## _wp16143

## Source details

**Canonical URL:** [_wp16143](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16143.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16143.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16143.pdf.json)

---

### Introduction and motivation
- Renewed interest since the global financial crisis in how collective bargaining affects economic performance, particularly employment responsiveness to shocks.
- Well-functioning collective bargaining can:
  - increase responsiveness of working conditions to economic shocks via coordination or decentralization (Blanchard et al., 2013; Dustmann et al, 2014).
  - support wages and thus consumption and aggregate demand where there is slack, low inflation, and limited macro policy space.
- Poorly functioning systems can reduce responsiveness of wages and working hours to aggregate shocks, increasing social costs of downturns through slower labor-market adjustment and greater labor shedding.

### Economic and institutional context (Portugal)
- Pre-2011 institutional facts:
  - Collective bargaining took place almost exclusively at the sectoral level.
  - Union density rate was estimated at 11% in 2009.
  - Coverage of sectoral collective agreements was very high due to administrative extensions; extensions entered into force retro-actively at the same time as underlying collective agreements.
  - Typical administrative delay in late 2010–early 2011 was about 180 days.
- Policy change exploited:
  - Government decision on 21 June 2011 to immediately suspend extensions, creating a sharp discontinuity in extension probability around 1 March 2011.
- 2012 reform changes:
  - Extensions reintroduced conditional on representativeness: employer association must represent firms that together account for 50% of the workforce of the relevant sector.
  - Extensions entered into force at the date of the administrative decision (no retroactivity).
- 2014 reform change:
  - Alternative representativeness criterion: met if more than 30% of firms affiliated to signatory employer association consisted of small or medium-sized enterprises (firms employing less than 250 employees).

### Natural experiment and empirical strategy
- Identification:
  - Exploit unexpected suspension of extensions (21 June 2011) and administrative delays to construct a Regression Discontinuity Design (RDD) around 1 March 2011.
  - Sharp RDD: drop two agreements signed between 24 January and 28 February 2011 that were not extended so that all agreements signed before 1 March 2011 were extended and all those signed after were not.
  - Fuzzy RDD: alternative that accounts for incomplete sharpness during 24 January–28 February 2011; estimated via 2SLS.
- Empirical implementation:
  - Pool data across affiliated and non-affiliated firms; estimate overall impact and differential effects by affiliation (interaction of treatment dummy and affiliation-status dummy).
  - Relative time effects modeled as Linear or Quadratic and allowed to differ across sides of the threshold.
  - Regressions weighted by the number of employees in 2010.
  - Standard errors robust and clustered by collective agreement and signature date.
- Validity checks:
  - Tests for anticipation (frequency of agreements per week) show no decline in average number of agreements after 1 March 2011.
  - Balancing tests on pre-determined variables reveal some statistically significant differences; therefore regressions include additional controls (e.g., pre-reform employment growth rate, average firm size, average export intensity) and focus on change in growth rate of employment to account for differential trends.

### Data and dataset construction
- Main sources:
  - Administrative Personnel Records (Quadros de Pessoal, QP): matched employer-employee panel covering private sector universe; focus period 2009-2013; worker info refers to October of each year; employer association membership available for 2010.
  - Collective agreements and extensions data from DGERT (Ministry of Labor): publication dates, signatory associations/unions, economic activities, regions, occupations.
- Empirical sample:
  - 38 collective agreements signed over October 2010 to August 2011, accounting for approximately 20% of the private sector workforce.
  - Focus on 31 agreements signed between 8 October 2010 and 20 June 2011 for the RDD analysis.
  - Resulting semi-aggregated dataset: 76 observations (38 agreements for each affiliation status group).
- Unit of aggregation: agreement × year × membership status.
- Variables in semi-aggregated dataset: employment, wages, sales, exports, productivity (sales per worker).

### Main empirical findings: employment effects
- Overall result:
  - Extensions have an adverse and statistically significant impact on employment growth.
  - Magnitude of estimated adverse impact: ranging from 5 to 10 percentage points depending on specification.
  - Large estimated effects likely reflect the 2010Q4-2012Q4 recession context.
- Selected reported regression statistics (preserving source formatting)
  - Panel A (Employment growth, p.p., 2010-2011 (ΔE_t)):
    - (1)(2)(1)(2) Treatment dummy- 0.0953-0.0500- 0.0578- 0.0420
      - ( 0.0446)( 0.0591)( 0.0262)( 0.0262)
    - Constant- 0.1042-0.1358- 0.1226- 0.1293
      - ( 0.0388)( 0.0274)( 0.0313)( 0.0236)
    - Relative time effects: Linear Quadratic Linear Quadratic
    - Observations62625858
    - R-squared0.39760.40370.42900.4359
    - Labels: Fuzzy Sharp
  - Panel B (Change in employment growth, p.p., (ΔE_t - ΔE_{t-1})):
    - (1)(2)(1)(2) Treatment dummy- 0.1574-0.1472- 0.1022- 0.1035
      - ( 0.0548)( 0.0709)( 0.0301)( 0.0295)
    - Constant- 0.1950-0.1985- 0.2238- 0.1926
      - ( 0.0380)( 0.0360)( 0.0279)( 0.0247)
    - Relative time effects: Linear Quadratic Linear Quadratic
    - Observations62625858
    - R-squared0.65790.65810.68090.6958
    - Labels: Fuzzy Sharp
  - Statistical significance notation in source: *, **, *** refer to statistical significance levels of 10%, 5% and 1% respectively.
- Sensitivity checks:
  - Results qualitatively similar across fuzzy and sharp RDD, and whether linear or quadratic relative time controls are used.
  - Excluding controls and extending bandwidth (Oct 2010–Aug 2011) do not qualitatively change baseline results.
  - Placebo (falsification) test with fictional reform date mid-May 2011 does not indicate discontinuities.

