Wage rigidities and business cycle fluctuations: A linked employer-employee analysis

Adamopoulou, Effrosyni; Bobbio, Emmanuele; De Philippis, Marta; Giorgi, Federico Article Wage rigidities and business cycle fluctuations: A linked employer-employee analysis IZA Journal of Labor Policy Provided in Cooperation with: IZA – Institute of Labor Economics Suggested Citation: Adamopoulou, Effrosyni; Bobbio, Emmanuele; De Philippis, Marta; Giorgi, Federico (2016) : Wage rigidities and business cycle fluctuations: A linked employer-employee analysis, IZA Journal of Labor Policy, ISSN 2193-9004, Springer, Heidelberg, Vol. 5, Iss. 22, pp. 1-32, https://doi.org/10.1186/s40173-016-0078-5 This Version is available at: https://hdl.handle.net/10419/194375 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ ORIGINAL ARTICLE Open Access Wage rigidities and business cycle fluctuations: a linked employer-employee analysis Effrosyni Adamopoulou * , Emmanuele Bobbio, Marta De Philippis and Federico Giorgi * Correspondence: Effrosyni.Adamopoulou@ bancaditalia.it Bank of Italy, Directorate General for Economics, Statistics and Research, Structural Economic Analysis Directorate, Via Nazionale 91, 00184 Rome, Italy Abstract: This paper analyses wage dynamics in Italy in the last 25 years with a special focus on the recent recession. Despite the rather rigid Italian institutional setting, using linked employer-employee data we find that wage rigidities, albeit always present, have been subdued during the recessionary years. Using complementary data, we verify that, although we only observe daily and not hourly wages, overtime hours are not the main mechanism behind this enhanced wage flexibility. We document the presence of a trade-off between wage and employment adjustments: firms historically displaying higher levels of wage rigidities were less able to modify wages but exhibited higher turnover. A higher share of temporary workers, whose contractual relationship may be costlessly terminated and whose wages are therefore more frequently negotiated, served instead as a significant wage flexibility enhancing margin. More broadly, we find that firms of larger dimension, with a higher share of blue collar workers, or belonging to a sector where bonuses represent a large part of annual earnings were the ones displaying a higher level of wage flexibility. JEL Classification: J31, J33 Keywords: Wage dynamics, Negotiated wages 1 Introduction Understanding what drives wage dynamics is important in order to explain why aggregate wages tend to be much less volatile over time than what standard macroeconomic models predict (Fig. 1). Moreover, it helps policymakers decide which policy interventions to prioritize during downturns. The relatively flat evolution of aggregate wages is usually explained through (i) the presence of wage rigidities, that is a well-known feature of many labour markets (see for example, Kahn 1997; Knoppik and Beissinger 2009; Devicienti et al. 2007; Dickens et al. 2007; and Holden and Wulfsberg 2008), and through (ii) cyclical changes in the composition of the workforce (Lemieux 2006), since lower-paid workers are usually more severely affected during recessions. Some recent literature (D’Amuri 2014; Adamopoulou et al. 2016; Daly et al 2011; Verdugo 2016) indeed finds that composition effects have driven up aggregate wages, particularly during the recent recession. This paper focuses on wage rigidities and studies the distribution of wage changes for job stayers over the last 25 years, with a particular focus on the Great Recession. © The Author(s). 2016 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 DOI 10.1186/s40173-016-0078-5 We evaluate the determinants of wage rigidities and we describe how firms, depending on their wage structure (e.g. share of bonuses on overall pay) and workforce composition, display very different levels of wage rigidity. Moreover, we study how firms reacted differently along the cycle, depending on their historical level of wage rigidity. In particular, we seek to answer whether firms which were structurally less able to adjust wages of job stayers reacted by adjusting employment more and whether these firms, by hiring new workers at a re-negotiated salary which corresponds more to the new cyclical conditions, managed to partially compress their average wage per employee, even in the presence of high levels of wage rigidity for stayers. To measure rigidities in daily wages, we use newly available administrative employeremployee matched data for Italy that cover the years between 1990 and 2014. We rely on measures of wage rigidity based on the asymmetry of the distribution of yearly wage changes for job stayers and we find important adjustments in wages during the recessionary years (2009–2013). These adjustments were mostly driven by large firms and were mainly affecting blue collars. Moreover, using a unique hand-collected dataset on negotiated wages for employees in the metalwork industry and in the wholesale and retail industry, we document that the majority of these wage adjustments were enacted through the part of the wages that is not nationally negotiated. In addition, we show that changes in overtime hours per day are not the main driver behind our results. In a second stage, we point out the large heterogeneity in the ability of firms to adjust wages and we study the determinants of this heterogeneity in the firm-level wage rigidity. We find that larger firms, with a higher share of blue collar workers, which belong to sectors whose wage structure is characterized by a larger amount of bonuses display more flexible wages. 5 7 9 11 13 15 17 19 21 23 25 27 29 5 7 9 11 13 15 17 19 21 23 25 27 29 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Germany Italy Spain France (current prices) Hourly Wages Source: Own calculations based on Eurostat data, National Accounts. Fig. 1 Evolution of hourly wages over time Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 2 of 32 Finally, we show that more rigid firms reacted to the shock by increasing turnover more, but only if they were endowed with a flexible enough workforce, i.e. with a large share of temporary workers before the crisis. Presumably firms managed, even in the presence of high wage rigidity, to lower their average cost per employee by workers’replacement and wage renegotiation. 