1 Introduction

The relationships between labour, technology, productivity and wages have usually been investigated in separate ways, without an integrated approach. Different streams of literature have examined the drivers of innovation, economic performance and the determinants of the quantity and quality of employment as well as wages, focusing on the specificity of the issues while neglecting the broader interactions between these economic processes.

This article aims to propose a virtuous circle model where positive relationships in each of these dimensions feed on each other, leading to a cumulative process of change and—possibly—to the good jobs-high innovation trajectory. To investigate these relationships in a more systematic way, considering not only one-way relationships but their interdependencies as well, we develop and test a simultaneous equation model at the level of industries.

The good jobs-high innovation virtuous circle is illustrated in Fig. 1. First, high employment quality, complementary to R&D efforts, contributes to greater product innovation—the outcome of a strategy of technological competitiveness. Second, product innovation is a key driver of labour productivity growth, alongside improved capital and labour inputs. Third, productivity gains are translated into higher wages and higher profits, distributing the benefits of growth. Higher wages, in turn, lead to improvements in the quality of jobs—with higher skills and more widespread standard employment relationships—and stimulate greater innovation efforts. In this way, the virtuous circle brings the economy to a higher growth trajectory that benefits all economic actors. In developing this model, we have to consider the high level of heterogeneity across economic activities; innovation, labour markets and productivity are crucially affected by structural factors and the diversity of sectors. Therefore, an appropriate focus for this investigation is the industry level; however, these processes operate also—with even greater heterogeneity—at the firm level, and further studies may address the presence of virtuous circles in firms.

Fig. 1
figure 1

Source: Authors’ illustration

The virtuous circle between high-quality labour, technological competitiveness, productivity and wages.

Full size image

In developing this approach, we connect different streams of research and build a more integrated perspective.

Non-standard jobs and innovation. Studies on industries and firms have investigated the impact of non-standard jobs (e.g. temporary, part-time, agency work), employment protection and job turnover on different measurements of innovation and economic performance, recognising that labour market flexibility is generally negatively associated with innovation outcomes (Cetrulo et al., 2019; Kleinknecht et al., 2014; Michie & Sheehan, 2003; Reljic et al., 2021) and productivity (Lisi & Malo, 2017; Lucidi & Kleinknecht, 2010; Ortega & Marchante, 2010).

Efforts to explore the job quality-innovation nexus in a more integrated way have been made by Duhautois et al. (2018), who found a positive association between technological innovation and job quality at the level of countries, industries and individuals; further factors relevant in explaining job quality include education, type of occupation and the presence of employee representation.

Technology and productivity. A large body of literature has explored the impact of technological activities—R&D, innovation and patents—on various measurements of productivity—from total factor productivity to labour productivity (see Hall, 2011 and Ugur & Vivarelli, 2021 for literature reviews)—identifying a significant contribution of innovation to improved economic performance. However, the diversity of strategies behind technological efforts has often been ignored, in the case of both R&D (Sterlacchini & Venturini, 2013; Syverson, 2011) and innovation; Pianta (2001, 2018) shows that two main innovation strategies can be identified with distinct effects on productivity—the search for technological competitiveness through new products and services, as opposed to cost competitiveness through the adoption of labour-saving technologies.

A useful contribution has come from Crépon et al. (1998), who developed a structural model where R&D efforts lead to innovation and innovation translates into productivity growth. This model has been widely used at the firm and industry level in the innovation literature (see Mohnen & Hall, 2013 for a review). However, the model adopts a static approach, which led to many cross-sectional investigations of firms and industries disregarding the presence of lags and feedback effects.

A dynamic approach has been proposed by Bogliacino and Pianta (2013), who investigated empirically the existence of a virtuous circle of technological progress using a simultaneous model of three equations at the industry level. Their results show that lagged R&D investments translate into successful product innovations that lead to higher profits, which are in turn reinvested into further R&D efforts, suggesting a feedback effect from retained earnings to new products and services. Bogliacino et al. (2017) tested the same model on Italian firms and found that the virtuous circle can be identified for the small group of persistent innovating firms alone. Exploring a panel of French and Dutch manufacturing firms, Raymond et al. (2015) found that innovation contributes to productivity; however, their results yield no signs of feedback effects, suggesting that more productive firms are not necessarily more successful innovators.

Productivity and wages. Different views exist on the productivity-wages nexus. Standard economic theory states that firms hire workers until the real wage equals the marginal product of labour. In other words, wage increases are driven by rises in marginal productivity. Conversely, the efficiency wage theory suggests that higher wages stimulate greater productivity (Akerlof & Yellen, 1990; Shapiro and Stiglitz, 1984). In addition, evolutionary perspectives point out that higher industry-wide wages might spur technological change as firms have to innovate to compensate for labour costs, while non-innovators exit the market (Nelson & Winter, 1982). Building on these insights, we explore the link between productivity and wages and the presence of a lagged ‘wage push’ effect on labour productivity.