### Heterogeneity by affiliation status and mechanisms
- Concentration of adverse effects:
  - Adverse impact of extensions tends to be concentrated among non-affiliated firms in most specifications, particularly when focusing on the change in employment growth (preferred specification).
  - Fuzzy RDD: extensions negative and statistically significant for non-affiliated firms, but positive and statistically significant for affiliated firms in some specifications.
  - Sharp RDD: mixed results across specifications.
- Representativeness:
  - Representativeness measure: share of workforce in affiliated firms in total sector employment.
  - Empirical finding: degree of representativeness does not appear to significantly affect the impact of extensions for either affiliated or non-affiliated firms (Table 5a: almost all interaction coefficients insignificant).
  - Possible explanations: limited number of agreements above 50% representativeness; majority-rule criteria may be insufficient; bargaining driven by market leaders with small affiliated firms.
  - Longer-term view: representativeness criteria may promote employer organization gradually even if immediate employment mitigation is not observed.
- Retro-activity and administrative delay:
  - Administrative delay measured as number of days between entry-into-force date among affiliated firms and publication date of extension.
  - Retro-activity amplifies adverse effects on non-affiliated firms because extensions applied retro-actively, generating wage arrears liabilities for non-affiliated firms.
  - Quantitative notes:
    - Typical administrative delay ≈ 180 days.
    - Interaction of treatment with administrative delay: negative effect for non-affiliated firms; insignificant or positive for affiliated firms.
    - Counterfactual (administrative delay = 0): fuzzy RDD indicates retro-activity accounts for a substantial part of the adverse impact—negative effect on employment growth in non-affiliated firms falls by approximately 40%, though effects remain sizeable and concentrated among non-affiliated firms. Sharp RDD: suggests no effect in absence of administrative delay.
  - Caveat: results rely on assumed linear relationship between administrative delay and employment impact.

### Effects on wages and wage inequality
- Method:
  - Analyze changes in residual wages and wage inequality within job categories, agreements, membership status and years.
  - Sample restricted to workers employed in both 2010 and 2011.
  - Dependent variables: changes in 5th, 10th, 15th, 20th percentiles of residual wage distribution and differences relative to the median.
- Main findings:
  - Extensions tend to increase wage growth at the bottom of the distribution and reduce inequality.
  - Example estimates:
    - Panel B sharp RDD: Treatment effect 0.0493 in ∆p5 (increase in wage growth for the fifth percentile).
    - Panel A fuzzy RDD: reports 0.0805 in ∆p5 raw.
  - Extensions tend to reduce growth rate of P50/P5 ratio by a similar amount (negative coefficients in ∆(p50-p5)).
  - Effects diminish moving up the distribution; effects on 20th percentile are insignificant in all specifications.
- Interpretation:
  - Evidence consistent with a binding extension mechanism that compresses wages at the bottom but creates a trade-off with employment.

### Policy-relevant conclusions and recommendations
- Key conclusions:
  - Extensions had an adverse role for employment growth during the sample recessionary period and likely amplified unemployment response to the global financial crisis.
  - Adverse effects mainly concentrated among non-affiliated firms, pointing to representativeness concerns.
  - Representativeness measures in the data do not show significant short-term mitigation of employment effects, but stricter criteria implemented in 2012 likely reduced the number of extensions and indirectly supported employment growth.
  - Retro-activity associated with extensions until 2012 appears harmful for employment among non-affiliated firms; removing retro-activity and shortening administrative delay can mitigate adverse employment effects while preserving sectoral bargaining.
  - Trade-off exists between employment costs and wage-compression benefits for low-wage workers.
- Policy implications explicitly noted in source:
  - Shortening administrative delay associated with issuing extensions can partly address concerns that removing retro-activity undermines sectoral bargaining objectives.
  - Representativeness criteria may be useful, especially if implemented gradually, to promote employer organization and improve industrial relations and trust—even if immediate mitigation of employment effects is not evident.
- Research implication:
  - Further research in other countries and periods is needed to assess generalizability.

*Authors’ calculations based on QP and DGERT data; document: _wp16143 - References*

### References .............................................................................................................

