - After sequestration*Sequester Exposure: 0.170* (t = 1.716).
  - After sequestration*Avg. Sequester: 1.836 (t = 1.283).
  - After sequestration*Wgt. Avg. Sequester: 1.879 (t = 1.310).
- Table 3, Panel B (Matched Sample, Observations 276):
  - After sequestration coefficients: -0.199* (t = -1.813); -0.232** (t = -2.418); -0.204** (t = -2.187); -0.207** (t = -2.211).
  - After sequestration*Sequester Exposure: 0.262** (t = 2.188).
  - After sequestration*Avg. Sequester: 3.487** (t = 2.022).
  - After sequestration*Wgt. Avg. Sequester: 3.565** (t = 2.066).

### Heterogeneity highlights (selected exact estimates)
- Industry concentration (Table 5, Low concentration subgroup, Full Sample):
  - After sequestration: -0.431*** (t = -2.732) to -0.289*** (t = -2.688).
  - Positive interactions between After sequestration and sequester measures significant.
- Firm-level competition (Table 6, Full Sample, High competition):
  - After sequestration: -0.509*** (t = -3.122).
  - After sequestration*Sequester Flag: 0.470*** (t = 2.816).
  - After sequestration*Sequester Exposure: 0.291* (t = 1.958).
- Agency-level competition (Table 7, Matched Sample, High agency competition):
  - After sequestration: -0.181* (t = -1.883) to -0.215** (t = -2.120).
  - After sequestration*Avg. Sequester: 4.642** (t = 2.035).

### Contracts and operating performance outcomes (selected exact figures)
- Contracts outcomes (Table 10, Full Sample):
  - Contracts (t,t+1) by lobbying-change quantiles: Q1 8.1; Q2 8.0; Q3 9.7; Q4 10.2; Q4-Q1 2.1 (t = 1.37).
  - Contracts (t,t+2) by quantiles: Q1 9.8; Q2 10.9; Q3 12.3; Q4 12.9; Q4-Q1 3.1** (t = 2.14).
- Operating performance (Table 11):
  - Lobbying main effect on ROA: 0.021 (t = 0.390) Full Sample; 0.037 (t = 0.452) Matched Sample.
  - Lobbying*After Sequestration on ROE: -0.056 (t = -0.571) Full Sample; 0.012 (t = 0.0828) Matched Sample.
  - Conclusion: contract gains for intense lobbyers without concomitant improvements in industry-adjusted ROA/ROE.

### Interpretation, contribution, and conclusion
- Converging evidence indicates a rent-seeking motive:
  - Predetermined, public sequester exposure reduces scope for private-information channels.
  - Short event window (quarter before vs quarter after) limits endogenous lobbying timing.
  - Highly exposed firms maintained or increased lobbying while low-exposure or non-sequestered firms cut lobbying.
  - Effects stronger under higher competition and for government-dependent firms.
  - Sequestered firms kept agency-directed lobbying and shifted toward firm-specific lobbying relative to associations.
  - Lobbying-intensive sequestered firms secured more contracts but did not show improved operating performance.
- Contribution:
  - Using an exogenous budget shock (sequestration) helps identify causal responses of firm lobbying to changes in prospective rents, distinguishing rent-seeking from pure information provision motives.
- Caveat:
  - Authors note it is not possible to fully rule out information motives, but the weight of evidence favors rent-seeking as a component of post-sequestration lobbying behavior.

*Source: wpiea2019172-print-pdf (section: 1. Federal Contracts).*

### 1. Federal Contracts .........................................................................................11

### 1. Federal Contracts

### Context and research question
- Research question: Do firms lobby to obtain firm-specific returns (rent-seeking) or to convey information to policymakers?
- Setting: Sequestration of federal budget accounts that took effect on March 1, 2013.
- Rationale for using procurement contracts:
  - Contracts are measurable at the firm level, avoiding some aggregation/free-rider problems.
  - The sequester was largely unanticipated and distributed by a predetermined formula across government accounts, creating plausibly exogenous variation in firms’ expected returns to lobbying.
- Event window focus: the quarter before and the quarter after sequestration to minimize private information channels.

### Identification and empirical strategy
- Treatment definitions:
  - Binary indicator: firm is “sequestered” = 1 if the firm obtained contracts prior to sequestration from federal accounts subject to cuts; 0 otherwise.
  - Continuous exposure measures: e.g., total dollar amount of a firm’s contracts in sequestered accounts scaled by the firm’s total contracts.
- Key identifying assumption: the sequester was exogenous to firms’ prior political activities; the predetermined formula redistributed funds independently of firms’ political efforts.
- Parallel trends assumption: confirmed in the data (referenced Figure 2), supporting a difference-in-differences design.
- Robustness strategies: alternative specifications, matching exercise, and a placebo test.

### Event evidence and market reaction
- Market evidence that the sequester was unanticipated:
  - Cumulative abnormal returns of government-dependent firms declined by 2.3 percent during a 3-day window (from 1 day before to 1 day after) around the event.
  - Government-dependent firms defined using NIPA-based exposure measure; figure plots firms above the 75th percentile as government-dependent.