Against this background, we aim to investigate these relationships in a more integrated way, taking into account the interdependencies among variables. Therefore, we develop and test a simultaneous equation model that links four key variables: employment quality, technological competitiveness, labour productivity and wages. The four equations of the model are presented in the next section. The empirical test is carried out at the industry level for 41 manufacturing and service sectors in six European economies—Germany, Spain, France, Italy, the Netherlands and the UK—over the 1994–2016 period, using the new version of the Sectoral Innovation Database with the NACE Rev.2 classification (Pianta et al., 2021).

Our approach offers several novelties. First, we consider the quality of jobs as a determinant of innovation performance, pointing out the complementarity between labour competences, R&D and other innovation inputs. Building on evolutionary perspectives, we expect that the diffusion of non-standard forms of employment—and the high labour turnover it entails—disrupts the accumulation of knowledge required for successful innovations, leading to a loss of the tacit knowledge ‘embodied’ in workers (Nelson & Winter, 1982). This role of job quality has so far received relatively little attention by the innovation literature.

Second, we emphasise the role of technological competitiveness, based on product innovation, as a driver of productivity growth complementary to the role played by improvements in capital and labour—proxied by fixed investment and lagged wages.

Third, we consider the distribution of the benefits of growth and the role of wages in these relationships. Higher wages contribute to a reduction in the share of non-standard jobs and an increase in labour productivity through the efficiency wage channel. Again, wage dynamics have so far received little attention by the innovation literature.

Finally, the combination of these four relationships in a simultaneous model, including lags and feedback effects, delivers an accurate representation of the good jobs-high innovation virtuous circle that goes beyond the linear, one-directional links typically explored in the literature.

The paper is organised as follows. Section 2 deals with the model specification; Sect. 3 presents data and methodology; results are discussed in Sect. 4; Sect. 5 concludes by summarising the main findings and policy implications. The Appendix provides the list of sectors we investigate; the time structure of the database organised in six periods; the description of the variables used; results of robustness checks of the simultaneous model.

2 Four explorations

The four relationships we investigate are presented here in the context of the relevant research streams, considering the key determinants and defining the model that is empirically tested. In Sect. 4, we first test each equation separately, assessing the robustness of results by including a broader range of control variables; we then carry out the simultaneous estimation with a simplified list of variables, showing that the key relationships—with lags, feedback loops and two-way links—are confirmed.

2.1 Non-standard work

The first equation in our model identifies the determinants of non-standard employment. The latter is proxied by the share of employees without a permanent full-time job—having either a full-time, fixed-term full-time job or part-time employment (Reljic et al., 2021). By considering different forms of non-standard contractual arrangements, this variable represents a relevant indicator of the employment quality in industries, capturing the disruption of workers’ learning and capabilities.

The fragmentation of labour markets and the rise of non-standard work have received considerable academic and policy interest. First, atypical forms of employment accounted for more than half of total employment growth over the last two decades in Europe (OECD, 2015). Second, they contributed to rising income inequalities associated with the gaps between standard and non-standard jobs in terms of wages, working conditions, career advancement opportunities and welfare protection (Kalleberg, 2011). Third, many advanced economies that introduced reforms aimed at liberalising labour markets—increasing ‘flexibility’ and non-standard jobs—have later experienced a slowdown in productivity growth; recent research has now shown that such policies have indeed contributed to lower economic performances (Lisi & Malo, 2017; Lucidi & Kleinknecht, 2010; Ortega & Marchante, 2010).

We regress industry-level non-standard employment on a set of institutional, economic and labour factors. Industrial relations studies have documented the crucial role of institutional settings—including unionisation and the bargaining power of workers—in shaping the spread of non-standard jobs (Hipp et al., 2015). Thus, we consider the declining unionisation rate and expect that a greater union representation is associated with a lower share of non-standard workers.

The main factors characterising the individual’s probability of being in involuntary non-standard employment are demographic, occupational and national. Green and Livanos (2017) have shown that in Europe the share is lower in the UK and Nordic countries and higher in Mediterranean countries, while non-standard jobs are disproportionately higher in some demographic groups.Footnote 1 Occupational categories also matter; their findings suggest that the share of non-standard jobs is much higher in elementary occupations, while managers, professionals and associate professionals report lower shares. In line with this, we control for the skill composition of the industry’s workforce. Our main proxy is the educational attainment—the share of university graduates and of workers with secondary education or less; alternatively, we use the top and bottom occupational groups—the share of managers, professionals and technicians and of manual workers. We expect industries with a higher share of skilled workers—university graduates or managers—to exhibit a lower presence of non-standard workers (OECD, 2015).