### References

### Introduction and motivation
- Renewed interest since the global financial crisis in how collective bargaining affects economic performance, particularly employment responsiveness to shocks.
- Well-functioning collective bargaining can:
  - increase responsiveness of working conditions to economic shocks via coordination or decentralization (Blanchard et al., 2013; Dustmann et al, 2014).
  - support wages and thus consumption and aggregate demand where there is slack, low inflation, and limited macro policy space.
- Poorly functioning systems can reduce responsiveness of wages and working hours to aggregate shocks, increasing social costs of downturns through slower labor-market adjustment and greater labor shedding.

### Extensions, coverage, and representativeness
- In many countries (particularly in Europe), collective bargaining coverage is often widened via administrative extensions that apply agreements beyond signatory parties (typically to all firms and workers in the same sector).
- Extensions have helped keep collective bargaining coverage high even as union density has declined.
- Empirical indicator: where the difference between bargaining coverage and union density is large (e.g., France), extensions tend to be important.
- Note on intra-firm coverage: agreements applying "erga omnes" within signatory firms extend coverage within firms irrespective of individual worker union membership; Portugal historically applied a "double affiliation principle" in the absence of extensions, where agreements only applied to unionized workers in affiliated firms.

### Theory and prior empirical evidence
- Theoretical concerns:
  - Moll (1996): with heterogeneous firms, extensions can benefit large/productive signatory firms by reducing competition from less productive firms and allowing lower wages; non-unionized low-productivity firms may be forced to pay higher wages, reducing low-wage competition.
  - Haucaup et al. (2001): employer associations can use extensions as anti-competitive devices to raise labor costs of non-organized rivals.
- Empirical findings from previous studies:
  - Murtin et al. (2014): using country-level panel data, excess coverage (bargaining coverage minus union density) is not systematically correlated with higher unemployment, though adverse effects may be stronger where the tax wedge is higher.
  - Magruder (2012): in South Africa, extensions in industries within certain districts reduce employment by around 10% relative to the same industry in uncovered neighbouring districts.
  - Martins (2014): for Portugal (2008-2011), extensions decrease sectoral employment by 2% over the four months following an extension, particularly among smaller firms.
  - Hartog et al. (2001): little role for extensions in the Netherlands.
  - De Ridder and Euwals (2016): suggestive evidence that extensions increase wages, but do not analyse employment effects.
- Gap in literature: prior work has not directly examined the role of representativeness by accounting for firms’ membership in employer associations.

### Natural experiment and empirical strategy
- The paper exploits a natural experiment in Portugal: the immediate suspension of extensions by the government that took office in June 2011.
- Method: regression discontinuity design (RDD) that leverages the administrative delay between agreement conclusion and extension issuance together with the June 2011 suspension.
- Key empirical leverage: a sharp and unanticipated decline in the probability an extension was issued several months prior to the policy change, creating plausibly exogenous variation.
- Data: matched employer-employee dataset merged with information on collective agreements and extensions including timings.

### Main empirical findings
- Extensions had a negative effect on employment growth during 2010-2011:
  - Amounting to 5 percentage points or more depending on specification.
- Negative effects are concentrated among non-affiliated firms, implying lack of representativeness of employer associations is an important factor behind adverse effects.
- Adverse effects of extensions are larger the longer the administrative delay in processing extensions.
  - This reflects the role of retro-activity (prior to the 2012 labor reform): non-affiliated firms were required to pay wage arrears from the original agreement entry-into-force to the time the extension was issued.
- Evidence suggests adverse employment effects may operate via impacts on wages in the bottom part of the wage distribution, pointing to a potential trade-off between wage and employment effects of extensions.

### Interpretation and mechanisms
- Representativeness: more representative employer associations may negotiate agreements better suited to majority of firms, yielding smaller adverse employment effects; the observed concentration of negative effects among non-affiliated firms is consistent with representativeness concerns.
- Retro-activity and administrative delay: longer delays increase the cumulative wage arrears liabilities for non-affiliated firms when an extension is eventually issued, amplifying negative employment effects.
- Context dependence: large estimated effects likely reflect the recessionary context during which the natural experiment occurred.

### Paper structure (as presented)
- Section 2: economic and institutional context around the natural experiment.
- Section 3: description of the natural experiment, the RDD approach, and validity considerations.
- Section 4: description of the matched employer-employee dataset and collective agreement/extension timing data.
- Section 5: empirical evidence on the impact of administrative extensions overall and separately for affiliated and non-affiliated firms; analysis of representativeness and retro-activity with administrative delay.

