Technological change is one of the main drivers of the transition to the knowledge economy. The consequences of this transition for the labour market feature prominently on the political agendas in many Western countries. Amongst the numerous policy reports which have been published about this topic, one of the publications that triggered the policy debate the most is arguably the OECD’s (Reference Bellani and Bosio2019) ‘Under Pressure: The Squeezed Middle Class’. The report depicts the unequal distributive effects of new technologies: the number of jobs involving routine-intense tasks, typically occupied by medium-educated workers, declined due to the increased applicability of computers and robotics. At the same time, non-standard employment is on the rise, in particular at the bottom of the labour market. For instance, recent studies report a growing number of part-time working individuals who wish to work more hours (Greve, Reference Greve2017). This has raised concerns as these jobs exhibit higher risks of in-work poverty (Gardiner and Millar, Reference Gardiner and Millar2006; Marx et al., Reference Marx, Vanhille and Verbist2012; Brülle et al, Reference Brülle, Gangl, Levanon and Saburov2019).
We contribute to the comparative political economy literature by providing a novel theoretical explanation that ties the declined demand for routine-intense labour and involuntary part-time employment together. Building on previous studies showing that replacement risks are a key determinant in explaining labour market outcomes (Eichhorst and Marx, Reference Eichhorst, Marx, Eichhorst and Marx2015; Reichelt, Reference Reichelt2015; Bellani and Bosio, Reference Bellani and Bosio2019; Weisstanner, Reference Weisstanner2021; Mattijssen et al., Reference Mattijssen, Pavlopoulos and Smits2020 ), we argue that those in low-skill employment also feel the squeeze in demand for routine-intense labour. Workers in this segment of the labour market have a relatively high replacement risk, even though their jobs involve tasks that cannot easily be performed by computers or robotics (Autor et al., Reference Autor, Levy and Murnane2003; Acemoglu and Autor, Reference Acemoglu, Autor, Card and Ashenfelter2011). However, the skills needed to perform these tasks typically require little investments in education or training (Goldthorpe, Reference Goldthorpe2000; Emmenegger, Reference Emmenegger2009). Hence, these workers can easily be replaced by someone else. This makes them relatively vulnerable to shifts in supply of labour (Eichhorst and Marx, Reference Eichhorst, Marx, Eichhorst and Marx2015). The increased supply for low-skill jobs, resulting from a substantial number of medium-educated workers that shifted into low-skill employment (Murphy, Reference Murphy2014; Cortes, Reference Cortes2016; Dauth et al., Reference Dauth, Findeisen and Suedekum2017; Kurer and Gallego, Reference Kurer and Gallego2019; Acemoglu and Restrepo, Reference Acemoglu and Restrepo2020), has been corrosive to the bargaining power of these workers. That in turn pushes them to accept part-time jobs that involve fewer than the desired number of hours.
This study empirically assesses the link between the size of middle-skill employment and the incidence of involuntary part-time employment across 16 European countries for the period 1999-2010. Our empirical analysis also examines the role of active labour market policies (ALMPs). In particular, social investment-oriented ALMPs – policies aimed at stimulating labour market participation – might cushion competition for low-skill employment as they aim to prevent new social risks that are associated with the transition to the knowledge economy from materialising (Taylor-Gooby, Reference Taylor-Gooby2004; Bonoli, Reference Bonoli2013). So far, the effectiveness of ALMPs is mainly examined in relation to labour market participation, either measured using the employment rate or the unemployment rate (Abrassart, Reference Abrassart2015; Benda et al., Reference Benda, Koster and Van der Veen2019; Bakker and Van Vliet, Reference Bakker and Van Vliet2021). However, an empirical assessment of whether these programmes have actually protected workers against new social risks, like possessing obsolete skills due to the automation of routine-intense labour, is lacking. Our analysis provides insights into the effectiveness of ALMPs in protecting workers from these risks. These insights are also relevant for governments’ employment policies.