1 Several previous studies have documented the existence of wage rigidities before the Great Recession. An important example is Kahn (1997), who uses US data and estimates that employees would experience nominal wage reductions 47% more frequently, absent wage rigidities. Dickens et al. (2007) analyse rigidities in the USA and in 15 European countries and find that the fraction of workers subject to wage rigidity is 28% on average, with very large heterogeneity across countries (from 4% in Ireland to 58% in Portugal, Italy is in the middle of the distribution). By analysing wage rigidities during the recent downturn, we complement recent findings for the USA, the UK and Europe that find evidence of increased flexibility during the recent recession (Kurmann et al 2014; Brandolini and Rosolia 2015; Elsby et al. 2016; Verdugo 2016). We contribute to this literature by investigating more thoroughly which are the main determinants of wage rigidity/flexibility. In addition, our paper speaks to the literature on the relationship between wage flexibility and employment. The available literature is much scarcer in this case and the evidence is mixed: Card and Hyslop (1997) find that wage rigidities have small effects on the economy, Pischke (2016) and Ehrlich and Montes (2015) find instead that wage rigidities are associated with lower employment levels. We contribute to this debate by exploiting the richness of our data in terms of variables and time span to study how firms responded to negative shocks, depending on their degree of wage rigidity and their workforce composition. The structure of the paper is as follows: Section 2 briefly discusses the Italian institutional setting; Section 3 describes the datasets used for the analysis, and Section 4 studies the presence of wage rigidities for job stayers. Section 5 analyses separately the evolution of the nationally negotiated and the residual part of the wages. Section 6 estimates the relationship between wage adjustments, employment adjustments and the average wage per employee at the firm level. Finally, Section 7 concludes. 2 Institutional setting The evolution of wages in Italy is strictly linked to its institutional setting of labour relations. Since the income policy agreements of 1993, the Italian industrial relations are organized around two pillars. The first pillar is the national collective agreement (Contratto Collettivo Nazionale di Lavoro, CCNL), a sector-specific labour contract negotiated between the sector employers’association and the main trade unions. Its functions include the definition of the framework that disciplines the main aspects of labour utilization and the safeguard of real wage levels. The contracts are renewed every 3 years (since 2009, it used to be every 2 years before 2009) by the main social partners. Nominal increases of the base (minimum) wage are benchmarked to an independent 3year-ahead forecast of inflation net of imported energetic goods. The social parties agree on the level and the evolution of the base (minimum) wage to be applied to each occupation category (distinguishing various types of blue and white collar workers, as Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 3 of 32 well as middle managers, depending on the sector) over the 3-year horizon. The second pillar is constituted by firmor area-level labour contracts. Pay negotiations at the firm level are intended to account for firm-specific developments and local conditions, such as improved productivity or the risk of job loss. These are subordinate to the national one and can (i) modify items related to labour utilization, if the national contract allows them to, or regulate aspects not explicitly covered by the first level and (ii) provide for additional wage increases, which should both redistribute firm-level productivity gains (possibly achieved by the very same firm-level contract through more efficient organization) and align the labour cost to the ongoing labour market conditions. However, only in well-delimited cases of firm’s restructuring or crisis, second-level deals can (temporarily) cut wages below the nationally set sectoral minimum. Still, although legally possible, there is little evidence of firm-level agreement envisaging a decrease in the wage below these minima during the period of our analysis (D’Amuri et al. 2015). Furthermore, despite the introduction of fiscal incentives to promote firmand locallevel bargaining, these agreements are not very widespread (only 20% of firms with more than 20 employees in 2010) and are limited to larger firms and to specific sectors, respectively (D’Amuri et al. 2015). This mechanism of wage negotiation inevitably results in some degree of downward wage rigidity as employers are constrained by the nationally negotiated minima agreed upon by the main social parties every 3 years. In case of a negative shock in firms’performance, this predetermination of wages over the contractual horizon by itself reduces the possibilities of timely adjustments, at least for the centrally bargained component. However, wages may still be flexible thanks to the part that is not centrally negotiated, representing on average 20% of wages. 2,3 For what concerns the form of labour contracts, since 1997, the Italian labour market is characterized by the presence of both permanent and temporary contracts. However, although temporary contracts are becoming more widespread, more than 85% of all employees are still permanent ones. Depending upon the size of the firm (whether it is above 15 employees), there is also a relatively more stringent employment protection legislation regarding permanent workers (Cappellari et al. 2012). This somewhat limits the ability of firms to fire workers in case of a negative shock. Therefore, wage rigidities and employment protection legislation interact in the evolution of wage and employment adjustments over the business cycle. 