*Source: _wp16143 - References*

### Section 6 presents some additional results in relation to wage inequality. Section 7 concludes.

### _wp16143 - Section 6 presents some additional results in relation to wage inequality. Section 7 concludes.

### II. ECONOMIC AND INSTITUTIONAL CONTEXT
- During the 2000s, Portugal experienced low growth, declining international competitiveness and deepening macroeconomic imbalances (Blanchard, 2007).
- The global financial crisis prompted large increases in public deficits and loss of market access, leading to a request for financial assistance in April 2011 directed towards the European Union, the European Central Bank and the International Monetary Fund.
- Financial support was conditional on structural reforms and adjustment measures, including measures on collective bargaining given the potential importance of real wage adjustment to minimize employment losses.
- Until mid-2011:
  - Collective bargaining took place almost exclusively at the sectoral level; firm-level bargaining was limited to state-owned and former state-owned companies and a small number of very large firms or holdings.
  - Union density rate was estimated at 11% in 2009.
  - Coverage of sectoral collective agreements was very high due to administrative extensions which widened reach beyond signatory parties.
- Extension procedure prior to June 2011:
  - A specific request was needed from an employer association or trade union for extension of a new or revised collective agreement.
  - Government assessed economic and social desirability, partly via empirical analysis of the number of workers potentially affected.
  - Procedure tended to delay issuance by several months, but extensions were "virtually never in doubt", resulting in the extension of almost all collective agreements.
  - Extensions entered into force retro-actively, at the same time as the underlying collective agreements, producing backdated wage increases for non-signatory firms' workers.
- Concerns raised:
  - Extensions played a key role in high and stable coverage despite low membership of social partners, effectively removing scope for low-wage competition between affiliated and non-affiliated firms.
  - Extensions were increasingly seen as a source of downward wage rigidity, particularly for smaller, younger, less productive firms not represented in bargaining.
- Reforms after June 2011:
  - Government that took office in June 2011 temporarily suspended issuance of extensions while preparing reform of extension procedures as part of the troika-agreed structural reform package.
  - The 2012 labor law reform reintroduced extensions with two key changes:
    - Extensions made conditional on representativeness: employer association must represent firms that together account for 50% of the workforce of the relevant sector.
    - Extensions entered into force at the date of the administrative decision (no retroactivity).
  - Figure 2 (described) shows that the large majority of collective agreements signed between September 2010 and August 2011 had affiliated-firm workforce shares well below the 50% threshold.
  - Re-introduction in 2012 led only to a modest pick-up in administrative extensions (Figure 3), partly reflecting a sharp decline in the number of new or revised sectoral agreements.
- Further reform in July 2014:
  - Introduced alternative representativeness criterion: met if more than 30% of firms affiliated to signatory employer association consisted of small or medium-sized enterprises (defined as firms employing less than 250 employees).
  - This largely represented a return to pre-2011 situation and resulted in a modest pick-up in extensions in 2015.

### III. METHODOLOGY
- Natural experiment exploited:
  - Government decision on 21 June 2011 to immediately suspend extensions created a sharp discontinuity around 1 March 2011 in the probability that a collective agreement was extended in the 12 months following conclusion.
  - Description of discontinuity:
    - Until 24 January 2011, about six weeks before cutoff, all collective agreements were extended during the subsequent twelve months.
    - From 24 January 2011 to 28 February 2011 some agreements were extended, but not all.
    - From 1 March 2011 until 20 June 2011, no collective agreements were extended during the subsequent twelve months.
  - Suspension was unexpected and could not be anticipated by unions and employer associations conducting bargaining.
- Regression discontinuity design (RDD):
  - Use sharp RDD by dropping two agreements signed between 24 January and 28 February 2011 that were not extended so that all agreements signed before 1 March 2011 were extended and all those signed after were not.
  - Sharp RDD model (formal):
    - Outcome y_i modeled as function of treatment dummy D_i (1 if extended in 2011, 0 otherwise), function of relative time s(.) and controls X_i.
  - Fuzzy RDD alternative:
    - Accounts for non-fully sharp decline in probability of extension during 24 January–28 February 2011.
    - Treatment equation (2a) models probability of extension as function of constant, threshold dummy T_i and relative time function.
    - Outcome equation (2b) models y as function of predicted probability of extension and relative time function.
    - Outcome and treatment equations estimated with 2SLS using same sample.
- Empirical implementation:
  - Pool data across affiliated and non-affiliated firms and perform two exercises:
    - Estimate impact of extensions on total employment growth while controlling for affiliation status.
    - Estimate differential effects by adding interaction between treatment dummy and affiliation-status dummy, allow different relative time effects across affiliation groups, and control for independent effect of affiliation status.
  - Standard errors correct for heteroskedasticity and are clustered by agreement and agreement date.
  - Relative time effects are controlled because:
    - Dependent variable often measured from October 2010 to October 2011; exposure to extensions depends on timing.
    - Economic conditions may affect timing and contents of agreements.
  - Relative time modeled as linear or quadratic and allowed to differ across threshold sides; given small sample, focus mainly on (potentially asymmetric) linear relative time effects.
  - Regressions weighted by the number of employees in 2010 to obtain average effect of extensions on total employment.
  - Firms covered by firm- and holding-level agreements are excluded.