### Main findings on lobbying behavior
- Divergent post-sequester lobbying responses by exposure:
  - Firms with low exposure to the sequester (less than 25th percentile) cut lobbying spending by 3.1 percent.
  - Firms with high exposure to the sequester (more than 75th percentile) increased lobbying spending by 3.3 percent.
- Interpretation: findings are consistent with rent-seeking—firms more affected by cuts intensified lobbying to compete for a larger share of a smaller pie.
- Robustness: results hold across alternative specifications, matching, and placebo tests.

### Additional evidence supporting rent-seeking interpretation
- Competition intensity:
  - The effect of spending cuts on lobbying spending is stronger when competition is more intense (industry-, firm-, and agency-level measures).
- Government dependence:
  - Lobbying responses are more pronounced for government-dependent firms.
- Targeting of lobbying:
  - Sequestered firms continued to target government agencies in their lobbying after sequestration.
  - Non-sequestered firms split lobbying efforts more evenly between government agencies and Congress.
- Firms vs associations:
  - In most industries, individual firm lobbying relative to industry association lobbying increased in the post-sequestration period, consistent with firms pursuing firm-specific benefits.
- Post-sequestration contracts and performance:
  - Sequestered firms that lobbied intensively obtained more contracts afterwards.
  - However, these intensively lobbying sequestered firms did not exhibit better operating performance after sequestration compared to sequestered firms that did not intensify lobbying.

### Interpretation and contribution
- The combination of:
  - an unexpected, formula-driven sequester (reducing scope for private-information channels),
  - a short window around the event (quarter before and quarter after),
  - observed increases in lobbying by highly exposed firms alongside no superior operating performance,
  - stronger effects under more intense competition and among government-dependent firms,
  - and changes in lobbying targets and firm-vs-association shares,
  - collectively supports a rent-seeking motive for the observed post-sequestration lobbying increases.
- Contribution to literature:
  - The exogenous shock helps establish causal direction: firms modify lobbying in response to changes in potential returns.
  - The predetermined nature of cuts helps distinguish preferential-treatment motives from information-provision motives.

*Source: wpiea2019172-print-pdf (section: 1. Federal Contracts).*

### Section IV describes the empirical methodology and presents the results. Section V concludes.

### wpiea2019172-print-pdf - Section IV describes the empirical methodology and presents the results. Section V concludes.

### Background and literature
- Lobbying defined as a legal activity aiming at changing existing rules or policies or procuring individual benefits.
- Private benefits examples cited: preferential access to credit, bailout guarantees, privileged access to licenses, or procurement contracts.
- Two strands of lobbying research:
  - Relationship between lobbying activities and specific policies (examples cited: trade, immigration, financial services).
  - Consequences of lobbying on firm-specific economic outcomes (examples cited: effects on loans, defaults, firm value, innovation).
- This study aligns more closely with the strand focusing on granting of contracts and lobbying.

### Sequestration institutional details
- Budget Control Act (BCA) signed into law on August 2, 2011; it stipulated a joint select committee to produce legislation to decrease the federal deficit by $1.2 trillion over 10 years; failure would trigger automatic across-the-board cuts ("sequestration").
- Super committee co-chairs released a statement on November 21, 2011 that the committee would be unable to come to a bipartisan agreement.
- Effective date of sequestration was known as January 2, 2013, though legislative actions and proposals cast doubt on implementation; American Taxpayer Relief Act (ATRA) delayed sequestration for two months.
- Sequestration cuts came into effect on March 1, 2013; on March 6, 2013 Congress passed a continuing resolution with some implementation flexibility for the Pentagon and the VA.
- Sequestration involved the following reductions:
  - 10 percent reduction in the caps on new discretionary appropriations for defense programs;
  - 7.8 percent reduction in the caps on new discretionary appropriations for nondefense programs;
  - 10 percent reduction in mandatory budgetary resources for nonexempt defense programs;
  - 7.8 percent reduction in mandatory budgetary resources for nonexempt nondefense programs (except Medicare);
  - 2 percent reduction in most Medicare spending.
- The Office of Management and Budget (OMB) published the account-level reductions and sequestrable amounts on March 1, 2013.

### Politically targeted activities and disclosure framework
- Two main channels for lobbyists to influence policy:
  - Direct engagement with the executive and legislative branches.
  - Campaign finance contributions, particularly through political action committees (PACs).
- Under the Lobbying Disclosure Act of 1995 (LDA), as modified by the Honest Leadership and Open Government Act of 2007, lobbyists meeting certain criteria must file quarterly reports to the Secretary of the Senate’s Office of Public Records (SOPR).
- Definition and registration criteria for a lobbyist (Congress guide to the LDA):
  - Any individual who (1) is either employed or retained by a client for financial or other compensation; (2) whose services include more than one lobbying contact; and (3) whose lobbying activities constitute 20 percent or more of his or her services during a three-month period.
- Lobbying reports disclose dollar amounts spent, issues lobbied on (allowing association of expenditures with targeted policy areas), and names of lobbyists working on specific issues.