3 Data The source for the data consists of social security payments made by legal entities to the Italian National Social Security Institute (INPS) for all employees with open-ended, fixed-term and apprenticeship contracts between 1990 and 2014. From this master data, INPS extracts two datasets. The first consists of the universe of firms with at least one employee at some point during a given calendar year—this extraction runs only up to 2013 and provides data at the firm level. The second consists of the employment histories of all workers born on the first or the ninth day of each month (24 dates). The firm extraction contains the fiscal code; information on the average number of employees over the year and the gross wage bill by occupational category—blue collar, white collar, middle and top manager; the two-digit sector code (NACE 2002) and the province code; and the date of entry and exit (if any). The worker extraction provides Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 4 of 32 information on demographics, the annual gross wage, the number of days worked, maternity and sick leave as well as short time work benefits (STWB, Cassa integrazione guadagni) but only for the period 2005–2014. We restrict attention to the nonagricultural business sector and workers aged between 20 and 64. We use the fiscal code as the definition of the firm. Tables 1 and 2 report the descriptive statistics for the sample of firms and workers respectively. Our sample of workers covers about 7% of the total workforce in the non-agricultural business sector while the firm data refer to the universe of firms with at least one employee (see Adamopoulou et al. 2016). 4 Rigidities of nominal wages This section studies the evolution over time of wage rigidities in Italy in the nonagricultural business sector between 1990 and 2014. In line with the existing literature on wage rigidities (for Italy Devicienti et al. 2007), we restrict the analysis to the sample of “super-stayers”in order to analyse wage changes net of composition effects: we keep only full-time workers aged 20–64, who have worked for two consecutive years (for at least 52 weeks and at least 200 days per year), in the same firm, with the same contract Table 1 Descriptive statistics, universe of firms paying contribution at INPS Year % of firms in industry % of firms in manufacturing Wage per employee Firm size Nfirms Nemployees (1000) Mean sd Mean sd 1990 0.49 0.32 1102 457 7.96 182.3 1,116,992 8891 1991 0.48 0.32 1217 495 7.96 181.0 1,120,621 8920 1992 0.48 0.31 1288 539 7.86 188.1 1,122,468 8823 1993 0.47 0.31 1334 556 7.80 184.2 1,084,614 8460 1994 0.47 0.31 1382 579 7.83 180.2 1,059,329 8295 1995 0.47 0.30 1441 620 7.87 179.1 1,063,816 8372 1996 0.47 0.30 1492 646 7.94 172.9 1,069,946 8495 1997 0.46 0.30 1550 670 7.96 163.1 1,058,116 8423 1998 0.46 0.29 1580 697 7.97 156.2 1,082,872 8630 1999 0.45 0.28 1595 711 7.86 138.3 1,136,162 8930 2000 0.44 0.27 1637 766 7.97 139.1 1,181,332 9415 2001 0.44 0.27 1675 821 7.98 140.1 1,222,383 9755 2002 0.44 0.26 1693 788 7.73 133.2 1,293,290 9997 2003 0.44 0.25 1728 819 7.70 130.0 1,325,115 10203 2004 0.43 0.24 1765 837 7.59 127.9 1,369,569 10395 2005 0.42 0.24 1816 892 7.56 128.7 1,380,837 10439 2006 0.42 0.23 1872 938 7.55 132.0 1,403,806 10599 2007 0.42 0.22 1898 994 7.53 133.5 1,474,110 11100 2008 0.41 0.22 1973 1030 7.57 129.0 1,496,808 11331 2009 0.40 0.22 1975 1006 7.48 146.9 1,478,586 11060 2010 0.39 0.21 2031 1055 7.43 169.6 1,471,068 10930 2011 0.38 0.21 2068 1070 7.46 165.1 1,467,732 10949 2012 0.37 0.21 2073 1086 7.35 167.6 1,468,611 10794 2013 0.36 0.21 2100 1139 7.44 169.1 1,414,664 10525 Source: own calculations on INPS data for the universe of firms. Statistics of wages are weighted by the number of employees in the firm Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 5 of 32 Table 2 Descriptive statistics on workers (at the contract level) Year Daily wage Age % Females % Full time workers % Temporary workers % Blue collars % White collars % Middle managersa% Middle managersa % Industry N employees N firmsbN firmsb mean s.d. mean s.d. 1990 47.28 43.96 35.96 10.99 0.29 0.96 0.65 0.31 0.63 682,208 307,372 1991 51.67 48.81 36.01 10.97 0.29 0.95 0.64 0.32 0.62 691,339 309,709 1992 55.59 99.31 36.19 10.92 0.29 0.95 0.64 0.32 0.62 690,862 310,198 1993 57.15 81.28 36.47 10.80 0.30 0.94 0.63 0.33 0.61 663,803 296,824 1994 60.43 151.11 36.44 10.70 0.31 0.93 0.63 0.33 0.60 655,417 293,078 1995 62.02 883.64 36.25 10.57 0.31 0.92 0.64 0.32 0.59 660,807 297,728 1996 61.11 53.90 36.29 10.52 0.32 0.91 0.64 0.31 0.02 0.58 672,283 302,918 1997 63.20 66.52 36.29 10.42 0.32 0.91 0.63 0.31 0.02 0.57 671,493 301,551 1998 65.46 263.59 36.31 10.42 0.32 0.90 0.12 0.62 0.31 0.02 0.57 683,133 305,565 1999 66.41 220.65 36.31 10.37 0.33 0.89 0.13 0.62 0.30 0.02 0.55 709,333 319,970 2000 66.96 182.25 36.34 10.34 0.33 0.89 0.14 0.62 0.30 0.02 0.53 755,548 340,240 2001 68.12 189.41 36.44 10.30 0.33 0.88 0.15 0.62 0.30 0.02 0.52 783,311 353,834 2002 69.07 142.84 36.49 10.27 0.33 0.87 0.15 0.63 0.29 0.02 0.51 819,908 375,084 2003 69.25 77.77 36.75 10.27 0.33 0.86 0.16 0.62 0.29 0.02 0.50 828,038 381,404 2004 71.30 112.71 36.99 10.24 0.34 0.85 0.17 0.62 0.29 0.02 0.50 836,470 388,024 2005 73.42 120.58 37.32 10.26 0.34 0.84 0.18 0.61 0.29 0.03 0.49 833,896 392,789 2006 74.90 93.32 37.61 10.29 0.34 0.83 0.20 0.61 0.29 0.03 0.47 847,912 399,949 2007 75.78 85.62 37.70 10.37 0.35 0.81 0.22 0.62 0.28 0.03 0.47 893,329 428,528 Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 6 of 32 Table 2 Descriptive statistics on workers (at the contract level) (Continued) 2008 79.31 93.75 38.04 10.44 0.35 0.80 0.22 0.61 0.29 0.03 0.45 920,759 435,389 2009 80.04 109.36 38.56 10.49 0.36 0.79 0.21 0.61 0.30 0.03 0.44 897,586 424,574 2010 81.84 209.37 38.82 10.55 0.36 0.78 0.23 0.61 0.30 0.03 0.43 892,964 423,298 2011 82.47 98.12 39.07 10.60 0.36 0.78 0.24 0.61 0.29 0.03 0.42 897,356 424,831 2012 83.09 96.54 39.44 10.67 0.37 0.76 0.24 0.62 0.30 0.03 0.41 887,777 420,333 2013 84.52 107.75 39.95 10.67 0.37 0.74 0.23 0.61 0.30 0.03 0.40 858,498 396,926 Source: own calculations on INPS data, data are summarized at the contract level and refer to all employees born on the first and ninth day of each month aData on middle managers and white collars are reported together before 1997 bNumber of firms where at least one worker in the sample transited in the considered year Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 7 of 32 (full time/part time and fixed term/open ended) and position (blue collar/white collar/ middle manager) and who were not under short time work benefits (CIG/fondo di solidarietà; only since 2005 when the information becomes available). 