### III.C VALIDITY
- Key identifying assumption: suspension of extensions was unexpected and could not be anticipated by social partners.
- Tests for anticipation:
  - Frequency of agreements:
    - Figure 5 plots number of agreements per week (8 October 2010 – 20 June 2011) and finds no decline in average number of collective agreements published per week after 1 March 2011; average is slightly higher after March 2011.
    - Given limited weekly counts, a formal test on continuity of density is not informative.
  - Balancing tests:
    - A series of balancing tests estimate equation (1) using pre-determined variables as dependent variables: degree of representativeness (share of sector workforce in affiliated firms), pre-reform employment growth rate (2009-2010), log number of employees potentially covered, log average firm size (employees per firm), log average hourly wage (within job title and year), log median hourly wage (within job title and year), export intensity (exports as share of revenue), and log labor productivity (revenue per worker).
    - Tests control for linear or quadratic relative time effects and are conducted for pooled sample and separately by affiliation status.
    - Table 1 results suggest statistically significant differences for some variables between groups; given small sample and many variables, this may imply the natural experiment does not fully equate to a randomized experiment.
    - To address imbalances, regression analysis includes additional controls (e.g., pre-reform employment growth rate, average firm size, average export intensity) and focuses on change in growth rate of employment to account for differential trends.

### IV. DATA AND IMPLEMENTATION
- Main data: Administrative Personnel Records (Quadros de Pessoal, QP)
  - Matched employer-employee panel covering universe of firms and workers in the private sector in Portugal.
  - Focus period: 2009-2013. Worker information refers to October of each year; firm information refers to full year.
  - QP provides applicable collective agreement for each worker and employer association membership of firms (membership information available only for 2010).
  - Analysis focuses on firms present in 2010 and their workers; new firms cannot be considered.
  - Firms covered by firm- and holding-level agreements excluded.
- Collective agreements and extensions data (DGERT)
  - Information publicly available from Ministry of Labor (DGERT) website.
  - Dataset includes publication dates of collective agreements and, if applicable, extensions; signatory employer associations and trade unions; economic activities (and regions/occupations).
  - Empirical dataset:
    - 38 collective agreements signed over October 2010 to August 2011, accounting for approximately 20% of the private sector workforce.
    - Main focus on 31 agreements signed between 8 October 2010 and 20 June 2011.
    - Rationale: extension procedure suspended when new government took office on 21 June 2011.
    - This yields a 15-week window from 1 March 2011 to 20 June 2011 during which 18 new or revised agreements were signed and not extended during the subsequent 12 months.
    - Preceded by a 5-week transition (24 January 2011 – 28 February 2011) where most agreements were extended but not all.
    - Period from 8 October 2010 to 24 January 2011 corresponds to a 15-week window during which all new or revised agreements were subsequently extended.
    - From 8 October 2010 to 28 February 2011, 13 new or revised agreements were signed.

*Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16143.pdf*

### 4.3 Combining the information into a semi-aggregated dataset

### 4.3 Combining the information into a semi-aggregated dataset

### Dataset construction and linkage procedure
- Purpose: construct a dataset with information on employment, wages, sales, exports and productivity (sales per worker) by agreement, year, and membership status, combining information from QP with information from collective agreements (DGERT).
- Linking steps:
  - Focus on the agreement that represents the largest share of workers in the firm, since workers in each firm may be covered by more than one agreement.
  - Use the name of the employer associations in which affiliated firms are affiliated to link collective agreements in QP and DGERT.
  - Extend the linking between the two collective agreement codes (QP and DGERT) to non-affiliated firms by establishing the domain of each collective agreement in terms of firm identifiers, first using 2010 data (for which employer affiliation information is available) and then the remaining years.
  - Collapse the firm-level data by year, agreement and employer association membership status.

- Resulting dataset: 76 observations (38 agreements for each affiliation status group).

### Aggregation and variables
- Unit of aggregation: agreement × year × membership status.
- Variables included: employment, wages, sales, exports, productivity (sales per worker).
- Collapsing method: firm-level data collapsed by year, agreement and employer association membership status to yield semi-aggregated observations.

### Regression specifications and estimation details (as reported)
- Estimation approaches: fuzzy RDD and sharp RDD.
- Time controls: relative time effects modeled as either Linear or Quadratic.
- Regressions are weighted by the number of employees in 2010.
- Included controls: log average firm size, export intensity and affiliation status.
- Standard errors: robust and clustered by collective agreement and signature date.
- Note on visualization: For ease of presentation, the results presented in Figure 6 do not include any controls apart from affiliation status and relative time effects. (Footnote 13)

### Empirical findings on employment growth
- Overall statement: Extensions have an adverse and statistically significant impact on employment growth.
- Magnitude: The impact of extensions on employment growth is potentially large, ranging from 5 to 10 percentage points depending on the specification.
- Contextual caveat: These large effects are likely to reflect at least in part the specific context during which the natural experiment took place, namely that of the 2010Q4-2012Q4 recession hitting firms still recovering from the global financial crisis. Because the need for adjusting working conditions was presumably greater under such circumstances, extensions may have had a larger adverse impact on employment than they would have had in normal times.

### Selected reported regression statistics (preserving source formatting)
- Panel A / Panel B structure reported; regressions shown under Fuzzy and Sharp RDD and under Linear and Quadratic relative time effects.