The dwindling shares of middle-skill employment
Recently, studies in the field of labour economics have shaken up the consensus that technological change mainly erodes low-skill employment (Autor et al., Reference Autor, Levy and Murnane2003; Spitz-Oener, Reference Spitz-Oener2006; Acemoglu and Autor, Reference Acemoglu, Autor, Card and Ashenfelter2011; Goos et al., Reference Goos, Manning and Salomons2014).Footnote 1 Focussing on a job’s task content rather than its skill level, these studies show that technological change is routine-biased. New technologies are especially suited to perform routine tasks, that can be characterised as repetitive, procedural and rule-based. As jobs involving these routine tasks typically lie in the middle of the skills distribution, medium-educated workers have been gradually substituted for computers and robotics. This has resulted in a squeezed demand for routine-intense labour, attested by dwindling shares of middle-skill employment in Western economies (Michaels et al., Reference Michaels, Natraj and Van Reenen2014; Gregory et al., Reference Gregory, Salomons and Zierahn2022).
In contrast, both the shares of low- and high-skill employment have grown during the same period. Again, the explanation for this trend is rooted in the task content of these jobs (Autor et al., Reference Autor, Levy and Murnane2003; Acemoglu and Autor, Reference Acemoglu, Autor, Card and Ashenfelter2011). On the higher end of the skills distribution, digital capital has complemented workers performing non-routine cognitive tasks. Accordingly, the demand for high-educated workers increased, fuelling the transition to the knowledge economy. On the other end of the skills distribution, jobs involve non-routine manual tasks that cannot easily be substituted by computers or robotics, like cleaning, renovating, or serving. This implies that these workers are relatively sheltered from automation risks. Besides, the demand for low-skill labour increased, predominately due to the growing demand for low-skill services (Goos and Manning, Reference Goos and Manning2007; Autor and Dorn, Reference Autor and Dorn2013).Footnote 2
These findings resonate with the comparative political economy literature analysing the service sector expansion that underpinned the transition to the knowledge economy (Iversen and Wren, Reference Iversen and Wren1998; Wren, Reference Wren and Wren2013; Hope and Martelli, Reference Hope and Martelli2019). On this reading, the adoption of new technologies in high-end service sectors increased the demand for high-educated workers, given their complementary skills. As a corollary, high-educated workers concentrated in sectors, like finance, business services and communication.Footnote 3 Meanwhile, low-educated workers shifted from shrinking traditional sectors, like agriculture and manufacturing, into low-end service sectors in which the diffusion of new technologies is relatively limited.
Our argument builds on these insights, but departs from the automation of routine-intense labour and its effect on employment possibilities for medium-educated workers. Note that jobs involving routine tasks were both prevalent in the manufacturing sector and high-end service sectors. Not only the number of blue-collar jobs, like machine operators and assemblers, but also white-collar jobs, like customer service employees and office clerks, have sharply declined during the last decades (Goos et al., Reference Goos, Manning and Salomons2009; Reference Goos, Manning and Salomons2014). This study therefore focusses on jobs instead of sectors in analysing the widespread effects of technological change.
The squeezed demand for routine-intense labour has affected medium-educated workers’ employment possibilities: they have become more likely to work in low-skill employment over the past decades (OECD, 2020; Van Vliet et al., Reference Van Vliet, Bakker, Van Doorn, Fischer and Strauss2021). This shift stems from the nature of these workers’ skills which are typically less suited to the tasks involved in high-skill jobs compared to those in low-skill jobs. This explains why a substantial number of displaced medium-educated workers shifted into low-skill employment, albeit some middle-educated workers managed to retain their job (Murphy, Reference Murphy2014; Cortes, Reference Cortes2016; Kurer and Gallego, Reference Kurer and Gallego2019; Acemoglu and Restrepo, Reference Acemoglu and Restrepo2020). Furthermore, young labour market entrants who completed medium-education are more likely to start working in low-skill jobs (Dauth et al., Reference Dauth, Findeisen and Suedekum2017). At this point, it is important to note that there might be variation in this regard between countries as ALMPs, in particular training, and vocational education and training (VET) systems might provide workers with the necessary skills to shift into high-skill jobs (Busemeyer and Trampusch, Reference Busemeyer and Trampusch2012; Wang, Reference Wang2020). Nevertheless, the overall picture shows a growing share of middle-educated workers in low-skill employment, increasing the competitive pool for jobs in this segment of the labour market.