### Data sources and sample construction
- Main datasets merged:
  - Federal contracts (www.usaspending.gov) mandated by the Federal Funding Accountability and Transparency Act of 2006.
  - Lobbying reports from the Senate Office of Public Records.
  - OMB report on sequestration ratios and sequestrable amounts (published March 1, 2013).
- Sample focus and event windows:
  - The event is the budget sequestration on March 1, 2013.
  - Lobbying activities are reported quarterly; pre-event quarter is the 4th quarter of 2012 and post-event quarter is the 2nd quarter of 2013.
  - The 1st quarter of 2013 is excluded because it contains pre-event, event, and post-event dates and is confounded.
- Federal contracts data processing:
  - Keep only new contracts; exclude modified contracts.
  - Exclude contracts below $150,000 (considered small business set-asides and do not require a formal process).
  - Create unique numeric identifiers for parent companies in the federal contract database for matching with lobbying and company financials.
  - Use federal contracts in fiscal years 2011 and 2012 (October 1, 2010 to September 30, 2012) to determine firms’ exposure to sequestration prior to the event.
  - For each contract, use the funding federal agency and program source account (assigned by the U.S. Department of Treasury) to merge at the account level with the OMB report to obtain sequester ratios applied to each federal agency account.
- Lobbying data processing:
  - Extract lobbying reports filed by federal contractors for the sample period around March 1, 2013.
  - Total lobbying spending by federal contractors is calculated for the quarters before and after the event.
  - The natural logarithm of lobbying spending is used in the analysis.
  - Classify targeted offices under two categories: the Senate and the House of Representatives, or Federal Agencies (Department of Defense, Department of Transportation, Department of Agriculture, etc.) to detect potential shifts in lobbying focus after sequestration.
  - Compile information on lobbying expenditures of industry associations; main analysis focuses on firms’ own lobbying (association spending is ignored in main analysis because of free-rider contamination and the difference-in-difference design). The text notes that association spending may provide suggestive evidence if sequestered firms switch more intensely from association to own spending than non-sequestered firms.
- OMB report and sequester ratio construction:
  - OMB report provides sequester ratios at the agency account level; ratios may differ across accounts of the same federal agency.
  - Example given (Federal Highway Administration accounts):
    - Accounts and sequester ratios: Emergency relief program (5 percent), payment to the Transportation Trust Fund (5.1 percent), Federal-aid Highways (5.1 percent).
    - Sequestrable amounts: $2.02 billion, $6.2 billion, and $739 million, respectively.
    - Sequestered amounts for these accounts: $101 million, $316 million, and $38 million.
  - Sequester ratio measures for agencies:
    - Simple average sequester ratio across accounts (example: average of 5, 5.1, and 5.1 percent = 5.07 percent for Federal Aviation Agency).
    - Weighted average sequester ratio based on sequestrable amount and sequester ratio (example calculation yields 5.08 percent for the described set of accounts using the formula ((5percent * $2.02 billion + 5.1percent * $6.2 billion + 5.1percent * $739 million) / ($2.02 billion + $6.2 billion + $739 million))).
  - Manual matching process to merge OMB agency-account identifiers to federal contract data:
    - Create unique identifiers for each agency in the OMB report.
    - Keep unique agencies in the federal contract data based on major funding agency category.
    - Manually screen and link agency names between the two datasets.
    - Add unique OMB agency identifiers into the federal contract data.

*Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019172-print-pdf.pdf*

### 4. Financial Variables

### 4. Financial Variables

### Financial variables and measurement
- Variables constructed from COMPUSTAT and other sources: firm size, R&D spending, industry concentration, government dependence, Tobin’s Q, cash flow.
- Frequency and aggregation:
  - Size, Tobin’s Q, cash flow, and R&D: computed as quarterly series (same frequency as lobbying data).
  - Industry concentration: Herfindahl-Hirschman index based on Text-based Network Industry Classifications from the Hoberg-Phillips Data Library, measured at the beginning of period.
  - Government dependence: calculated following Belo, Gala, and Li (2013) using NIPA tables at the 2-digit SIC level.
- Sequestration exposure measures (based on federal contracts in fiscal period 2011–2012, September 2010–October 2012):
  - Sequester Flag: = 1 if firm obtained pre-event contracts from agencies that were sequestered on March 1, 2013; = 0 otherwise.
  - Sequester exposure: total dollar amount of a firm’s contracts in sequestered accounts scaled by total dollar amount of all contracts.
  - Average sequester ratio and weighted average sequester ratio: firm-level ratios derived by applying agency account-level sequester ratios from the OMB report to each firm’s contracts and scaling by total federal contracts for that firm.

### Empirical methodology
- Identification strategy: difference-in-differences around the sequestration shock on March 1, 2013.
- Outcome variable: Lobbyity_it — natural logarithm of total lobbying amount of firm i at quarter t (pre-event quarter = October 1, 2012–December 31, 2012; post-event quarter = April 1, 2013–June 30, 2013).
- Baseline regression (controls and fixed effects):
  - Interaction of After_sequestration_t (indicator for post-event quarter) with Sequester variable_i (alternative definitions).
  - Control vector X_it: firm and industry level variables (firm size, R&D spending, industry concentration, growth opportunities, cash flow).
  - Firm fixed effects n_i; robust standard errors clustered at the firm level.
- Samples:
  - Full balanced panel: 221 firms (203 sequestered, 18 not sequestered), total observations 442.
  - Matched sample via propensity score matching (3-nearest neighbor, each matched firm included once): 122 sequestered firms, 16 non-sequestered firms, overall sample size 276.
  - Propensity score predictors: natural logarithm of contracts obtained over previous two fiscal years (September 2010–October 2012), natural logarithm of total lobbying expense over the last two quarters before pre-event (2nd and 3rd quarters of 2012), firm size at beginning of sample (end of 3rd quarter of 2012), industry (12 Fama-French sectors).