Recent literature has explored the impact of non-standard forms of work on innovation and productivity (Reljic et al., 2021), but much less is known about the reverse causality. Grande et al. (2020) found a positive impact of innovation on the composite indicator of job quality that accounts for its different dimensions (pay, intrinsic job quality, employment quality and workplace risks). Malgarini et al. (2013), using data on Italian manufacturing firms, suggested that the impact of innovation on firms’ demand for temporary workers depends on the phase of the business cycle; innovating firms hire more on a permanent basis in upswings, while in downswings they rely more on temporary contracts. We therefore consider the relationship that (lagged) productivity has with employment quality in industries.

Finally, wage levels are introduced; the variable we use is the percentage difference between the average wage in an industry and the average pay in the country’s sector with the highest wages; this measurement of relative distance locates the industry in the national labour market and we expect that the share of non-standard workers will be higher in low-paying sectors, that is in industries with a greater distance from top wages.

Considering the differences in the incidence of non-standard forms of employment across countries, we control for distinct (time-invariant) institutional settings with country fixed-effect. Moreover, we also include a set of dummies for time periods, industry effects and manufacturing.

Formally, the non-standard work equation can be written as follows:

$${\text{QNSW}}_{{\text{i,j,t}}} = \alpha_{0} + \alpha_{{1}} {\text{Union}}_{{\text{i,j,t}} - 1} + \alpha_{{2}} {\text{Skills}}_{{\text{i,j,t}} - 1} + \alpha_{{3}} \Delta {\text{LabProd}}_{{\text{i,j,t}} - 1} + \alpha_{{4}} {\text{WageDist}}_{{\text{i,j,t}} - 1} + \mu_{{\text{i}}} + \chi_{{\text{j}}} + \tau_{{\text{t}}} + \varepsilon_{{\text{i,j,t}}}$$
(1)

where i, j and t are indices for industry, country and time periods, respectively; \({\mu }_{i}\) stands for the industry effects that are controlled for by dummies for the Revised Pavitt classesFootnote 2; \({\chi }_{j}\) for country fixed effect, and \({\tau }_{t}\) for the time dummies, while ε is the error term.

2.2 Product innovation

The next equation in our model is product innovation, proxied at the industry level by the number of firms that introduced new or significantly improved goods and services over the total number of firms in the industry. Product innovation is regarded as an outcome of a technology-driven competitiveness strategy, driven by R&D efforts, the level of skills and the quality of jobs.

R&D is a generally used measurement of innovation input, regardless of the approach that is adopted (Dosi et al., 1990; Griliches, 1979; Van Reenen, 1996). We introduce a lag to account for the time needed for R&D results to emerge.

Human knowledge is essential in the innovation process and is reflected in the skills of workers; as above, we proxy them either with educational attainment or with occupational groups variables.

However, an industry’s knowledge is more than a mere sum of individual competences; organisational factors, learning by doing, and the accumulation of experience and capabilities are important in driving innovation (Dosi & Nelson, 2010; Lundvall, 2016). We include employment quality (proxied by the share of non-standard workers) as a variable reflecting some of these processes; we expect product innovation to be higher in industries with a lower share of non-standard workers.

Formally, the product innovation equation can be written as follows:

$${\text{ProductInnov}}_{{{\text{i}},{\text{j}},{\text{t}}}} = \beta_{0} + \beta_{{1}} {\text{R}}\& {\text{D}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \beta_{{2}} {\text{Skills}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \beta_{{3}} {\text{NSW}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \mu_{{\text{i}}} + \chi_{{\text{j}}} + \tau_{{\text{t}}} + \, \varepsilon_{{{\text{i}},{\text{j}},{\text{t}}}} $$
(2)

Details are the same as those provided for Eq. (1).

In addition to our baseline Eq. (2), we include additional control variables; Alongside the ‘technology push’, ‘demand pull’ is another relevant driver of innovation (Bogliacino & Pianta, 2013). To account for this, we introduce industry-level demand proxied by the rate of change of value added. Moreover, in line with Schumpeterian literature, we may expect product innovation to be higher in more concentrated industries, where larger firms exert greater market power (Schumpeter, 1942); therefore, we also include average firm size among regressors.

2.3 Labour productivity

The third equation in the model is labour productivity, measured as the average annual compound growth rate of value added per hour worked. The number of hours worked is a more appropriate measure of labour inputs as it accounts for differences in working time across countries and for the increasing share of part-time jobs. The key explanatory variables include fixed capital investment, product innovation, and wage growth.

