Enhancing sectoral collective bargaining in Chile

4. Recent OECD evidence from macro and micro data

The economic literature has long debated the role of collective bargaining for labour market performance but paid little attention to the system of collective bargaining. In the early eighties, attention focussed very much on the role of centralisation, but the debate then also looked at other features such as the degree of wage co‑ordination across bargaining units.

The main results from macro-data are based on country-level data on labour market outcomes and a characterisation of collective bargaining systems developed by the OECD. Five categories of collective bargaining systems are identified OECD (2019[1]):

• Predominantly centralised and weakly co‑ordinated systems: Sectoral agreements play a strong role, extensions are relatively widely used, derogations from higher-level agreements are possible but usually limited or not often used, and wage co‑ordination is largely absent.

• Predominantly centralised and co‑ordinated systems: As in the previous category, sectoral agreements play a strong role and the room for lower-level agreements to derogate from higher-level ones is quite limited. However, wage co‑ordination is strong across sectors.

• Organised decentralised and co‑ordinated systems: Sectoral agreements play an important role, but they also leave significant room for lower-level agreements to set the standards – either by limiting the role of extensions, leaving the design of the hierarchy of agreements to bargaining parties or allowing opt-outs. Co‑ordination across sectors and bargaining units tends to be strong.

• Largely decentralised systems: Firm-level bargaining is the dominant bargaining form, but sectoral bargaining or wage co‑ordination also play a role. Extensions are very rare.

• Fully decentralised systems: Bargaining is essentially confined to the firm or establishment level with no co‑ordination and no (or very limited) influence by the government. Chile is part of this group, together with Colombia, Costa Rica, Mexico, New-Zealand and the United States among others.

The analysis compares labour market outcomes under these different collective bargaining systems relative to the fully decentralised system, while controlling for the level of bargaining coverage as well as the possible role of the business cycle, the characteristics of the workforce and persistent country-specific features. The results also account for other policy reforms that occurred at the same time, in the areas of labour taxation, product market regulation, job dismissal regulation, minimum wages and unemployment benefits.

4.1. Better employment outcomes with co‑ordinated bargaining systems

Co‑ordinated bargaining systems are associated with higher employment and lower unemployment relative to fully decentralised systems (Panel A of Figure 8). This is particularly the case for predominantly centralised systems, while for organised decentralised systems the result on unemployment is somewhat smaller and less robust. Centralised but weakly co‑ordinated systems and largely decentralised systems hold an intermediate position, with better employment outcomes than in fully decentralised ones but similar unemployment outcomes. The difference between the employment and unemployment results suggests that such systems are linked with higher employment and labour force participation.

It is sometimes argued that collective bargaining delivers good labour market outcomes for “insiders” (notably prime‑age male full-time workers with a permanent contract) at the expense of jobs for “outsiders”, such as youth, women and low skilled. The evidence, however, suggests that, in most cases, co‑ordinated systems – either centralised or organised decentralised – are associated with better labour market outcomes for vulnerable groups (Panels B and C of Figure 8). The unemployment rates of youth, women and low-skilled workers appear to be consistently lower (or at least not higher) in co‑ordinated systems than in decentralised ones.

Figure 8. Collective bargaining systems and employment outcomes

Difference in percentage points with respect to fully decentralised systems

Note: ***, **, *: statistically significant at the 1, 5 and 10% levels, respectively. Results are based on OLS regressions including country and year dummies, collective bargaining coverage, log of average years of education, female employment share and institutional variables: (tax wedge, product market regulation, employment protection legislation (both temporary and permanent), ratio of minimum wage to median wage and gross unemployment benefit replacement rate).

Source: OECD estimates.

4.2. Lowest wage dispersion with sectoral bargaining

Drawing on country-level data and the same collective bargaining taxonomy, collective bargaining systems that are not fully decentralised are found to be correlated with lower wage inequality for full-time employees (Figure 9), as measured by the D9/D1‑ratio, i.e. the ratio of the wage at the 9^th decile of the wage distribution to the wage at the 1st decile. This association is present both in the lower and upper half of the wage distribution.1

Strengthening the bargaining power of low-wage workers is one of the core missions of collective bargaining, so it is not surprising that empirically collective bargaining is associated with lower levels of inequality. Detailed pay scales, where they are defined, can compress wages in the middle and top of the distribution to compensate for higher wages at the bottom. These mechanisms are particularly relevant when bargaining covers a substantial share of the working population.

Figure 9. Collective bargaining and wage dispersion

Point difference with respect to fully decentralised systems

Note: ***, **, *: statistically significant at the 1, 5 and 10% levels, respectively. Results are based on OLS regressions including country and year dummies, collective bargaining coverage, log of average years of education, female employment share and institutional variables: tax wedge, product market regulation, employment protection legislation (both temporary and permanent), ratio of minimum wage to median wage and gross unemployment benefit replacement rate. Earnings inequality measures are based on gross earnings of full-time wage and salary workers. D1, D5 and D9 stand for the first, fifth and ninth decile of the wage distribution.