- Panel A (Employment growth, p.p., 2010-2011 (ΔE_t)):
  - (1)(2)(1)(2) Treatment dummy- 0.0953-0.0500- 0.0578- 0.0420
    - ( 0.0446)( 0.0591)( 0.0262)( 0.0262)
  - Constant- 0.1042-0.1358- 0.1226- 0.1293
    - ( 0.0388)( 0.0274)( 0.0313)( 0.0236)
  - Relative time effects: Linear Quadratic Linear Quadratic
  - Observations62625858
  - R-squared0.39760.40370.42900.4359
  - Labels: Fuzzy Sharp

- Panel B (Change in employment growth, p.p., (ΔE_t - ΔE_{t-1})):
  - (1)(2)(1)(2) Treatment dummy- 0.1574-0.1472- 0.1022- 0.1035
    - ( 0.0548)( 0.0709)( 0.0301)( 0.0295)
  - Constant- 0.1950-0.1985- 0.2238- 0.1926
    - ( 0.0380)( 0.0360)( 0.0279)( 0.0247)
  - Relative time effects: Linear Quadratic Linear Quadratic
  - Observations62625858
  - R-squared0.65790.65810.68090.6958
  - Labels: Fuzzy Sharp

- Statistical significance notation in source: *, **, *** refer to statistical significance levels of 10%, 5% and 1% respectively.

### Interpretation and limitations highlighted in the source
- Results are qualitatively similar across fuzzy and sharp RDD, and whether linear or quadratic controls for relative time are included and whether the pre-reform trend is controlled for or not.
- The unusually large estimated adverse effects likely reflect the recessionary context (2010Q4-2012Q4) and may overstate effects that would obtain in normal times.

*Authors’ calculations based on QP and DGERT data.*

### Section 5.3, the retro-active entry-into-force of extensions in a context where many firms are

### _wp16143 - Section 5.3, the retro-active entry-into-force of extensions in a context where many firms are

### Effects of extensions on employment growth (overall)
- Regression discontinuity design (sharp and fuzzy RDD) indicates extensions had an adverse impact on employment growth during the sample period.
- Results are estimated from a sharp RDD using employment weights; standard errors robust and clustered by collective agreement and signature date.
- Sensitivity checks:
  - “Controls”: includes only affiliation status (omits average firm size and export intensity).
  - “Bandwidth”: extends observation window to October 2010–August 2011.
  - “Falsification”: placebo reform date (mid-May 2011) using March–August 2011 agreements.
  - Excluding controls and extending bandwidth do not qualitatively change baseline results.
  - Falsification test does not point to discontinuities around the fictional reform date.
- Note: coefficients can lose statistical significance when controlling for quadratic relative time effects when focusing on employment growth but remain largely unchanged when focusing on the change in employment growth.

### Heterogeneity by affiliation status (affiliated vs non-affiliated firms)
- Key empirical patterns:
  - Adverse impact of extensions on employment growth tends to be concentrated among non-affiliated firms in the majority of specifications, particularly when focusing on the change in employment growth (preferred specification).
  - Fuzzy RDD: impact of extensions is negative and statistically significant for non-affiliated firms, but positive and statistically significant for affiliated firms.
  - Sharp RDD: mixed results—employment growth results suggest effects concentrated among affiliated firms, while change in employment growth results suggest effects concentrated among non-affiliated firms.
- Quantitative notes from Table 4 (selected signs and significance):
  - Non-affiliated firms: negative treatment effects on employment growth and change in employment growth in many specifications (e.g., non-affiliated treatment dummy values in panels show negative coefficients).
  - Affiliated firms: in some specifications treatment effects are positive (indicating possible increases in pay for non-unionised workers or increased market power).
- Interpretation:
  - Affiliated firms may shape use of extensions to their benefit at expense of non-affiliated outsiders.
  - Representativeness of employer associations appears relevant: wage floors negotiated in collective agreements may be less appropriate for affiliated firms than for non-affiliated firms.

### Role of representativeness
- Representativeness measure: share of workforce in affiliated firms in total sector employment (consistent with 2012 reform criteria).
- Empirical findings:
  - Degree of representativeness does not appear to matter significantly for impact of extensions for either affiliated or non-affiliated firms (Table 5a: almost all interaction coefficients insignificant).
  - Robustness checks with representativeness thresholds (30% and 50%) yield similar results (additional robustness checks available upon request).
- Potential explanations offered:
  - Technical: identification issues due to very limited number of agreements with representativeness levels above 50%.
  - Substantive: majority-rule representativeness criteria may be insufficient—50% threshold may not be high enough or bargaining may be driven by market leaders with small affiliated firms lacking influence.
  - Representativeness criteria may need fine-tuning or supplementation (e.g., test of public interest as in the Netherlands).
- Longer-term role:
  - Representativeness criteria may promote employer organization over time, improving industrial relations and trust, even if short-term mitigation of adverse effects on employment is not evident.