Productivity increases when value added grows faster (or declines slower) than hours worked, based on different drivers. Investment aimed at expanding production capabilities is crucial for value added growth; some investments may focus on labour-saving processes that reduce labour inputs with little effect on value added. Product innovation leads to new markets that expand value added even when there is no increase in labour inputs. The efficiency wage approach shows that higher wages support productivity growth as they may increase the effort of employees, decrease shirking and attract more productive workers (Akerlof & Yellen, 1990; Shapiro and Stiglitz, 1984).Footnote 3

The productivity equation can be written as follows:

$$\Delta {\text{LabProd}}_{{{\text{i}},{\text{j}},{\text{t}}}} = \gamma_{0} + \gamma_{{1}} {\text{GFCF}}_{{{\text{i}},{\text{j}},{\text{t}}}} + \gamma_{{2}} {\text{ProductInnov}}_{{{\text{i}},{\text{j}},{\text{t}}}} + \gamma_{{3}} \Delta {\text{Wages}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \mu_{{\text{i}}} + \chi_{{\text{j}}} + \tau_{{\text{t}}} + \, \varepsilon_{{{\text{i}},{\text{j}},{\text{t}}}} $$
(3)

Details are the same as those provided for Eq. (1).

In the single equation model, we introduce several control variables. First, the skill level of labour employed in the industry is considered, using the shares of managers and manual workers; we expect skills to be complementary to capital investment and innovation in driving productivity.

Second, we include the relevance of non-standard jobs, building on the studies documenting its link to declining productivity growth (Kleinknecht, 2020; Lisi & Malo, 2017; Ortega & Marchante, 2010). The argument is that greater reliance on low-paid temporary and part-time work leads firms and industries to more labour-intensive regimes with less investment and innovation, slowing down productivity dynamics.

Third, we also test whether a catching-up process allows faster productivity growth in laggard countries; we calculated the percentage difference between an industry’s productivity and that of the country with the highest productivity levels in the same industry (Bogliacino & Pianta, 2011).

2.4 Wages

The fourth equation deals with wages. Considering the persisting large differences in wage levels across industries and countries and the very slow wage dynamics in the period we investigate, we focus on relative wages, using the percentage difference between the average wage in an industry and the average pay in the country’s sector with the highest wages; this measurement accounts for the conditions of national industries and labour markets.

In fact, empirical evidence suggests that wages significantly differ across industries for workers with the same characteristics—age, experience, education, occupation, gender and race (Krueger & Summers, 1988; Thaler, 1989)—as a result of the structural factors associated with technological capabilities, economic performance, union presence, etc.

The main drivers of relative wages include labour productivity, product innovation and workers’ skills. The gains from higher productivity are distributed between wages and profits on the basis of the bargaining power of capital and labour in industries. Product innovation opens up the possibility of higher rewards for workers involved in technological activities and learning processes. A higher level of skills—proxied by educational attainment—is likely to be found in the industries paying top wages.

The wage distance equation can be written as follows:

$${\text{WageDist}}_{{{\text{i}},{\text{j}},{\text{t}}}} = \delta_{0} + \delta_{{1}} \Delta {\text{LabProd}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \delta_{{2}} {\text{ProductInnov}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \delta_{{3}} {\text{Skills}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \mu_{{\text{i}}} + \chi_{{\text{j}}} + \tau_{{\text{t}}} + \, \varepsilon_{{{\text{i}},{\text{j}},{\text{t}}}} $$
(4)

Details are the same as those provided for Eq. (1).

In the single equation model, we introduce additional controls. First, skills are also proxied by the shares of managers and manual workers to assess the robustness of the relationship with wages. Second, union density is introduced to account for the institutional setting of industrial relations in industries. Third, firm size accounts for firms’ market power.

2.5 The virtuous circle model

The previous subsections have explored the drivers of employment quality, innovation, productivity and wages separately. We can now move beyond one-way relationships and combine the four equations in a simultaneous model that can summarise the good jobs-high innovation virtuous circle. In short, higher employment quality (and fewer non-standard jobs) complements technological activities and skills, leading to more product innovations that increase productivity growth; in turn, higher productivity brings wages close to those of top-paying industries, also improving employment quality. The four-equation simultaneous model is shown below:

$$\begin{gathered} {\text{QNSW}}_{{{\text{i}},{\text{j}},{\text{t}}}} = \alpha_{0} + \alpha_{{1}} {\text{Union}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \alpha_{{2}} {\text{Skills}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \alpha_{{3}} \Delta {\text{LabProd}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \alpha_{{4}} {\text{WageDist}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \mu_{{\text{i}}} + \chi_{{\text{j}}} + \tau_{{\text{t}}} + \varepsilon_{{{\text{i}},{\text{j}},{\text{t}}}} \hfill \\ {\text{ProductInnov}}_{{{\text{i}},{\text{j}},{\text{t}}}} = \beta_{0} + \beta_{{1}} {\text{R}}\& {\text{D}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \beta_{{2}} {\text{Skills}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \beta_{{3}} {\text{NSW}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \mu_{{\text{i}}} + \chi_{{\text{j}}} + \tau_{{\text{t}}} + \, \varepsilon_{{{\text{i}},{\text{j}},{\text{t}}}} \hfill \\ \Delta {\text{LabProd}}_{{{\text{i}},{\text{j}},{\text{t}}}} = \gamma_{0} + \gamma_{{1}} {\text{GFCF}}_{{{\text{i}},{\text{j}},{\text{t}}}} + \gamma_{{2}} {\text{ProductInnov}}_{{{\text{i}},{\text{j}},{\text{t}}}} + \gamma_{{3}} \Delta {\text{Wages}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \mu_{{\text{i}}} + \chi_{{\text{j}}} + \tau_{{\text{t}}} + \, \varepsilon_{{{\text{i}},{\text{j}},{\text{t}}}} \hfill \\ {\text{WageDist}}_{{{\text{i}},{\text{j}},{\text{t}}}} = \delta_{0} + \delta_{{1}} \Delta {\text{LabProd}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \delta_{{2}} {\text{ProductInnov}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \delta_{{3}} {\text{Skills}}_{{{\text{i}},{\text{j}},{\text{t}} - {1}}} + \mu_{{\text{i}}} + \chi_{{\text{j}}} + \tau_{{\text{t}}} + \, \varepsilon_{{{\text{i}},{\text{j}},{\text{t}}}} \hfill \\ \end{gathered}$$
(5)

The novelties in this simultaneous model have already been pointed out in presenting individual equations. In linking the equations together, we have to pay particular attention to the feedback loops that are present and to the structure of lags that is included. Several tests of the robustness of the model have been carried out, and are discussed with the results in Sect. 4.

3 Data and econometric strategy

3.1 Data overview

The empirical analysis uses the new version of the Sectoral Innovation Database with the NACE Rev.2 classification, which merges information on industries’ economic performance from different sources, including economic data from the OECD’s Structural Analysis database, innovation activity from the Community Innovation Survey (CIS), labour market variables from EU Labour Force Survey (Pianta et al., 2021). The investigation is carried out on 18 manufacturing and 23 service industries (listed in Table 6 in the Appendix) of six major European economies—Germany, Spain, France, Italy, the Netherlands and the UK—over the period 1994–2016.

The lack of annual innovation surveys led us to develop a periodical and balanced panel with six time periods corresponding to upswings and downswings of the business cycle. As shown in Table 7 in the Appendix, innovation variables (sourced from six different CIS waves) were matched with economic and labour market data with lags to account for the time necessary for innovations to develop their economic effects.Footnote 4

Finally, as the statistical classification of economic activities (NACE) moved from Rev. 1.1 to Rev. 2 in 2008 all data for years before 2008, expressed in terms of NACE Rev. 1.1, have been converted to NACE Rev. 2 using the conversion matrices provided by Pianta et al. (2021). Furthermore, all monetary variables have been deflated, converted to euros and adjusted for purchasing power parities to allow for cross-country comparability. The list of variables, a description of the sources, and the methodology used for their construction are reported in Table 8 in the Appendix, while Table 9 reports summary statistics.

3.2 Econometric strategy

First, we test each equation separately, with several controls and different specifications. Second, we test the simultaneous model with the system of equations with lags and feedback effects.

With regard to the single equations, we adopt the following identification strategy. First, all specifications include country, period and industry dummies to control for institutional, time and structural heterogeneities. Time dummies are needed to control for the business cycle and to prevent time-specific effects, otherwise captured by the error term, from rising endogeneity concerns. As industries differ in their technological regimes, we introduce dummies for the four Revised Pavitt classes and for the manufacturing-services divide, as they account for the structural characteristics of industries while avoiding the risk of multicollinearity potentially induced by the inclusion of a large number of sector-specific dummies. In the product innovation equation, dummies for technological regimes alone are included, as they properly account for heterogeneity.

Second, most explanatory variables are lagged by one period (3–4 years) to take into account the time required for economic effects to fully emerge in innovation processes, production systems and labour markets; this also reduces the risk of simultaneity-related endogeneity bias (Van Reenen, 1996).

Third, considering that industry data are grouped data of unequal size, we use the Weighted Least Squares (WLS) estimator (Wooldridge, 2002) to prevent sectors of small size and modest economic significance from contributing equally to other sectors in terms of information. We use the total number of employees in industry as weights; an alternative weight would be value added, but the former is preferred as it is not affected by prices.

In the simultaneous model, we estimate the system of equations using the three-stage least squares (3SLS) estimator because it allows us to account for cross-equation correlation among the errors. This method estimates all coefficients simultaneously and has a relative advantage with respect to 2SLS, which estimates each equation separately.Footnote 5 In the simultaneous model, we include country dummies to control for different institutional environments, a manufacturing dummy for structural differences, and time-fixed effects.

4 Results

The results of the individual equations on employment quality, product innovation, labour productivity and wage distance are presented in Tables 1, 2, 3 and 4, where different specifications with additional controls are included. The results of the simultaneous model are then shown in Table 5 and Table 11 (see Appendix).