4 Excluding workers under short time work benefits is important when analysing the cyclicality of wage changes for stayers as these schemes are strongly anti-cyclical. Some other forms of measurement error in daily wage (level and/or change) remains inevitable, for instance, because of episodes of maternity or sick leave or because of adjustments in overtime hours. As long as this measurement error is not correlated with the business cycle, it should not pose a serious concern for our analysis that studies wage rigidity over time. We further show in a robustness exercise that changes in overtime hours that are likely to be correlated with the cycle do not affect our results. We construct the daily wage for each worker by dividing the total annual wage by the number of days worked during the year, and we calculate the percentage change in daily wages for the sample of workers defined above. 5,6 We exclude the outliers of the daily wage changes distribution (1st and 99th percentile). The upper panel of Fig. 2 plots the distributions of the annual change in daily wage for 2006 (before the first part of the recession) and 2009 (during the first part of the recession). Each plot includes a solid vertical line at 0 to denote the threshold for nominal wage rigidities, a dotted vertical line at the inflation rate to denote the possible threshold for real wage rigidities and a dashed one that is a proxy of aggregate productivity developments (annual percentage change of value added per worker). We observe that in 2006, the distribution is skewed to the right, implying that the mass of employees who experienced a wage change above the median is larger than the mass of those who experienced a wage change below the median. The peak of the distribution is around the inflation rate, and there is an “excessive”concentration between zero and the inflation rate. Moreover, the share of employees who received a daily wage cut in 2006 is significantly lower. Instead, in 2009, the distribution shifts to the left and becomes much more symmetric. The lower panel of Fig. 2 presents the distributions of the annual change in daily wage for 2010 (a period of slight recovery) and 2013 (during the second part of the recession). We observe that the mass of the distribution shifted to the right as soon as there was a slight recovery in 2010 but shifted again to the left in 2013. 4.1 The measures of wage rigidity In this paper, we adopt a flexible approach for the estimates of wage rigidities: the only assumption we make is that the distribution of the notional wage changes is symmetric around the median. 7 This implies that the only thing that distinguishes a scenario of rigid wages from a scenario of fully flexible wages is the presence of asymmetries in the distribution of wage changes. The assumption that the distribution of the notional wage changes is symmetric is in line with a large strand of the literature that adopts more parametric approaches (Dickens et al. 2007; Goette et al. 2007; Devicienti et al. 2007; Card and Hyslop 1997). Our methodology, adopted also by Verdugo (2016) to study wage cyclicality in Europe, differs because it does not make any parametric assumption on the mean and variance of the notional wage changes distribution. We just assume that, absent wage rigidities, the bottom half of the distribution would mirror the upper Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 8 of 32 look at wage changes instead of wage levels, since the agreed percentage changes are usually similar across broad positions. Figures 11 and 12 show the evolution of the part of the wage that excludes the nationally negotiated minima in each sector by the employees’position (black line). Among metalworkers, we observe a large drop in 2009, i.e., at the onset of the crisis, for blue collars and white collars while middle managers were only mildly affected. In 2014, there is another drop, but only for blue collars. By contrast, those that suffered mostly in 2009 among the employees in the wholesale and retail trade were the middle managers. Blue collars and white collars experienced instead a large drop in 2014. These different reactions in the residual part of wages may reflect sectorial differences in the structure of wages (bonuses) and/or overtime hours. The nonnegotiated part of wages is larger for large firms, for workers in the metalwork industry and for middle managers. We then repeat the exercise of the previous section and calculate measures of the skewness using the part of the wage that is not negotiated at the national level rather than the total wage. As expected, the residual part of the wage is on average, more -0.10 -0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 1-9 employees 10-49 employees 50-49 employees 250 employees or more Source: own calculations on INPS data (on the sample of job sta y ers). Fig. 9 Wage rigidity by firm size, Kelley’s skewness 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 1-9 employees 10-49 employees 50-49 employees 250 employees or more Source: own calculations on INPS data (on the sample of job stayers). Fig. 10 Negative wage changes by firm size, % employees Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 15 of 32 flexible (the level of skewness associated with it is lower and the share of wage cuts is higher). In particular, in 2006, the Kelley’s skewness in the wholesale and retail trade sector was 0.07 for the residual part and 0.31 for the overall wage (0.02 and 0.14 for the metalwork industry). Moreover, in line with our previous graphs, we observe that the skewness of the residual part of wages decreases, even towards negative values, in 2009, especially among blue collars in the metalworkers contract and among middle managers in the wholesale and retail trade sector. The information on the negotiated and residual part of the wage also allows us to investigate whether an increase in the