### Descriptive statistics (key figures)
- Average lobbying (full sample):
  - Overall sample average: $800,000 (text summary) and $818,000 (Table 1, Panel A summary).
  - Before sequestration average: $842,000.
  - After sequestration average: $794,000.
  - Sequestered firms average lobbying: $838,000.
  - Non-sequestered firms average lobbying: $596,000.
- Exposure prevalence and dispersion:
  - 92 percent of firms (203 firms) exposed to sequestration; 18 firms had zero exposure.
  - Coefficient of variation across continuous exposure measures:
    - Sequester exposure: 45 percent.
    - Average sequester ratio and weighted average sequester ratio: 49 percent.
- Matched sample: federal contracts and lobbying amounts between sequestered and non-sequestered firms are closer; decrease in lobbying after event is less pronounced.

### Analyses and findings

#### A. Univariate results
- Pre/post logging (means in logs, Table 1, Panel C):
  - Sequestered firms: 12.52 (before) and 12.53 (after).
  - Non-sequestered firms: 12.86 (before) and 12.65 (after).
  - Interpretation: sequestered firms kept lobbying expenses roughly constant, non-sequestered firms reduced lobbying after sequestration.
- Event-time pattern (Figure 2): parallel trends pre-event; clear break at event (2013Q1 as event period (0)): sequestered firms maintained lobbying; non-sequestered firms cut lobbying dramatically.

#### B. Panel estimations (baseline and matched samples)
- Baseline result pattern:
  - Baseline ("after sequestration" without interactions) is negative: firms with zero or low sequester exposure cut lobbying spending post event.
  - Significant, positive interaction of sequester measures with After_sequestration: sequestered firms cut lobbying less or increased lobbying relative to low-exposure firms.
- Magnitudes (examples reported in text):
  - Column (1), Panel A: non-sequestered firms cut lobbying spending by 21 percent after sequestration; sequestered firms on average increased spending by 0.1 percent.
  - Column (2): firms with low exposure (less than 25th percentile) cut lobbying spending by 3.1 percent; firms with high exposure (more than 75th percentile) increased lobbying spending by 3.3 percent.
  - Columns (3) and (4): larger magnitudes for high-exposure firms, though statistically weaker.
- Matched sample (Panel B, Table 2):
  - Results align with full sample; statistically stronger for continuous measures of sequester exposure despite smaller sample size.

#### Placebo tests
- Two placebo exercises (Table 3):
  - Placebo event date: March 1, 2014 (post-event quarter April 1–June 30, 2014; pre-event quarter September 1–December 31, 2013).
  - Random assignment of sequester status in the baseline period (maintaining counts of 18 not sequestered and 203 sequestered).
- Findings: no significant coefficients in placebo exercises (Panels A and B); placebo tests support that results are driven by the actual sequestration shock.

#### C. Interpretation and robustness checks supporting rent-seeking motive
- Information vs. rent-seeking:
  - Sequestration exposure is predetermined and publicly available; little private information to convey—findings are therefore less consistent with an information-provision motive.
  - Evidence consistent with rent-seeking: sequestered firms maintain/increase lobbying to compete for larger share of reduced pie.
- Competition dimension:
  - Results driven by firms in industries with low concentration (high competition) (Table 5).
  - Firm-level competition measure (share of competitive contracts): effects stronger for firms facing high competition (Table 6).
  - Agency-level competition: matched sample supports stronger effects for firms obtaining contracts from more competitive agencies (Table 7).
  - Overall: lobbying spending increases more post-event for firms facing higher competition.
- Government dependence:
  - Table 8: full sample weak; matched sample (Panel B) shows stronger evidence that more government-dependent sequestered firms increased lobbying more; weaker or no effect for low-dependence firms.
- Target of lobbying — agencies vs. Congress:
  - Method: split lobbying spending equally across listed entities each firm lobbies; compute relative lobbying towards agency vs. Congress.
  - Evidence (Table 9): relative lobbying directed towards agencies declined for both groups after event but declined less for sequestered firms (significant at the 11 percent level). Suggestive evidence that sequestered firms kept up agency-directed lobbying.
- Persistence of agency relationships:
  - Typical contractor works with the same two agencies on average (sample 2008–2012).
  - Sequestered firms maintain lobbying with the same agency after sequestration.
- Association vs. firm-level lobbying:
  - Aggregate evidence from top 50 industry lobbyists (Center for Responsive Politics) for selected industries (air transportation, banking, chemicals, manufacturing, oil and gas, telecommunications): firm lobbying relative to association lobbying picks up after sequestration (Figure 3). Suggestive of firms shifting toward firm-specific lobbying.
- Outcomes: contracts and operating performance
  - Contracts (Table 10): sample divided into four quantiles based on difference in lobbying (post - pre). Firms with more intense lobbying obtain more contracts over one quarter and two quarters after sequestration.
  - Quantile cutoffs:
    - Full sample: -0.2, 0, and 0.18.
    - Matched sample: -0.19, 0, and 0.16.
  - Operating performance (Table 11): industry-adjusted ROA and ROE (industry mean at 2-digit SIC subtracted from firm ROA and ROE) over one and two quarters after sequestration show no improvement for lobbying firms in either full or matched samples.
  - Interpretation: lobbying-intensive firms do not exhibit better operating performance after sequestration, consistent with rent-seeking rather than information-based efficiency gains.