Table 1 Regression results for Non-standard work
Full size table
Table 2 Regression results for Product innovation
Full size table
Table 3 Regression results for Labour productivity
Full size table
Table 4 Regression results for Wage distance
Full size table
Table 5 Results of the simultaneous model of job quality, product innovation, productivity, and wages
Full size table

4.1 Single equation models

The non-standard work equation. Table 1 shows that our model for the share of non-standard workers is supported by the empirical evidence on the industries of major European economies; two specifications are offered, and all variables are significant with the expected signs. More unionised sectors have higher employment quality (and fewer non-standard workers), confirming the important role of unions in building an appropriate institutional setting for industrial relations. In the specification of column 2, skill levels are proxied by the shares of university graduates and employees with secondary education or less, with opposing signs; sectors with more highly educated workers have fewer non-standard workers. Higher productivity growth also contributes to reducing non-standard jobs, confirming previous results (Grande et al., 2020) on the little-explored link between economic performance and employment quality. In the specification reported in column 1, we also include product innovation, finding the expected negative relationship with non-standard workers. Finally, the wage gap from the top-paying industry is positively associated with the relevance of non-standard work in both specifications.

Product innovation. Table 2 presents the results for the product innovation equation; three specifications are offered, confirming our expectations. In line with earlier studies, R&D expenditure is a key driver of new products in all versions of the model (Bogliacino & Pianta, 2011). We proxy skills by formal educational levels in columns 1 and 2 and by occupational groups in column 3; the share of university graduates always has a positive and significant effect on the ability of industries to introduce new products, while the share of workers with lower education is not significant, as they are less involved in the innovation process. In column 3, the share of managers, professionals and technicians in the total number of employees is positive but statistically insignificant, while the share of manual workers is negative and insignificant. In all versions of the model, employment quality contributes to product innovation; the estimates of the share of non-standard workers are always negative and highly significant, in line with Reljic et al. (2021). In columns 2 and 3, we introduced additional controls; firm size always has a positive and significant effect on new products, confirming the results of a large Schumpeterian literature. Conversely, the demand-pull effect, proxied by the change in industries’ value added, is never significant.

Labour productivity. Table 3 reports the results for labour productivity, again with three specifications of the model. Capital accumulation, proxied by gross fixed capital formation per hour worked, always has—as expected—a positive and significant effect on productivity change. Product innovation has a positive but non-significant coefficient in all versions. Efficiency wages are confirmed to be a significant driver of labour productivity in all equations. This result is consistent with the findings of Vergeer and Kleinknecht (2014), who investigated the impact of wage growth on labour productivity using a panel of 20 OECD countries over the period 1960–2004; the coefficients we find are remarkably similar to theirs, albeit marginally lower. In columns 2 and 3, we add additional controls. The catching-up effect of productivity is significant in both versions. This means that industries lagging behind Europe’s top performer are able to increase faster their hourly output, learning from and imitating other countries. Conversely, size effects are statistically insignificant. The role of labour in productivity growth is further confirmed by the role of employment quality; a higher share of non-standard workers significantly slows down productivity growth in both equations. In Eq. 3, we also added occupational groups as proxies of labour skills, finding that a higher share of manual workers significantly slows down productivity growth, while the share of managers has no significant effect.

Wages. Table 4 reports the results for the wage distance equation, with three specifications of the model that confirm our expectations. The gap of an industry’s average wage from the top-paying sector in the country is significantly reduced by faster productivity growth in the industry, in all versions of the model. The same result is found for product innovation, as innovative success contributes to increasing labour remuneration. Moreover, in all specifications, the latter grows faster where unions are present.

Moving to different specifications of the model, we include labour skills, in column 2 with formal education variables and in column 3 with occupational groups. The share of university graduates has a non-significant effect on relative wage increases, while a large share of employees with a secondary degree or less is associated with a higher distance from top wages in the economy. In column 3, we find that the share of managers and the share of manual workers in industry employees have contrasting and significant effects, reducing and expanding—in this order—the gap from the best-paying industries. Lastly, we introduce firm size as a control in column 2. It exhibits a positive and significant effect on wage distance; although firm-level evidence suggests that large firms pay more than smaller ones, at the sectoral level gaps emerge between the highest-paying knowledge-intensive services—finance, real estate and business services—where small average firm size dominates, and the labour-intensive industries with lower skills, where the average firm size is relatively higher.

The findings of the separate estimations of the four equations show that the model we proposed is supported by econometric results; the additional controls we introduced provide evidence of the robustness of the fundamental relationships and highlight further dimensions that contribute to these processes.