### Role of retro-activity (administrative delay)
- Definition: degree of retro-activity measured by number of days between entry-into-force date of collective agreement among affiliated firms and publication date of extension to non-affiliated firms.
- Empirical facts and interpretation:
  - Typical administrative delay in Portugal in late 2010–early 2011 was about 180 days (about six months).
  - Retro-activity plays an important role in explaining adverse impact of extensions on employment among non-affiliated firms (Table 5b).
  - Negative average treatment effect reflects impact for the typical administrative delay (180 days).
  - Interaction of treatment effect with administrative delay gives impact of a 1 day increase in administrative delay on change in employment growth:
    - Negative for non-affiliated firms.
    - Insignificant or even positive for affiliated firms.
  - Economic logic: retro-activity hits non-affiliated firms directly (liquidity constraint bite) while potentially benefiting affiliated firms indirectly by reducing competition.
- Counterfactual (no retro-activity / zero administrative delay):
  - Re-estimating with administrative delay defined in absolute value (so treatment dummy captures effect when administrative delay = 0) reduces overall impact of extensions on employment growth and reduces concentration among non-affiliated firms.
  - Fuzzy RDD: retro-activity accounts for a substantial part of adverse impact—negative effect on employment growth in non-affiliated firms falls by approximately 40%, though effects remain sizeable and concentrated among non-affiliated firms.
  - Sharp RDD: suggests no effect on employment growth in either group in absence of administrative delay.
- Contextual interpretation:
  - If no uncertainty about extension and firms not liquidity-constrained, retro-activity would not be expected to slow employment growth.
  - In a rapidly deteriorating economy with many liquidity-constrained firms, retro-activity substantially slowed employment growth—consistent with evidence of large job losses in four months after extension issuance (Martins, 2014), mainly from reductions in hirings.
- Caution: results rely heavily on assumed linear relationship between length of administrative delay and impact on employment growth.

### Effects on wages and wage inequality
- Method:
  - Focus on changes in residual wages and wage inequality within job categories, agreements, membership status and years.
  - Sample restricted to workers employed in both 2010 and 2011 to limit composition effects.
  - Dependent variables: changes in 5th, 10th, 15th, 20th percentiles of residual wage distribution and their differences relative to the median.
- Main findings (Table 6):
  - Extensions tend to increase wage growth at the bottom of distribution and reduce inequality.
  - Extensions increase wage growth by about 5 percentage points for workers in the fifth percentile of residual wage distribution (Panel B sharp RDD: Treatment effect 0.0493 in ∆p5; Panel A fuzzy RDD reports 0.0805 in ∆p5 raw).
  - Extensions tend to reduce growth rate of P50/P5 percentile ratio by a similar amount (negative coefficients in ∆(p50-p5)).
  - Effects diminish moving up distribution; effects on 20th percentile are insignificant in all specifications.
- Interpretation:
  - Findings support a binding interpretation of the extension mechanism.
  - Evidence of trade-off between adverse employment effects and beneficial wage-compression effects for low-wage workers.

### Policy-relevant conclusions and implications
- Key summary conclusions:
  - Extensions had an adverse role for employment growth during the sample period and likely amplified unemployment response to the global financial crisis—caveat that results may not generalize to normal times or different institutional settings.
  - Adverse effects mainly concentrated among non-affiliated firms, suggesting extensions favor affiliated firms and highlighting representativeness concerns.
  - Degree of representativeness measured in the data is not found to significantly affect employment outcomes; however, stricter representativeness criteria implemented in 2012 likely reduced number of extensions and indirectly contributed positively to employment growth.
  - Retro-activity associated with extensions until the 2012 reform appears harmful for employment among non-affiliated firms; removing retro-activity and shortening administrative delay can mitigate adverse employment effects while addressing sectoral bargaining concerns.
  - Trade-off exists between employment costs and wage/compression benefits for low-wage workers.
- Policy implications explicitly noted:
  - Shortening administrative delay associated with issuing extensions can partly address concerns that removing retro-activity undermines sectoral bargaining objectives.
  - Representativeness criteria may be useful, especially if implemented gradually, to promote employer organization and improve industrial relations and trust—even if immediate mitigation of employment effects is not evident.
- Research implication:
  - Further research in other countries and periods is needed to assess generalizability.

*Authors’ calculations based on QP and DGERT data; results reported in the source document.*