### Synthesis of evidence and main conclusion
- Empirical pattern:
  - Sequestered firms lobbied more (or cut less) after the March 1, 2013 sequestration shock compared to non-sequestered or low-exposure firms.
  - Effects are more pronounced when competition is intense (industry, firm, or agency level) and in government-dependent sectors.
  - Sequestered firms that increased lobbying obtained more contracts after sequestration but did not exhibit superior operating performance relative to peers.
- Inference:
  - Multiple strands of evidence (predetermined exposure, competition sensitivity, agency-directed lobbying, association vs. firm lobbying, contract gains without operating performance gains) point toward a rent-seeking motive for increased lobbying after sequestration.
  - The authors acknowledge it is not possible to fully rule out information motives, but the evidence more strongly supports rent-seeking as a component of lobbying behavior in this setting.

*Source: IMF Working Paper — Chapter 4 (Empirical analysis of lobbying and sequestration).*

### REFERENCES

### wpiea2019172-print-pdf - REFERENCES

### Literature and theoretical background
- Citations focus on political connections, lobbying, and regulation including:
  - Acemoglu et al., 2016; Adelino and Dinc, 2014; Akcigit, Baslandze, and Lotti, 2018; Belo, Gala, and Li, 2013; Bernheim and Whinston, 1986; Bertok, 2008; Bertrand, Duflo, and Mullainathan, 2004; Bertrand, Bombardini, and Trebbi, 2014; Blanes-Vidal, Draca, and Fons-Rosen, 2012; Claessens, Feijen, and Laeven, 2008; Facchini, Mayda, and Mishra, 2008; Faccio, 2006; Faccio and Parsley, 2009; Fisman, 2001; Goldberg and Maggi, 1999; Goldman, Rocholl, and So, 2013; Grossman and Helpman, 1994, 2001; Hill et al., 2013; Hoberg and Phillips, 2016; Johnson and Mitton, 2003; Kerr, Lincoln, and Mishra, 2014; Khwaja and Mian, 2005; Kroszner and Stratmann, 1998; Kroszner and Strahan, 1999; Lohmann, 1995; Ludema, Mayda, and Mishra, 2010; Potters and van Winden, 1992; Raddatz and Braun, 2009; Stigler, 1971.

### Event definition and measurement
- Event: announcement of budget sequester on March 1, 2013 (event day 0).
- Government-dependent firms: those above 75th percentile of government spending exposure based on 2digit SIC codes (Belo, Gala, and Li, 2013).
- Sequestration event quarter: 1st quarter of 2013 (January 1–March 30, 2013).
- Pre-event quarters: -4 to -1 relative to event; post-event quarters: 1 to 2 relative to event.
- Sequester Flag: 1 for firms exposed to sequestration (any contracts in 2011–2012 fiscal period exposed).
- Sequester Exposure: total amount of contracts exposed to sequester as a ratio of total contracts.
- Average Sequester Ratio (%) and Weighted Average Sequester Ratio (%) defined as in the text using 2011 and 2012 federal contracts.

### Figures — event responses (preserved descriptions)
- Figure 1: Market reaction to the announcement of budget sequester on March 1, 2013; abnormal returns based on market model; government-dependent firms per Belo, Gala, and Li (2013).
- Figure 2: Mean of lobbying amount for quarters before and after sequestration; event quarter is 1st quarter of 2013; pre-event (-4 to -1) and post-event (1 to 2).
- Figure 3: Ratio of firm lobbying amount to total lobbying (firm + association) around sequestration; pre-event = 4th quarter of 2012, post-event = 2nd quarter of 2013; association lobbying determined by top 50 lobbyists in Center for Responsive Politics.