The above-described trends are confirmed by Figure 1, which shows employment data of 16 European countries in the period between 1999 and 2010. First, low-skill employment as a share of total employment indeed increased. The United Kingdom is leading the pack, just ahead of Finland, Norway, and Spain, with an increase in the share of low-skill employment of eight percentage points. The only exceptions are Italy and Luxembourg – in these countries the share of low-skill employment declined, respectively with four and one percentage points. More importantly, the figure confirms the expected shift of medium-educated workers into low-skill employment. In fact, the shares of these workers in low-skill jobs increased by at least 10 percentage points in the majority of the countries. The largest increases can be found in Belgium, Greece and Finland, where their share increased by approximately 18 percentage points. Interestingly, Denmark, Norway, and Sweden – the other three Scandinavian countries – are the only countries bucking this trend.
Replacement risks and shifting supply
Although sheltered from automation risks, the squeezed demand for routine-intense labour also worsens the employment possibilities for workers in low-skill employment. The skills needed to perform tasks associated with low-skill jobs typically require little investment in training or education. Due to this skill profile workers at the bottom end of the labour market have a relatively high replacement risk: employers can replace these workers relatively easily (Goldthorpe, Reference Goldthorpe2000; Emmenegger, Reference Emmenegger2009; Eichhorst and Marx, Reference Eichhorst, Marx, Eichhorst and Marx2015). As replacement is lurking, the bargaining power of these workers is relatively limited. In the same vein, as the burden of finding a replacement is relatively low, employers face few incentives to bind workers to the firm through favourable working conditions, like permanent full-time contracts. This mechanism explains differences in workers’ job trajectories (Eichhorst and Marx, Reference Eichhorst, Marx, Eichhorst and Marx2015; Reichelt, Reference Reichelt2015; Mattijssen et al., Reference Mattijssen, Pavlopoulos and Smits2020), and relative wage risks resulting from labour market flexibilisation (Bellani and Bosio, Reference Bellani and Bosio2019; Weisstanner, Reference Weisstanner2021).
The skill profile of workers in low-skill employment makes them relatively vulnerable to the shifts in labour supply stemming from the automation of routine-intense labour. The described inflow of medium-educated workers in the competitive pool for low-skill jobs implies a growing number of potential substitutes for workers depending on this type of employment (Acemoglu and Restrepo, Reference Acemoglu and Restrepo2020). This in turn has a corrosive effect on these workers’ already limited bargaining power, pushing them to accept part-time jobs that involve few hours to prevent unemployment. Moreover, recall that the growing shares of low-skill employment are mainly driven by the increased demand for low-skill services, which includes a lot of jobs that require flexible working times to meet customers’ needs (Hipp et al., Reference Hipp, Bernhardt and Allmendinger2015). The increased supply of labour for low-skill jobs gives employers greater leverage to achieve this flexibility through the use of part-time contracts.
This dynamic implies a macro-level increase in the number of part-time employed workers who wish to work more hours, in particular at the bottom end of the labour market. Studies presenting descriptive evidence show that their numbers are indeed rising in Europe (Greve, Reference Greve2017). Besides, the share of involuntary part-time employment is highest amongst low-skilled service workers (Peugny, Reference Peugny2019). The labour market data presented in Figure 2 tell the same tale. Whereas the share of involuntary part-time employment is relatively stable in middle- and high-skill employment, the share has increased in low-skill employment. Hence, we expect that a decline in the size of middle-skill employment is associated with an increase in the incidence of involuntary part-time employment at the bottom end of the labour market.