Source: OECD estimates.

Worker-level data on collective bargaining coverage are also available for 21 OECD countries. Besides distinguishing workers covered by collective bargaining from those who are not, the micro-data separately identify workers whose wage is primarily determined by a firm- as opposed to a sectoral agreement. This makes it possible to distinguish three bargaining levels: i) individual or no collective bargaining; ii) firm-level bargaining; and iii) sectoral bargaining. Analysis based on these micro data provides further evidence on the positive role of collective bargaining for wage equality. On average, earnings dispersion is lower with collective bargaining, when accounting for compositional differences (Figure 10). In the first group of countries where all three bargaining levels co‑exist, wage dispersion is highest among workers not covered by collective bargaining, followed by firm-level and then sectoral bargaining. By contrast, for the second group of countries where there is no sectoral bargaining, wage dispersion among workers covered and those not, at least on average, is the same. A cross-country comparison of the averages for the first two groups suggests that firm-level bargaining is only effective in lowering wage dispersion when it comes on top of sectoral bargaining. One possible explanation for this may be that companies characterised by firm-level bargaining are in most cases also covered by sectoral bargaining. Nevertheless, overall, they are consistent with those drawing on country-level data which suggested that the economy-wide distribution of wages is less equal in systems without scope for sectoral or higher-level bargaining (see Figure 9).

Figure 10. Accounting for the differences in wage dispersion with and without collective bargaining

Change in the ratio of the 9th to the 1st earnings decile relative to employees not covered by collective bargaining (adjusted for composition), 2014

Note: Results are based on Juhn-Murphy-Pierce decompositions using workers without a collective agreement as the reference group and controlling for gender, age groups, educational attainment, industry, occupation, firm size, type of contract and job tenure. Data are from 2012‑16, depending on the country (2006 for Germany). For countries with three bargaining types, data are available for firm- and sectoral bargaining and no collective bargaining. For countries with two bargaining types, data are available for firm-level bargaining and no collective bargaining. “Sectoral bargaining” for Australia refers to the use of Modern Awards. A proper sectoral bargaining does not exist in Australia.

Source: OECD calculations based on the European Structure of Earnings Survey for European countries, the Household, Income and Labour Dynamics survey for Australia, the Labour Force Survey for Canada, the Labour and Income Panel Study for Korea, the National Survey of Occupation and Employment for Mexico and the Current Population Survey Merged Outgoing Rotation Groups for the United States.

The inequality results in this report complement previous findings that point in the same direction, from earlier studies by Blanchflower and Freeman (1993[4]), Blau and Kahn (1999[5]), Card, Lemieux and Riddell (2004[6]) and DiNardo and Lee (2004[7]) to more recent ones including OECD (2011[8]), ILO (2015[9]) and Jaumotte and Buitron (2015[10]).

4.3. Wage premium of firm-level bargaining

Workers are paid more with firm-level bargaining, while sectoral bargaining is not associated with relatively higher pay on average (Figure 11). This is not surprising as firm-level negotiations can often only raise wages relative to sectoral agreements. The differences in wages may also signal higher productivity in companies with firm-level bargaining. The results are in line with a large body of the literature which finds that sectoral bargaining is not linked with higher wages on average – see Dell’Aringa and Lucifora (1994[11]), Hartog, Leuven and Teulings (2002[12]), Rycx (2003[13]) and Cardoso and Portugal (2005[14]).

The variation for sectoral bargaining across countries is large, with a positive premium in some countries and a negative one in others. By contrast, wages of workers covered by firm-level agreements are higher than those of uncovered workers in all countries except Latvia. In countries with low collective bargaining coverage, wage inequality can thus rise as firm-level bargaining expands to include more workers, even if wage dispersion is smaller among workers covered by firm-level bargaining than among those who are not.

Figure 11. Wage premium by level of collective bargaining

Composition-adjusted difference in average wages relative to no collective bargaining, 2014

Note: Results are based on Juhn-Murphy-Pierce decompositions using workers without a collective agreement as the reference group and controlling for gender, age groups, educational attainment, industry, occupation, firm size, type of contract and job tenure. Data are from 2012‑16, depending on the country (2006 for Germany). “Sectoral bargaining” for Australia refers to the use of Modern Awards. A proper sectoral bargaining does not exist in Australia.

Source: OECD calculations based on the European Structure of Earnings Survey for European countries, the Household, Income and Labour Dynamics survey for Australia, the Labour Force Survey for Canada, the Labour and Income Panel Study for Korea, the National Survey of Occupation and Employment for Mexico and the Current Population Survey Merged Outgoing Rotation Groups for the United States.

4.4. Lower productivity growth in centralised systems with limited scope to firm level bargaining

The stronger wage compression with collective bargaining documented above may reflect a misalignment of wages with a firm’s or sector’s productivity because centralisation or co‑ordination makes pay also determined by factors other than the firm or sector.