### Descriptive statistics (Table 1; Panel A full sample; preserve exact values)
- Sample: balanced sample of 221 firms over two time periods (before and after event).
- Lobby ($millions): Mean 0.818; Median 0.300; Std. Dev. 1.323; p25 0.083; p75 1.049; Before 0.842; After 0.794; Sequestered 0.838; Non-sequestered 0.596.
- Lobby (log): Mean 12.543; Median 12.612; Std. Dev. 1.589; p25 11.326; p75 13.863; Before 12.549; After 12.537; Sequestered 12.524; Non-sequestered 12.754.
- Sequester Flag: Mean 0.919; Median 1.000; Std. Dev. 0.274; p25 1.000; p75 1.000; Before 0.919; After 0.919; Sequestered 1.000; Non-sequestered 0.000.
- Sequester Exposure: Mean 0.754; Median 0.925; Std. Dev. 0.337; p25 0.623; p75 1.000; Before 0.754; After 0.754; Sequestered 0.821; Non-sequestered 0.000.
- Average Sequester Ratio (%): Mean 5.087; Median 5.264; Std. Dev. 2.491; p25 3.760; p75 7.422; Before 5.087; After 5.087; Sequestered 5.538; Non-sequestered 0.000.
- Weighted Average Sequester Ratio (%): Mean 5.064; Median 5.119; Std. Dev. 2.487; p25 3.753; p75 7.422; Before 5.064; After 5.064; Sequestered 5.513; Non-sequestered 0.000.
- Size: Mean 9.614; Median 9.661; Std. Dev. 1.751; p25 8.484; p75 10.684; Before 9.614; After 9.614; Sequestered 9.618; Non-sequestered 9.572.
- R&D: Mean 0.006; Median 0.000; Std. Dev. 0.010; p25 0.000; p75 0.008; Before 0.007; After 0.004; Sequestered 0.006; Non-sequestered 0.010.
- Tobin's Q: Mean 1.609; Median 1.391; Std. Dev. 0.741; p25 1.137; p75 1.840; Before 1.570; After 1.648; Sequestered 1.605; Non-sequestered 1.660.
- ROA (industry adjusted): Mean 0.230; Median 0.099; Std. Dev. 0.396; p25 0.025; p75 3.169; Before 0.237; After 0.223; Sequestered 0.219; Non-sequestered 0.355.
- ROE (industry adjusted): Mean 0.016; Median 0.009; Std. Dev. 0.494; p25 -0.027; p75 0.043; Before -0.001; After 0.032; Sequestered 0.015; Non-sequestered 0.026.
- Industry Concentration: Mean 0.227; Median 0.156; Std. Dev. 0.204; p25 0.090; p75 0.043; Before 0.219; After 0.234; Sequestered 0.231; Non-sequestered 0.178.
- Firm Level Competition: Mean 0.539; Median 0.611; Std. Dev. 0.401; p25 0.006; p75 0.999; Before 0.539; After 0.539; Sequestered 0.533; Non-sequestered 0.613.
- Government Agency Level Competition: Mean 0.526; Median 0.516; Std. Dev. 0.058; p25 0.479; p75 0.546; Before 0.526; After 0.526; Sequestered 0.525; Non-sequestered 0.536.
- Government Dependence: Mean 0.131; Median 0.112; Std. Dev. 0.085; p25 0.074; p75 0.173; Before 0.131; After 0.131; Sequestered 0.133; Non-sequestered 0.108.
- Federal Contracts ($millions) - Beginning of Period: Mean 320.706; Median 68.960; Std. Dev. 1195.820; p25 1.309; p75 105.435; Before 320.706; After 320.706; Sequestered 348.866; Non-sequestered 3.129.
- Federal Contracts (log) - Beginning of Period: Mean 16.322; Median 16.008; Std. Dev. 2.738; p25 14.085; p75 18.474; Before 16.322; After 16.322; Sequestered 16.546; Non-sequestered 13.797.

### Difference-in-differences and main regression findings (Table 2 and Table 3)
- Table 2 (Panel B matched sample): Difference-in-difference on Lobby (log): 0.208* (t-statistic 1.90).
- Table 3 — Lobbying after Sequestration:
  - Panel A: Full Sample (Observations 442; R-squared values 0.015 to 0.011)
    - After sequestration coefficients (columns (1)–(4)):
      - After sequestration: -0.207** (t = -2.032); -0.137* (t = -1.760); -0.101 (t = -1.322); -0.103 (t = -1.342).
      - After sequestration*Sequester Flag: 0.214** (t = 1.993) in column with interaction.
      - After sequestration*Sequester Exposure: 0.170* (t = 1.716).
      - After sequestration*Avg. Sequester: 1.836 (t = 1.283).
      - After sequestration*Wgt. Avg. Sequester: 1.879 (t = 1.310).
  - Panel B: Matched Sample (Observations 276; R-squared 0.022 to 0.039)
    - After sequestration coefficients (columns (1)–(4)):
      - After sequestration: -0.199* (t = -1.813); -0.232** (t = -2.418); -0.204** (t = -2.187); -0.207** (t = -2.211).
      - After sequestration*Sequester Exposure: 0.262** (t = 2.188).
      - After sequestration*Avg. Sequester: 3.487** (t = 2.022).
      - After sequestration*Wgt. Avg. Sequester: 3.565** (t = 2.066).
    - R&D coefficients notable: -12.33 (t = -1.561) to -16.60** (t = -1.979) in some specifications.

### Placebo tests (Table 4)
- Panel A: Full Sample placebo with alternative sequester date March 1, 2014 and random sequester assignment (Observations 410 and 442; R-squared up to 0.007):
  - After sequestration coefficients in placebo tests: -0.125 (t = -0.759); -0.0284 (t = -0.265); -0.0223 (t = -0.221); -0.0246 (t = -0.243); random sequester assignment column: -0.0989 (t = -0.928).
  - Interaction terms largely insignificant (e.g., After sequestration*Sequester Flag 0.136 (t = 0.795)).
- Panel A matched sample and Panel B matched sample reported similar null placebo patterns (e.g., After sequestration -0.162 (t = -0.963) to -0.0904 (t = -0.670)).

### Heterogeneity by industry concentration (Table 5)
- Panel A: Full Sample — High concentration subgroup:
  - After sequestration coefficients: 0.0843 (t = 0.753) to 0.146* (t = 1.753); interactions: After sequestration*Wgt. Avg. Sequester 4.886*** (t = 2.902).
- Panel A — Low concentration subgroup:
  - After sequestration coefficients: -0.431*** (t = -2.732) to -0.289*** (t = -2.688); interactions significant and positive for sequester measures.
- Panel B: Matched Sample shows similar patterns with High and Low splits; for Low concentration matched sample, After sequestration coefficients -0.440*** (t = -2.651) to -0.405*** (t = -3.476); After sequestration*Avg. Sequester 5.811*** (t = 2.731).