ALMPs: cushioning competition
Active labour market policies (ALMPs) might cushion the competition that stems from the increased supply of labour for low-skill jobs. These policies took off in the 1990s when many governments transformed their welfare states against a backdrop of growing concerns regarding their carrying capacity, and the emergence of new social risks stemming from the transition to the knowledge economy (Hemerijck, Reference Hemerijck2013; Nelson, Reference Nelson2013; Clasen et al., Reference Clasen, Clegg and Goerne2016).Footnote 4 Accordingly, the rationale underlying labour market policies became the promotion of labour market participation through activation and investment in human capital policy (Bonoli and Natali, Reference Bonoli, Natali, Bonoli and Natali2012:9).
Following Bonoli (Reference Bonoli2013), we distinguish two types of ALMPs. On the one hand, there are so-called social investment-oriented ALMPs, which invest in human capital and have a pro-market employment orientation. These policies are designed to increase the quantity and quality of the labour force. In particular, two policies are relevant in this regard: training and employment incentives. First, training increases an individual’s employability by human capital enhancement, which has been associated with an increase in labour market participation (Kluve, Reference Kluve2010; Card et al., Reference Card, Kluve and Weber2018). Training can also help workers, who found themselves possessing obsolete skills due to the automation of routine-intense labour, acquiring the necessary skills to shift into high-skill jobs (Rodrik and Stantcheva, Reference Rodrik and Stantcheva2021). This might limit the inflow of redundant workers in the competitive pool for low-skill employment. As a result, we expect that high levels of effort on training attenuate the rise in involuntary part-time employment by cushioning competition.
Second, employment incentives also aim to stimulate labour market participation (Graversen and Van Ours, Reference Graversen and Van Ours2011). However, we expect that the effect differs regarding the prevalence of involuntary part-time employment. The bulk of spending in this category includes making-work-pay-policies, like income maintenance and support payments, and back-to-work bonuses. This entails payments to formerly unemployed individuals who have taken up part-time or full-time employment and as a result experience an income loss relative to unemployment benefits. Therefore, they are encouraged to accept (part-time) jobs even though earnings might be lower than the level of benefits due to a lower wage or fewer hours (Haapanala, 2021). Thus, we expect that effort on employment incentives increases the incidence of involuntary part-time employment.
On the other hand, there are demand-side ALMPs that stimulate labour market participation by increasing employment possibilities, like public job creation schemes. The creation of these jobs offers workers dependent on low-skill employment an alternative to exit unemployment. This reduces the need to accept a part-time job in the private sector to prevent unemployment. However, the effectiveness of these schemes in relation to labour market participation is inconclusive (Kluve, Reference Kluve2010; Card et al., Reference Card, Kluve and Weber2018). This might call the attractiveness of these public jobs as a realistic alternative into question. Nevertheless, we expect that effort on direct job creation mitigates competition and is thus associated with a decrease in the incidence of involuntary part-time employment.
3. Method, measures and data
Using a partial adjustment model, we regress the share of involuntary part-time employed workers in a country on indicators measuring the automation of routine-intense labour, active labour market policies, and institutional and economic factors (see Appendix 1 for a technical explanation of our model). As this model captures both transitory and permanent effects (De Boef and Keele, Reference De Boef and Keele2008; Williams and Whitten, Reference Williams and Whitten2012), we are able to analyse not only the immediate impact of the decline of middle-skill jobs but also the way this contributes to the structural change in the dynamics at the bottom end of the labour market. Note that our model controls for serial correlation, panel-heteroscedasticity and contemporaneous spatial correlation (Beck and Katz, Reference Beck and Katz2011).
We define our dependent variable as the number of part-time employed workers who wish to work more than the current number of hours as a share of the total number of workers in low-skill employment (see Appendix 2 for the operationalisation of all variables and sources). To define low-skill employment, we follow Goos et al. (Reference Goos, Manning and Salomons2014) and categorise jobs based on their mean wage rank using two-digit International Standard Classification for Occupations (ISCO) codes.Footnote 5 The jobs that are included in this category are typically low-paying and involve few routine tasks, meaning that they are not easy to automate.