The extent to which wages in a particular firm or sector correspond to the productivity in the form or sector can be estimated with available data. By comparing countries with one another, OECD (2019[1]) provides suggestive evidence2 that wages tend to be less aligned with labour productivity in countries where collective bargaining institutions have a more important role. Examining the different features of collective bargaining systems that may matter for wage‑productivity alignment, the study concludes that co‑ordination, collective bargaining coverage and centralisation jointly predict lower wage‑productivity alignment. The empirical evidence, which is based on cross-country comparisons, is however not enough for proving that such features of collective bargaining are the driving or causal factors behind the differences in age‑productivity alignments. It is nonetheless suggestive that collective bargaining has an important role for how wages in a sector respond to sector performance.

Overall, in countries where wage co‑ordination has an important role or wages are more centralised at sectoral level, the correlation of wages with productivity at the sub-national level is weaker. This suggests that wage co‑ordination “works”, in the sense that it co‑ordinates wages and, by partially delinking wages from productivity, may end up in a less dispersed wage distribution. Centralisation and co‑ordination may also affect how wages can respond to individual firm performance. In the longer term, such delinking of wages from productivity could have potentially important implications for productivity growth. It could reduce incentives for workers to innovate, work hard and move to a better-paid job.

However, stronger misalignments of wages from productivity do not need to have such negative effects; for example, they may even increase innovation incentives, if firms would reap the full benefits of productivity gains. Box 2 summarises the existing literature on collective bargaining and productivity.

Box 2. Collective bargaining and productivity growth

How does collective bargaining influence productivity? Theory suggests that effects could go either way. On the one hand, collective bargaining can increase aggregate productivity by setting higher wage floors (and making it more difficult to cut costs through lower wages) which may force unproductive firms to exit the market (Braun, 2011[15]). More rigid wages may also increase the incentives of the firms’ owners to innovate, as they would reap the full benefits of productivity gains – see Acemoglu and Pischke (1999[16]) and Haucap and Wey (2004[17]). Other ways through which collective bargaining could promote productivity growth are higher “efficiency” wages, better non-wage working conditions and the possibility for workers to voice concerns.

On the other hand, a more compressed wage structure may reduce the incentives to work hard and move to a more productive firm, harming firm productivity and the efficient reallocation of workers. Union power could also allow workers appropriating the benefits of investments by employers, giving rise to the so-called “hold-up” problem (Malcomson, 1997[18]) and reducing investment incentives for firms. Further, limitations to adjustments in the organisation of work (such as in working time, shifts or leave) could lower productivity. Finally, decentralisation of bargaining may promote productivity through a more frequent use of incentive schemes (such as performance pay).

The empirical literature has examined quite extensively the role of union coverage for productivity. According to a meta‑analysis (Doucouliagos, Freeman and Laroche, 2017[19]), the evidence overall suggests that union coverage increases productivity in non-manufacturing industries, but not in manufacturing industries. Some papers studied empirically the relevance of collective bargaining for the “hold-up” problem and investment, with inconclusive results overall. Card, Devicienti and Maida (2014[20]), using matched employer-employee data from Italy’s Veneto region, obtain little evidence of hold-up. Based on sector-level data for OECD countries, Cardullo, Conti and Sulis (2015[21]) find that union coverage reduces investment in sunk-capital-intensive industries relative to others.

The OECD empirical results reported in this section suggest that certain collective bargaining systems can be associated with stronger misalignments of pay and productivity, with possible consequences for productivity growth. However, few papers have directly studied the role of different features of bargaining systems, such as centralisation or co‑ordination, for productivity, in part due to lack of suitable data. A few results from empirical papers on this topic are summarised in OECD (2019[1]). Taking the evidence together, decentralisation appears to improve firm productivity, while it may slow down the cleansing effect of higher wages and therefore, due to composition effects, not translate in higher aggregate productivity growth. When investigating the links of centralisation and co‑ordination with productivity growth, results suggest that centralised bargaining may come at the expense of lower productivity growth, although analysis beyond these empirical explorations is needed to examine the links between bargaining regimes and productivity further.

While further empirical work seems necessary to assess the link between bargaining regimes and productivity more in depth, OECD results tend to call for combining sectoral agreements with some role for further negotiation at firm level in order to better align productivity and pay, as well as foster productivity by promoting some wage differentiation across firms and workers. This result suggests that the articulation between sector and firm-level bargaining needs to be carefully fine‑tuned.

Notes

← 1.  While decreasing wage inequality among full-time workers, collective bargaining may increase earnings inequality between full-time employees and other workers, in the spirit of an insider-outsider model above on the limited empirical backing of such model). Since the data in this analysis are based on hourly wages of full-time workers, they cannot be used to study effects on overall earnings inequality among all workers.

← 2.  The analysis relies on insights using sector-level data, examining the correlation between wage sand productivity across sectors.