### Heterogeneity by firm-level competition (Table 6)
- Panel A: Full Sample — High competition:
  - After sequestration: -0.509*** (t = -3.122); After sequestration*Sequester Flag 0.470*** (t = 2.816); After sequestration*Sequester Exposure 0.291* (t = 1.958).
- Panel A — Low competition:
  - After sequestration: 0.00420 (t = 0.0451) to -0.0765 (t = -0.686); interactions generally not significant.
- Panel B: Matched Sample shows similar contrasts; High competition matched sample After sequestration -0.491*** (t = -2.825); interaction After sequestration*Sequester Flag 0.468** (t = 2.619).

### Heterogeneity by agency-level competition (Table 7)
- Panel A: Full Sample — High agency competition:
  - After sequestration coefficients: 0.0711 (t = 1.434) to -0.236 (t = -1.120); R&D strongly negative and significant: -13.43*** to -19.35*** across specifications.
- Panel A — Low agency competition:
  - After sequestration coefficients: -0.211* (t = -1.912) to -0.126 (t = -1.551); interactions show After sequestration*Sequester Flag 0.209* (t = 1.799) in some specs.
- Panel B: Matched Sample — High agency competition:
  - After sequestration: -0.181* (t = -1.883) to -0.215** (t = -2.120); After sequestration*Avg. Sequester 4.642** (t = 2.035).
- Panel B — Low agency competition: After sequestration coefficients not significant.

### Heterogeneity by government dependence (Table 8)
- Panel A: Full Sample — High dependence:
  - After sequestration coefficients: -0.223 (t = -1.498) to -0.0468 (t = -0.523); After sequestration*Sequester Flag 0.260* (t = 1.688) in some specs.
- Panel A — Low dependence:
  - After sequestration coefficients: -0.0916 (t = -0.706) to -0.136 (t = -0.992); interaction terms not strongly significant.
- Panel B: Matched Sample results reported with Observations 160 for High and 160 for Low; R-squared up to 0.099.

### Agency versus Congress lobbying intensity (Table 9)
- Agency Lobbying to Congress Lobbying Ratio — Full Sample:
  - After sequestration: -0.265* (t = -1.667); -0.206* (t = -1.793); -0.228** (t = -2.028); -0.229** (t = -2.036).
  - After sequestration*Avg. Sequester: 4.676** (t = 2.156); After sequestration*Wgt. Avg. Sequester: 4.691** (t = 2.163).
- Panel B matched sample (High/Low groups) shows mixed patterns; matched sample difference-in-differences reported:
  - Sequestered vs Non-sequestered Difference in Difference: 0.04 (t = 0.21) for Agency Lobbying to Congress Lobbying Ratio.

### Contracts outcomes after sequestration — sequestered firms (Table 10)
- Contracts (t,t+1) (Full Sample) by lobbying-change quantiles:
  - Q1 (Most Negative): 8.1
  - Q2: 8.0
  - Q3: 9.7
  - Q4 (Most Positive): 10.2
  - Q4-Q1: 2.1 (t = 1.37)
- Contracts (t,t+2) (Full Sample):
  - Q1: 9.8
  - Q2: 10.9
  - Q3: 12.3
  - Q4: 12.9
  - Q4-Q1: 3.1** (t = 2.14)
- Matched Sample Contracts (t,t+1):
  - Q1: 5.7; Q2: 5.5; Q3: 5.2; Q4: 7.3; Q4-Q1 1.6 (t = 0.92).
- Matched Sample Contracts (t,t+2):
  - Q1: 7.9; Q2: 9.0; Q3: 9.8; Q4: 10.2; Q4-Q1 2.3 (t = 1.3).

### Operating performance for sequestered firms (Table 11)
- Dependent variables: ROA, ROA(6mo), ROE, ROE(6mo); industry-adjusted measures.
- Selected coefficients and means (Full Sample / Matched Sample):
  - Lobbying main effect on ROA: 0.021 (t = 0.390) Full Sample; 0.037 (t = 0.452) Matched Sample.
  - After Sequestration main effect on ROA(6mo): 0.267 (t = 0.633) Full Sample; 0.716 (t = 1.322) Matched Sample.
  - Lobbying*After Sequestration on ROE: -0.056 (t = -0.571) Full Sample; 0.012 (t = 0.0828) Matched Sample.
  - R&D coefficients in operating performance regressions vary widely (e.g., -9.498 to -31.81) with t-statistics indicating limited significance in many specs.
- Sample sizes and fit:
  - Observations: Full Sample ROA regressions 400; ROE regressions 384; Matched Sample smaller (e.g., 240, 236); Number of num_parent_id 200, 192, 120, 118 as provided.
  - R-squared values range from 0.002 to 0.082 across specifications.

*Source: wpiea2019172-print-pdf - REFERENCES (IMF working paper PDF content).*

### Appendix 1. Variable Descriptions and Data Sources

### Appendix 1. Variable Descriptions and Data Sources

### Lobbying and Sequestration Indicators
- Lobby (log)
  - Natural logarithm of the total lobby amount for each quarter
  - Source(s): Lobbying Disclosure Act database provided by the United States Senate Office of Public Records (www.senate.gov/legislative/opr.htm).
- After Sequestration
  - Indicator variable that takes a value of 1 for the quarter after sequestration (post-event period: April 1–June 30, 2013) and is zero for the quarter before sequestration (pre-event period: October 1–December 31, 2012).
  - Sequestration happened on March 1, 2013.
  - Source(s): By authors
- Sequester Flag
  - Indicator variable that takes a value of 1 in the post-event period, April 1–June 30, 2013 for a sequestered firm. If any contracts obtained in the 2011–2012 fiscal period is exposed to sequestration, the firm is considered to be exposed to sequestered.
  - Source(s): usaspending.gov; Office of the Management and Budget (OMB) (https://www.whitehouse.gov/omb/budget/Historicals)