negotiated wage induces firms to adjust the other component of the wage. The red line in Figs. 11 and 12 represents the negotiated part of the wages. Note that due to the institutional setting, the evolution of the negotiated part of the wages is sticky as it is usually predetermined for 3 years. As a result the negotiated part of the wages continued to rise even during years of economic downturn 49 54 59 64 69 74 79 84 89 94 99 104 109 114 119 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Residual Part of Wages Negotiated Wages Middle Manager 21 26 31 36 41 46 51 56 61 66 71 76 81 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Residual Part of Wages Negotiated Wages White Collar 8 13 18 23 28 33 38 43 48 53 58 63 68 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Residual Part of Wages Negotiated Wages Blue Collar 0.10 0.20 0.30 0.40 0.50 0.60 0.70 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Blue Collar White Collar Middle Manager Residual Part of Wages over Total (%) Daily wages: Metal Workers Source: own calculations on INPS data (on the sample of job stayers). Fig. 11 Changes in the nationally negotiated and residual part of wages over time, by workers’ position. Metalworkers Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 16 of 32 (grey area). 16 However, a simple graphical inspection of the evolution of the negotiated (red line) and residual (black line) part of the wages shows that in 2009 and in 2014, firms compensated for the increase of the negotiated part of the wages by reducing the residual component. 17 6 Wage rigidities, employment adjustments and firms’average employee compensation Wage rigidities may induce firms to perform adjustments along the employment margin as an alternative way to react to shocks. Devicienti et al. (2007) find that in the 1990s, firms with higher downward rigidities in Italy tended to display higher worker reallocation rates in terms of turnover. In this section, we study whether firms that are historically characterized by high wage rigidities adjusted employment more during the recent recession and whether in this way they managed to contain their average wage per employee. 58 63 68 73 78 83 88 93 98 103 108 113 118 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Residual Part of Wages Negotiated Wages Middle Manager 3 8 13 18 23 28 33 38 43 48 53 58 63 68 73 78 83 88 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Residual Part of Wages Negotiated Wages White Collar 3 8 13 18 23 28 33 38 43 48 53 58 63 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Residual Part of Wages Negotiated Wages Blue Collar -0.00 0.10 0.20 0.30 0.40 0.50 0.60 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Blue Collar White Collar Middle Manager Residual Part of Wages over Total (%) Daily wages: Wholesale and Trade Source: own calculations on INPS data (on the sample of j ob sta y ers) Fig. 12 Changes in the nationally negotiated and residual part of wages over time, by workers’position. Workers in the wholesale and trade sector Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 17 of 32 To do so, we make use of an alternative dataset that covers the universe of employees for a sample of firms with more than 20 employees. Observing the universe of employees in each firm is crucial in order to define an accurate measure of wage rigidities at the firm level. These data come from INPS as well, and refer to the universe of employees in the firms that belong to the Bank of Italy’s yearly survey on industrial and non-financial service firms (INVIND), and consist of around 4000 firms per year. Another advantage of using this dataset of employees is that we can enrich it with extra variables at the firm level that come directly from the INVIND survey. For example, INVIND allows us to separately observe accessions and separations that could only indirectly and partially be proxied by monthly employment changes in the INPS firm data. Moreover, INVIND provides us with additional information, like total sales (in euros) that can serve as a demand-shift control. There is also information on per capita overtime hours that can help us verify that the wage adjustment is not operated through this channel only. 18 6.1 Wage rigidity: firms’heterogeneity We now dig into the determinants of wage rigidity. In particular, we study how the degree of wage rigidity differs among firms and which characteristics of the firm determine such heterogeneity. In order to perform this analysis, we need to compute a measure of wage rigidity at the firm level. In particular, we aim at constructing a measure that summarizes the firms’ability to adjust wages of incumbents when hit by a shock, for example, because of higher bonuses or higher bargaining power. Similarly to what we did in Section 4, we base our measure on the asymmetry of the distribution of wage changes before the crisis. For each firm belonging to the INVIND sample, we compute the Kelley’s skewness of the distribution of yearly wage changes for job stayers and we average across the years between 2003 and 2008. This measure relies on the same assumptions described in Section 4. Positive values of the skewness are associated with the presence of wage rigidities, while values close to 0 point towards wage flexibility. We exclude all firms with few job stayers (less than 80 in the period 2003–2008) because the skewness measure would not be reliable if the number of observations in the distribution of wage changes is too small. 19 Figure 13 shows how our firm-level measure of wage rigidity is distributed. The mean is larger than 0 (it is 0.13), implying that it exists some form of wage rigidity on average. Moreover, there is a high level of heterogeneity across firms (the standard deviation is 0.19). We consider three main factors behind differences in wage rigidity across firms: the presence of a large component of bonuses; the extensive use of overtime hours, which are easier to adjust during downturns; and the actual ability to adjust the base salary of job stayers, due to high bargaining power, for instance. Figure 14 shows how our measure of rigidity is correlated with the amount of bonuses (over annual earnings) at the sector level. The figure displays on the x-axis the share of bonuses over earnings as obtained from the Structure of Earnings Survey in 2006 and on the y-axis our measure of wage rigidity averaged at the sector level. There is a negative correlation: those sectors whose wage structure is on average characterized by a high share of bonuses are the ones that according to our measure of wage rigidity are more flexible. Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 18 of 32 Table 3 shows the determinants of our measure of wage rigidity. Column 1 shows that larger firms, whose average wages are less often flattened upon the national contractual wages, tend to be less rigid, which is in line with the results of Adamopoulou et al. (2016) and with those in Section 4. Again in line with the results of our previous section, column 2 shows that firms are less rigid if they are characterized by a high share of blue collars. Column 3 shows instead that our measure of wage rigidity is not related with the amount of overtime hours employed by each firm. This suggests that wages are not flexible simply because of the firms’ abilitytoadjustovertimehours.Column4showsthatthereisapositiveassociation between firms’productivity and firms’rigidity.Thisresultisconsistentwith the theory of efficiency wages (Akerlof 1982; Stiglitz 1986; Campbell and Kamlani 1997). Firms may decide not to cut wages because wage cuts reduce workers’effort and overall productivity. Productivity would therefore be positively associated with higher level of rigidity both because workers not experiencing a wage cut are more productive and because more productive, and profitable, firms may have enough 0.02 .04 .06 .08 Fraction -1 -.5 0 .5 1 w rigidity 03-08 Note: own calculations from INPS data, INVIND sample. Mean = 0.13, standard deviation = 0.17. Fig. 13 Distribution of firm level wage rigidity (Kelley’s skewness, 2003–2008) Mining and quarrying Manufacturing Electricity Construction Wholesales and retail trade Hotels and restaurants Transport Financial intermediation Real estate .1 .12 .14 .16 .18 Wage rigidity 2003-2008 .05 .1 .15 .2 Quota of bonuses on annual pay Note: quota of annual bonuses on annual earnings obtained from the Structure of Earnings survey, 2006. The fitting line is computed excluding the construction sector. Fig. 14 Wage rigidity and share of bonuses at the sector level Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 19 of 32 Table 3 Determinants of firm level wage rigidity Dep. var: Skewness 2003–2008 size cat = 2 −0.019** (0.08) −0.016* (0.009) −0.016* (0.009) size cat = 3 −0.021* (0.012) −0.023* (0.013) −0.026* (0.013) % blue coll = 2 −0.026*** (0.007) −0.018** (0.009) −0.016 (0.010) % blue coll = 3 −0.046*** (0.007) −0.045*** (0.010) −0.042*** (0.010) overtime pw = 2 0.002 (0.007) 0.000 (0.009) 0.001 (0.009) overtime pw = 3 −0.006 (0.007) −0.016* (0.009) −0.018* (0.009) va pw = 2 0.015** (0.007) 0.018** (0.009) 0.022** (0.010) va pw = 3 0.017** (0.007) 0.022** (0.010) 0.029*** (0.010) % temporary = 2 0.001 (0.007) 0.003 (0.009) 0.002 (0.009) % temporary = 3 0.016** (0.007) 0.024*** (0.009) 0.019** (0.009) Observations 2170 3650 3650 3519 3650 2103 2101 Sector FE No No No No No No Yes Note: *p < 0.1; **p < 0.05; ***p < 0.01. All controls are the average of the considered characteristic between 2003 and 2008. Categories refer to tertiles of the distribution (3 = highest tertile). size cat refers to the number of employees, % blue coll is the share of blue collar over all employees, overtime pw is the average number of overtime hours per worker, va pw is the value added per worker and % temporary is the share of temporary workers. Robust standard errors in parenthesis Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 20 of 32 resources to decide not to cut wages. Finally, column 5 shows that firms with a high share of temporary workers tend to be more rigid: we interpret this as a sign that rigid firms, envisaging their difficulty in adjusting wages of job stayers, react by hiring a large share of temporary workers, easier to fire. Lastly, we conducted a robustness check in order to verify that observing daily instead of hourly wages does not undermine our results of Section 4. We use the information provided by the INVIND survey on the amount of overtime hours per capita in 2008 (i.e. before the crisis) at the firm level and we classify firms into more and less overtime-intensive (above and below the median). We then compute the skewness of the distribution of annual changes of daily wages for all workers belonging to each group of firms for the years 2006 and 2009. We find that for both groups, the skewness was almost identical in 2006 and it declined sharply in 2009 by almost the same amount (Fig. 15). Therefore, we conclude that overtime hours do not seem to drive the observed increase in wage flexibility in 2009. In the next section, we study how firms characterized by different levels of wage rigidity respond to the recent recession. 6.2 Firms’wage rigidity and employment adjustments We perform regressions at the firm level to evaluate the relationship between the measure of wage rigidity described above and firms’adjustments in the employment margin and in their average wage per employee during the recent recession. We estimate the following equation: Δyi;t−08 ¼αþβskewnessiþγXiþεið1Þ where y i,t−08 are flows of the outcome variables during the recent recession (between 2008 and t= 2009, 2011 and 2013); in particular, we look at: turnover accessionstþseparationst employment2008  ; accessions accessionst employment2008  ;separations separationst employment2008  20 ; the increase in the average wage per year per employee av waget av wage2008 −1  and the probability of exiting the market. skewness i is the Kelley’s skewness measure that refers to the 6 years before the recession, as described in Section 6.1. Since we want in principle to isolate the effect of wage rigidity from the effect of other factors such as the firm size, that may be associated with both a more flexible wage structure and different dynamics of our dependent variables, we control in all specifications for firms’value added per worker, firms’age, size, level of overtime hours per employee, share of temporary workers, province and sector of activity. All controls refer to the year 2008, i.e. right before the arrival of the crisis, in order to exclude endogenous changes of these variables correlated with the effect of interest. Moreover, in some specifications, we include the variation of total sales during the recession as additional control, in order to correct for the heterogeneity of the shocks across firms. Standard errors are clustered at the firm level. 