4.2 The simultaneous equation model

The simultaneous model combining the four equations on non-standard work, product innovation, productivity and wages has been estimated with three-stage least squares; the results are presented in Table 5. The simultaneous equations confirm the results of the individual equation estimations and are able to introduce cross-effects and feedback loops, providing a fuller and more integrated picture of the joint relationships between non-standard work, product innovation, productivity and wages.

Looking first at employment quality (column 1), we find that the share of non-standard work decreases with unionisation rate and with productivity growth, while it increases with a larger presence of employees with lower education and by relatively lower wages; all coefficients are significant. Conversely, the share of university-educated workers is not significant.

The product innovation equation (column 2) confirms that knowledge and competences developed in R&D activities and high labour skills are the fundamental drivers of innovation. R&D expenditure and the share of university graduates have a positive effect, while the shares of employees with lower education and of non-standard workers have negative coefficients; all are significant.

The labour productivity equation (column 3) shows that capital investment, technology and wage growth all have positive and significant coefficients. This simplified representation captures the key drivers of improved performances.Footnote 6

The wage distance equation (column 4) confirms that labour skills, productivity and new products shape the evolution of relative wages in industries. The distance of an industry’s wages from the top-paying sector of a country is increased by a high share of employees with lower education, while the share of university graduates does not emerge as significant. Labour productivity growth and product innovation are major engines of wage convergence; all variables are lagged to allow for the time required for changes in wage setting.Footnote 7

Taken together, we find that the joint system of equations should be preferred over the single equation approach because cross-equation correlations are significant.Footnote 8

The estimation results—concerning manufacturing and service industries of major European countries—confirm the existence of a virtuous circle between employment quality, product innovation, productivity and wages. The good jobs-high innovation virtuous circle is rooted in the complex, positive interactions that link together the characteristics of high-quality labour—in terms of skills, employment contracts, unionisation and wages—and high-quality capital—considering investments, R&D, product innovation and productivity.

Considering the diversity of industries, we further explored the presence of the virtuous circle in higher-tech vis-à-vis lower-tech sectors. The criteria used to group industries is the degree of cumulativeness of knowledge, which includes an important dimension of workers’ learning; we adopt here the classification developed by Peneder (2010) in this regard. In a further empirical analysis, we tested the four-equation model separately for the group of industries with medium or high cumulativeness of knowledge, as opposed to those with low cumulativenessFootnote 9 (Cetrulo et al., 2019; Kleinknecht et al., 2014). The results are reported in Table 11 in the Appendix. The results show that the virtuous circle is typical of industries with medium and high cumulativeness of knowledge—and more broadly higher technological levels—leading to the same findings as in Table 5. Conversely, when industries with lower learning processes are considered, some coefficients maintain their sign and significance, but the mechanism of the virtuous circle disappears; non-standard work, product innovation, labour productivity and wage distance lose their significance in the equations where they are introduced as regressors. This implies that the circular and cumulative nature of the relationships investigated between the quality of jobs and economic performance operates only in industries where technological and learning processes are more relevant (i.e. about two-thirds of industries). Instead, when industries have inadequate levels of these activities, the virtuous circle does not emerge.

Summarising our results, with our conceptual approach and econometric models, we have documented four key sets of relationships. First, high employment quality is driven by employees’ educational levels, unionisation, productivity and wage levels. Second, in turn, employment quality and high labour skills are combined with R&D efforts resulting in higher product innovation, the outcome of a strategy of technological competitiveness. Third, in turn, product innovation, investment and wage growth lead to greater labour productivity. Fourth, in turn, productivity gains are translated into higher wages, with labour skills and product innovation also driving up workers’ pay.

5 Conclusions

Several new insights emerge from the approach we developed and the results we presented. The most important novelty is the joint consideration of the quality of labour and the quality of capital and technology as drivers of progress. Most economic research has focused on R&D, innovation and investment as engines of growth, disregarding the essential need for human labour, knowledge and learning that is behind the same R&D, innovation and production activities (Lundvall, 2016). Long term success in these directions requires both high-quality labour and higher-quality capital.

Much research has disregarded this basic fact and has considered labour and wages simply as costs for business, which may reduce competitiveness relatively to low-wage producers in emerging countries. This narrow view of competitiveness goes hand in hand with a generic—and equally misleading—view of innovation, where no distinction is made in the types and goals of innovative activities.Footnote 10 Our results confirm the importance of the distinction between technological competitiveness, based on the development of product innovation with high-quality labour, and cost competitiveness, using new processes to reduce and deskill labour (Pianta, 2001). Only the former contributes to the virtuous circle that may bring sound, long-term growth.