### Measures of Sequester Exposure
- Sequester Exposure
  - Firm’s exposure to sequestration is defined as the total dollar amount of a firm's sequestered contracts, scaled by total dollar amount of a firm's all contracts. This ratio is calculated based on the federal contracts obtained in the 2011–2012 fiscal period (October 1, 2010–September 30, 2012).
  - Source(s): usaspending.gov; OMB (https://www.whitehouse.gov/omb/budget/Historicals)
- Average Sequester Ratio
  - Formula (descriptive): ∑∑ (contract amount_ikl × simple average seqratio_kl)_{l=1..L, k=1..K} / Sum of all contract amount_i. i denotes firm, k denotes federal agency and l denotes federal agency account. Average sequester ratio for each firm. For each firm, the sum of the federal contract amount multiplied by the average sequestration ratio of the corresponding Federal agency's account scaled by the total dollar amount all federal contracts obtained by that firm. The calculation is based on the federal contracts obtained in the 2011–2012 fiscal period (October 1, 2010–September 30, 2012).
  - Source(s): usaspending.gov; OMB (https://www.whitehouse.gov/omb/budget/Historicals)
- Weighted Average Sequester Ratio
  - Formula (descriptive): ∑∑ (contract amount_ikl × weighted average seqratio_kl)_{l=1..L, k=1..K} / Sum of all contract amount_i. i denotes firm, k denotes federal agency and l denotes agency account. Weighted average sequester ratio for each firm. In the OMB file, some accounts have different sequestration ratios for the same Federal Agency. Weighted average of these ratios based on the amount granted by each Federal Agency account in a Federal Agency is constructed. Weighted Average Sequester Ratio reports, for each firm, the sum of the federal contract amount multiplied by this weighted average sequestration ratio of the corresponding Federal agency's account scaled by the total dollar amount all federal contracts obtained by that firm. The calculation is based on the federal contracts obtained in the 2011–2012 fiscal period (October 1, 2010–September 30, 2012).
  - Source(s): usaspending.gov; OMB file (https://www.whitehouse.gov/omb/budget/Historicals)

### Contract and Firm-Level Controls
- Federal Contracts
  - A firm’s total federal contract amounts in 2011–2012 fiscal periods (October 1, 2010–September 30, 2012)
  - Source(s): usaspending.gov
- Size
  - Natural logarithm of the total assets (ATQ), measured at the beginning of the period
  - Source(s): COMPUSTAT
- Tobin's Q
  - (Total assets (ATQ) + Quarter-end share price (PRCC_Q)*Number of shares outstanding (CSHOQ) - Book value of equity (CEQQ))/Total assets, measured at the beginning of the period
  - Source(s): COMPUSTAT
- R&D
  - Research and Development Expense (XRDQ), scaled by beginning-of-period total assets (ATQ)
  - Source(s): COMPUSTAT
- ROA
  - Return on assets. Return on assets is (Net Income (NIQ)/Beginning of period Total assets (ATQ).
  - Source(s): COMPUSTAT
- ROA (Industry adjusted)
  - ROA adjusted to industry by subtracting mean industry ROA values at the 2-digit SIC level are subtracted from firm ROA values.
  - Source(s): COMPUSTAT
- ROE
  - Return on equity. Return on assets is (Net Income (NIQ)/Beginning of period stockholders’ equity (TEQQ).
  - Source(s): COMPUSTAT
- ROE (Industry adjusted)
  - ROE adjusted to industry by subtracting mean industry ROE values at the 2-digit SIC level are subtracted from firm ROE values.
  - Source(s): COMPUSTAT

### Industry and Competition Measures
- Industry Concentration
  - Annual basis Herfindahl-Hirschman Index based on Text-Based Network Industries (TNIC), measured at the beginning of the period
  - Source(s): Hoberg-Phillips Data Library
- Firm Level Competition
  - Ratio of a firm’s competitive contracts to total contracts in 2011-2012 fiscal periods. A contract is competitive if “Extend Competed” is either A: Full and Open Competition or CDO: Competitive Delivery Order
  - Source(s): usaspending.gov
- Agency Level Competition
  - Agency level competition is as follows:
    - 푎푣푔.푐표푚푝푒푡푖푡푖표푛.푎푔푒푛푐푦_푖 = ∑_{j=1..m} N_{ij} C_j / ∑_{j=1..m} N_{ij}
    - In this calculation, C_j is the ratio of competitive contracts to total contracts in agency j. N_ij is the total number of contracts firm i gets from agency j. m is the number of firms in the sample.
  - Source(s): usaspending.gov
- Government Dependence
  - Government dependence is the industry exposure to government spending determined by Belo-Gala-Li (2013) at the sector level based on the Benchmark Input-Output Accounts released by Bureau of Economic Analysis. This measure is used at the 2-digit SIC level for the sample.
  - Source(s): Belo, Gala and Li (2013)

*Source: Appendix 1. Variable Descriptions and Data Sources (wpiea2019172-print-pdf).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019172-print-pdf.pdf_