21 Therefore, the βcoefficient of our regressions measures the relationship between wage rigidity at the firm level and employment adjustments, net of the effect of other potential (observable) factors that may spur our estimate. 22 Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 21 of 32 Table 4 displays the main descriptive statistics. Firms in the INVIND sample mostly belong to the industrial sector and are of rather large dimension. Temporary workers represent on average 7% of their workforce. Average turnover in 2009 was 20% with respect to the number of employees in 2008 and separations were slightly larger than accessions. Table 5 presents our main results on turnover. In line with Devicienti et al. (2007), we find that firms with higher wage rigidities are characterized by higher turnover in 2009 (column 1). In particular, an increase in the firm-level skewness by one standard deviation (0.16) is associated with a turnover rate in 2009 of almost 0.7 percentage points higher. One may worry about possible omitted variables in this regression. In particular, it may be that more rigid firms may be hit Firms with low level of overtime hours per capita in 2008 skewness=0.5 kewness=0.19 Firms with high level of overtime hours per capita in 2008 skewness=0.5 s3 s4 kewness=0.12 0 .01 .02 .03 .04 .05 .06 .07 -.3 -.2 -.1 0.1 .2 .3 .4 2006 0 .01 .02 .03 .04 .05 .06 .07 -.3 -.2 -.1 0.1 .2 .3 .4 2009 0 .01 .02 .03 .04 .05 .06 .07 -.3 -.2 -.1 0.1 .2 .3 .4 2006 0 .01 .02 .03 .04 .05 .06 .07 -.3 -.2 -.1 0.1 .2 .3 .4 2009 Source: own calculations on INVIND data (on the sample of job stayers). Fig. 15 Distribution of percentage annual change of daily wages, by firms’level of overtime hours per capita in 2008 Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 22 of 32 differently by the crisis. Column 2 shows that the results are robust to the inclusion of the percentage change of sales, our proxy for the size of the firm-level shock. However, this control of the demand shock is potentially endogenous, as it refers to 2009 and may be considered as an outcome itself. We therefore exclude it from our preferred specification. We also examine whether the effect of wage rigidities persists over time by considering turnover in 2011 and in 2013 (computed as the cumulative rate of accessions and separations with respect to 2008). Indeed, the effect is still present and increasing both in 2011 and 2013. This may be reconciled with a delayed employment adjustment for firms that start exhausting any margin of wage adjustment. At this point, it is crucial to understand the nature of this increased turnover. In particular, it is interesting to explore whether it is mainly driven by separations or accessions and which may be the channel behind it. One possible mechanism acts through the increased turnover of temporary workers (see Bulligan and Viviano, 2016). Firms that are constrained by wage rigidities may exploit the turnover of temporary workers either to renegotiate their wage or to hire a different worker, whose wage will reflect the new cyclical conditions. In this way, firms can exploit flexibility on the employment margin in order to adjust their average labour cost. Table 4 Descriptive statistics of INVIND firms: mean and standard deviation Characteristics Mean s.d. Skewness 2003-2008 0.138 0.161 Turnover 2009 0.194 0.258 % Accessions 2009 0.087 0.142 % Separations 2009 0.107 0.132 Turnover 2011 0.579 0.751 % Accessions 2011 0.276 0.391 % Separations 2011 0.300 0.365 Turnover 2013 0.905 1.198 % Accessions 2013 0.439 0.629 % Separations 2013 0.466 0.585 %Δ(firm’s average wage) 2009 −0.010 0.080 %Δ(firm’s average wage) 2011 0.061 0.089 %Δ(firm’s average wage) 2013 0.092 0.119 % Temporary employees 2008 0.068 0.115 Firm size 2008 468 2152 % Industry 0.760 – Firm age 2008 25.89 13.15 Value added per worker 2008 71.09 74.28 Overtime hours per worker 2008 5336 3062 %Δ(sales) 2009 0.068 0.115 Source: The skewness is taken from INPS data on the population of workers belonging to firms in the INVIND sample. Turnover, accessions, separations, share of temporary workers, overtime hours per worker and sales from the INVIND survey. Average wage, firm size, age and sector from INPS data on the population of firms. Value added per worker from CERVED. Turnover 2011, accessions 2011 and separations 2011 are cumulative for the years 2009–2011, turnover 2013, accessions 2013 and separations 2013 are cumulative for the years 2009–2013. All variables in %Δare defined with 2008 as base year Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 23 of 32 Table 5 Relationship between turnover and wage rigidity controlling for demand shocks Turnover 2009 Turnover 2011 Turnover 2013 (1) (2) (3) (4) (5) (6) Skewness 2003-2008 0.116* (0.0651) 0.107 (0.0651) 0.444** (0.201) 0.425** (0.201) 0.711* (0.364) 0.649* (0.365) Overtime hours per wrk 2008 0.0196 (0.0215) 0.0202 (0.0214) 0.121* (0.0719) 0.119* (0.0707) 0.283*** (0.0860) 0.290*** (0.0830) Value added per wrk 2008 0.000313 (0.000412) 0.000316 (0.000412) 0.000112 (0.00121) 4.59e−05 (0.00122) −0.00211* (0.00119) −0.00241** (0.00117) Share temporary wrk 2008 1.003*** (0.286) 0.994*** (0.287) 3.057*** (0.883) 3.016*** (0.888) 4.176*** (0.929) 4.098*** (0.916) %Δ(sales) 2009 0.0923*** (0.0348) 0.356*** (0.115) 0.723*** (0.209) Firms’size, sector and province dummies, age Yes Yes Yes Yes Yes Yes Adj. R 2 0.373 0.376 0.425 0.431 0.492 0.503 N1792 1792 1321 1321 1019 1019 Source: Turnover 2011 and Turnover 2013 are cumulative for the years 2009–2011 and 2009–2013. All columns include controls for: firms’value added per worker, firms’age, size, level of overtime hours per employee, share of temporary workers, province and sector of activity, all referring to 2008. Turnover, overtime hours, share of temporary workers and sales are taken from the INVIND survey. Skewness from INPS data on the population of workers belonging to firms in the INVIND sample. Firm’s size, age, sector and province from INPS data on the population of firms. Value added per worker from CERVED. Robust standard errors in parenthesis *p< 0.1; **p< 0.05; ***p< 0.01 Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 24 of 32 possible to cut wage further and not because of their type of labour contract (and associated firing costs). We find however that, even controlling for the average wage level in 2008 (which proxies the share of workers close to the contractual minimum threshold), the turnover is much higher in firms characterized by a higher share of temporary workers in 2008 (results available upon request). 