Moving to the specific findings of this article, an important novelty is that enhancement of job quality should be seen as both the means and the end of higher innovation capabilities and higher productivity. We find that employment quality and R&D efforts are essential for successful innovation that in turn contributes—alongside investment—to higher productivity growth. The latter allows for higher wages that, in turn, improve working conditions by decreasing demand for non-standard workers. This is in line with empirical studies showing that labour market flexibilisation significantly reduces productivity and discourages R&D investments, patent applications and innovation (Cetrulo et al., 2019; Hoxha & Kleinknecht, 2020; Kleinknecht et al., 2014; Reljic et al., 2021). However, these works have focused on the one-way relationships from non-standard employment to innovation or productivity, disregarding the reverse causality. The model used in this article allows a more systematic investigation, identifying the role of employment quality in the good jobs-high innovation virtuous circle.

A further novelty concerns the inclusion of wages into the picture. In a context of stagnating incomes, we focus on wage levels relative to the top-paying industry in a country, accounting for the national labour market specificities. We revive the notion of efficiency wage and find that higher remunerations have a key role in contributing to higher productivity growth in industries. They also support employment quality, as the share of non-standard workers falls in higher-pay sectors. Moreover, higher wages go hand in hand with higher labour skills—measured either by education or by occupational groups—in their economic effects. In contrast, industries that rely more on low wages and non-standard forms of employment may shift towards a labour-intensive, low innovation regime, with a consequent slowdown of productivity. A key lesson from our approach and results is that wage dynamics—as part of the broader income distribution—are a crucial part of the explanation of economic change, including rising inequalities. For the virtuous circle of growth to operate, wages have to increase alongside the skill of workers and the quality of jobs.

Our findings are in line with many of the stylised facts of the empirical literature. The employment quality in industries increases with educational attainments, union representation and wages. Product innovation results from R&D and skilled labour. Labour productivity is driven by capital investment and product innovation. The latter also increases with a larger average firm size. Wages are supported by productivity, educational levels and innovation. While these stylised facts have usually been identified in isolation from one another, we provide here an integrated approach that links them all together.

We should also point out that our analysis of manufacturing and service sectors of major European economies confirms the importance of industry-level studies. They are able to account for the dynamics of structural change, the specificities of technological regimes and labour market institutions, all aspects that can hardly be captured either by aggregate analyses on national economies or by firm-level investigations on highly heterogeneous enterprises.

Taken together, our findings provide an interpretation of changes in European economies that is significantly different from the mainstream view that has long considered labour as a cost for firms, higher wages as a hindrance to competitiveness and labour market flexibility as a key driver of greater productivity. Building on an evolutionary perspective on innovation and industrial dynamics, and on institutional insights on labour, we have obtained the opposite results, summarised in our ‘virtuous circle’ narrative.

The explanation emerging from our conceptual framework and empirical findings is also supported by the results of studies on specific relationships within the virtuous circle we describe. Mainstream explanations of the productivity slowdown have been challenged by Kleinknecht (2020), who showed that sluggish wage growth has contributed to the slowdown through a more limited diffusion of labour-saving innovations; moreover, supply-side labour market reforms—leading to larger non-standard employment and greater turnover—have reduced workers’ learning processes, further contributing to the productivity slowdown.

The neoclassical argument that more flexible labour markets improve static efficiency in the allocation of labour is challenged when dynamic efficiency is considered and the ability of labour to contribute to innovation is brought into the picture. Several studies have shown that high flexibility—and in particular larger non-standard employment—has slowed down the innovative performances of firms and industries, measured by R&D expenditures, patenting and innovations, with further negative effects on productivity performances (Cetrulo et al., 2019; Kleinknecht et al., 2014; Reljic et al., 2021).

Finally, our analysis also brings a policy message. We have investigated European industries in two decades of sluggish growth and stagnant wages, when the good jobs-high innovation virtuous circle produced modest results for the aggregate economy. Key drivers of that growth trajectory were in fact missing, with declining capital investment and worsening employment quality. Two decades of 'structural reforms' in Europe's labour markets have resulted in the large expansion of non-standard employment and in stagnant wages. This has made it possible for many small, low-productivity firms to survive in the market without improvements in technologies, organisational capabilities and labour skills, with non-standard employees largely excluded from learning processes and accumulation of competences. The quality of employment and the dynamics of wages emerge from our analysis as relevant—but often disregarded—factors in explaining the performance of European industries.

The explanation we provide in our good jobs-high innovation narrative proposes an integrated approach to labour market, innovation and productivity policies, thereby moving beyond separate actions in each of these areas. Recent policies of labour market flexibilisation and wage compression should be reversed, opening up space for the faster productivity growth associated with good jobs. Technology and industrial policies should pay more attention to the development of new products and new markets, within a strategy of technological competitiveness, as opposed to mainly encouraging labour-saving and wage-cutting process innovations (ILO, 2016; Pianta et al., 2020; Rodrik & Stantcheva, 2021). After decades of crises and stagnation, policies that jointly improve high-quality innovation and the conditions of labour appear as crucial tools for reviving the good jobs-high innovation virtuous circle in Europe.