### References

### References

### Bibliographic references
- Addison, J., P. Portugal & H. Vilares, 2015, “Unions and Collective Bargaining in the Wake of the Great Recession”, IZA Discussion Paper No. 8943.
- Blanchard, O., 2007, “Adjustment within the euro. The difficult case of Portugal”, Portuguese Economic Journal, 6(1), p. 1-21.
- Blanchard, O., F. Jaumotte & P. Loungani , 2014, "Labor market policies and IMF advice in advanced economies during the Great Recession," IZA Journal of Labor Policy, vol. 3(1), pp. 1-23.
- Cardoso, Ana Rute & Pedro Portugal, 2005, "Contractual Wages and the Wage Cushion under Different Bargaining Settings," Journal of Labor Economics, 23(4), 875-902, October.
- Dustmann, Christian, Bernd Fitzenberger, Uta Schönberg, & Alexandra Spitz-Oener, 2014, "From Sick Man of Europe to Economic Superstar: Germany's Resurgent Economy." Journal of Economic Perspectives, 28(1), 167-88.
- De Ridder, M. and R. Euwals, 2016, “What are the wage effects of extending collective labour agreements? Evidence from the Netherlands”, CPB Background Document, April 2016.
- Guimarães, P., F. Martins & P. Portugal, 2015, “Upward nominal wage rigidity”, paper presented at DNB-IMF conference in Amsterdam, 29-30 October 2015.
- Hahn, J., Todd, P., Van der Klaauw, W., 2001, “Identification and estimation of treatment effects with a regression-discontinuity design”, Econometrica, 69(1), pp. 201-209.
- Hartog, J., E. Leuven and C. Teulings, 2002, "Wages and the bargaining regime in a corporatist setting: the Netherlands," European Journal of Political Economy, Vl. 18(2), pp. 317-331.
- Haucap, Justus & Pauly, Uwe & Wey, Christian, 2001, "Collective wage setting when wages are generally binding An antitrust perspective," International Review of Law and Economics, 21(3), pp. 287-307, September.
- Hijzen, A., P. Martins and J. Parlevliet, 2016, “A comparison of collective bargaining in Portugal and the Netherlands”, mimeo.
- IMF World Economic Outlook, Chapter 3, 2016, “Time for a supply boost? Macroeconomic Effects of Labor and Product Market Reforms in Advanced Economies”, April.
- Lee, D. S. & Card, D., 2008, "Regression discontinuity inference with specification error," Journal of Econometrics, Vol. 142(2), pp. 655-674.
- Lee, D.S. and Lemieux, T., 2010, “Regression Discontinuity Design in Economics”, Journal of Economic Literature, 48(2), pp. 281-355, June.
- Magruder, Jeremy R., 2012, "High Unemployment Yet Few Small Firms: The Role of Centralized Bargaining in South Africa," American Economic Journal: Applied Economics, 4(3), pp. 138-66, July.
- Martins, P. S., 2009a, "Dismissals for Cause: The Difference That Just Eight Paragraphs Can Make," Journal of Labor Economics, 27(2), pp, 257-279.
- Martins, P. S., 2009b, "Rent sharing before and after the wage bill," Applied Economics, 41(17), 2133-2151.
- Martins, P. S., 2014, “30,000 Minimum Wages: The Economic Effects of Collective Bargaining Extensions”, IZA Discussion paper No. 8540, October 2014.
- Martins, P. S., Solon. G. and Thomas, J. T., 2012, "Measuring What Employers Do about Entry Wages over the Business Cycle: A New Approach," American Economic Journal: Macroeconomics, 4(4), pp 36-55.
- Moll, P.,1996, “Compulsory Centralization of Collective Bargaining in South Africa”, American Economic Review, 86(2), pp. 326-329.
- Murtin, F., A. de Serres, and A. Hijzen. “Unemployment and the coverage extension of collective wage agreements.” European Economic Review 71:C., 2014, pp. 52–66.

### Annex Table A1. Collective agreements and extensions
- Agreement sector Date agreement Date extension
- Car sale 08-10- 2010 10-01- 2011
- Viana do Castelo retail 08-10- 2010 29-12- 2010
- Clinical analysis labs 08-11- 2010 28-02- 2011
- Wine trade sector 22-11- 2010 28-02- 2011
- Football clubs (players) 15-12- 2010 22-03- 2011
- Cork industry, North, Office workers 29-12- 2010 26-04- 2011
- Wine industry, cellars 10-01- 2011 26-04- 2011
- Textile industry 24-01- 2011
- Hotels and restaurants, Centre and South 24-01- 2011 23-05- 2011
- Aveiro retail 22-02- 2011 23-05- 2011
- Electric and electronic products 22-02- 2011 23-05- 2011
- Ropes industry 28-02- 2011
- Chemical and pharmaceutical retail 28-02- 2011 30-05- 2011
- Wood 09-03- 2011
- Pharmaceutical products retail 29-03- 2011
- Merchandising firms 29-03- 2011
- Viseu retail 08-04- 2011
- Wheat 08-04- 2011
- Coffee 08-04- 2011
- Driving schools 08-04- 2011
- Fish preserve industry 26-04- 2011
- Bread manufacturing, Lisbon 26-04- 2011
- Guarda, retail 29-04- 2011
- Poultry, slaughter 09-05- 2011
- Farming 09-05- 2011
- Meat, retail, South 16-05- 2011
- Retail store houses 23-05- 2011
- Bread manufacturing 23-05- 2011
- Fishing 30-05- 2011
- Non-alcoholic beverages 30-05- 2011
- Cement 08-06- 2011
- Shoe manufacturing 15-07- 2011
- Farming, Abrantes 15-07- 2011
- Farming, Beja 22-07- 2011
- Construction sector 08-08- 2011
- Private schools 16-08- 2011
- Clothing 16-08- 2011
- Textile industry 16-08- 2011

*Document: _wp16143 - References*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16143.pdf_