24 Firing is very difficult for firms with more than 15 employees in Italy. 25 Additional estimates (available upon request) indicate that more rigid firms have a higher probability of exiting the market in 2013. This would entail that the samples of more rigid firms with and without temporary workers are not fully comparable in 2013. Acknowledgements Many thanks to an anonymous referee and the editor, Matteo Bugamelli, Francesco D’Amuri, Raffaella Nizzi, Alfonso Rosolia, Paolo Sestito, Eliana Viviano and seminar participants at the Bank of Italy for useful comments and help. The views expressed in this paper are those of the authors and do not necessarily reflect those of the Bank of Italy. Responsible editor: Juan Jimeno Competing interests The IZA Journal of Labor Policy is committed to the IZA Guiding Principles of Research Integrity. The authors declares that they have observed these principles. Received: 22 June 2016 Accepted: 24 November 2016 References Adamopoulou E, Bobbio E, De Philippis M, Giorgi F (2016) Allocative efficiency and aggregate wage dynamics in Italy, 1990-2013. In: Bank of Italy Occasional Paper., p 340 Akerlof GA (1982) Labor contracts as partial gift exchange. Q J Econ 97:543–569 Brandolini A and Rosolia A (2015) The Euro area wage distribution over the crisis. Banca d’Italia, Rome Mimeo Bulligan G, Viviano E (2016) Has the wage Phillips curve changed in the Euro area? Evidence from four countries. In: Banca d’Italia occasional working paper., p 355 Campbell C, Kamlani K (1997) The reasons for wage rigidity: evidence from a survey of firms. Q J Econ 112:759–789 Cappellari L, Dell’Aringa C, Leonardi M (2012) Temporary employment, job flows and productivity: a tale of two reforms. Econ J 122:188–215 Card D and Hyslop D (1997) Does inflation “grease the wheels of the labor market?”In: Romer C and Romer D (eds) Reducing Inflation: Motivation and Strategy, vol 30. National Bureau of Economic Research Studies in Business Cycles, Cambridge D’Amuri F (2014) Composition effects and wage dynamics in Italy. Banca d’Italia, RomeMimeo Daly M, Hobijn B, Wiles T (2011) Dissecting aggregate real wage fluctuations: individual wage growth and the composition effect. In: Federal Reserve Bank of San Francisco., p 23 D'Amuri F, Fabiani S, Sabbatini R, Tartaglia Polcini R, Venditti F, Viviano E, Zizza R (2015) Wages and prices in Italy during the crises: the firms’perspective. In: Bank of Italy Occasional Papers., p 289 Devicienti F, Maida A, Sestito P (2007) Downward wage rigidity in Italy: micro-based measures and implications. Econ J 117:F530–F552 Dickens W, Goette L, Groshen E, Holden S, Messina J, Schweitzer M, Turunen J, Ward M (2007) How wages change: micro evidence from the International Wage Flexibility Project. J Econ Perspect 21(2):195–214 Ehrlich G and Montes J (2015) Wage rigidity and employment outcomes: evidence from German administrative data. University of Michigan, Ann ArborMimeo Elsby M, Shin D, Solon G (2016). Wage adjustment in the great recession and other downturns: evidence from the United States and Great Britain. J Labor Econ. Forthcoming. http://www.journals.uchicago.edu/doi/abs/10.1086/ 682407. Eurostat (2002). "Statistical Classification of Economic Activities in the European Community, Rev.1.1 (2002) (NACE Rev. 1.1)", Luxembourg. Gertler M, Trigari A (2009) Unemployment fluctuations with staggered Nash wage bargaining. J Polit Econ 117:38–86 Goette L, Sunde U, Bauer TK (2007) Wage rigidity: measurement, causes and consequences. Econ J 117:469–477 Guvenen F, Ozkan S, Song J (2014) The nature of countercyclical income risk. J Polit Econ 122:621–660 Hibbs DA Jr, Locking H (1996) Wage compression, wage drift and wage inflation in Sweden. Labour Econ 3:109–141 Holden S (1998) Wage drift and the relevance of centralised wage setting. Scand J Econ 100:711–731 Holden S, Wulfsberg F (2008) Downward nominal wage rigidity in the OECD. B E J Macroecon (Advances) 8(1):15 Kahn S (1997) Evidence of nominal wage stickiness from microdata. Am Econ Rev 87:993–1008 Knoppik C, Beissinger T (2009) Downward nominal wage rigidity in Europe: an analysis of European micro data from the ECHP 1994-2001. Empir Econ 36:321–338 Kurmann A, McEntarfer E, Spletzer J (2014). The nature of wage adjustment in U.S. firms: new evidence from workerfirm linked data. Drexel University, Philadelphia Mimeo Lemieux T (2006) Increasing residual wage inequality: composition effects, noisy data, or rising demand for skill? Am Econ Rev 96:461–498 Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 31 of 32 Manacorda M (2004) Can the scala mobile explain the fall and rise of earnings inequality in Italy? A semiparametric analysis, 1977–1993. J Labor Econ 22:585–613 Pischke JS (2016) Wage flexibility and employment fluctuations: evidence from the housing sector. In: NBER Working Paper., p 22496 Postel-Vinay F, Robin JM (2004) To match or not to match? Optimal wage policy with endogenous worker search intensity. Rev Econ Dyn 7:297–331 Stiglitz JE (1986) Theories of wage rigidity. In: Butkiewicz J, Koford K, Miller J (eds) Keynes’Economic Legacy: Contemporary Economic theories. Praeger Publishers, New York, pp 153–206 Verdugo G (2016) Real wage cyclicality in the Euro zone before and during the Great Recession: evidence from micro data. Eur Econ Rev 82:46–69 Adamopoulou et al. IZA Journal of Labor Policy (2016) 5:22 Page 32 of 32