Focussing on low-skill employment is relevant for two reasons. First, the automation of routine-intense labour intensifies competition at the bottom end of the labour market (Acemoglu and Restrepo, Reference Acemoglu and Restrepo2020). Workers in these jobs are especially vulnerable to competitive pressures as they have a relatively high replacement risk. Second, our measure is relevant in the context of the growing number of working poor in Europe, as insufficient working hours, especially in low-skill employment, are one of the main determinants of in-work poverty (Gardiner and Millar, Reference Gardiner and Millar2006; Marx et al., Reference Marx, Vanhille and Verbist2012; Brülle et al., Reference Brülle, Gangl, Levanon and Saburov2019).
For our measure of the size of middle-skill employment, the main independent variable, we again use two-digit ISCO codes to categorise jobs based on the ranking provided by Goos et al. (Reference Goos, Manning and Salomons2014).Footnote 6 Subsequently, we use the relative number of hours worked in this category to measure the size of middle-skill employment: this measure is frequently used to analyse the labour market structure (Verdugo and Allègre, Reference Verdugo and Allègre2020; Maarek and Moiteaux, Reference Maarek and Moiteaux2021). In this way, we are able to capture the decrease in demand for routine-intense labour, which is associated with an increase in the competitive pool for low-skill jobs.
The data underlying our dependent variable and the main independent variable come from the European Labour Force Survey (Eurostat, 2019). We aggregated micro-level data to create time-series cross-sectional data for 16 countries between 1999 and 2010.Footnote 7 Due to a break in the occupational classification in 2011 (from ISCO88 to ISCO08), our analysis focuses on the longest consistent time series available (1999-2010). This could be a potential limitation of our analysis. However, a sensitivity analysis shows that our results also hold if we extend the period to 2018 by applying a crosswalk to link both classifications (see the robustness tests). Furthermore, we restricted our sample to individuals of working-age (15-64), excluding full-time students, unpaid family workers and the agricultural sector. Note that the results of our analysis are not sensitive to the exclusion of these categories.
To analyse the effect of ALMPs, we focus on the three previously described policies: two social investment-oriented ALMPs, training and employment incentives, and public job creation schemes. Effort on each policy is operationalised as expenditures corrected by the number of unemployed relative to GDP per capita. In this regard, the unemployed serve as a proxy for the number of recipients (Van Vliet and Koster, Reference Van Vliet and Koster2011; Kuitto, Reference Kuitto2016). The data underlying our measure are from the OECD’s Labour Market Programmes, and the National Accounts databases.
In our analysis, we also control for a number of institutions and economic explanations of involuntary part-time employment. First, passive labour market policies (PLMPs) are associated with a shrinkage of the labour supply (Bassanini and Duval, Reference Bassanini and Duval2009). Hence, we include a measure of PLMPs, which comprises unemployment benefits and early retirement programmes, to control for this. Second, the strictness of employment protection legislation (EPL) might impact hiring decisions of employers (Kalleberg, Reference Kalleberg2003). We control for this by including the OECD’s EPL indicator for regular contracts. Third, the degree of firm involvement in the provision of VET determines the development and quality of medium-educated worker’s skills (Busemeyer and Trampusch, Reference Busemeyer and Trampusch2012 ). The theoretical skills that are important for (high-end) service sector jobs are less provided if firms are heavily involved in the provision of VET, compared to a school-based setting or on-the-job-learning (Anderson and Hassel, Reference Anderson, Hassel and Wren2013). To account for these differences in the skill formation process, we follow Busemeyer and Iversen (Reference Busemeyer, Iversen, Busemeyer and Trampusch2012) and control for the share of students in vocational training schemes that combine school- and workplace-based VET. Fourth, the effect of automation on labour market outcomes is conditional on the strength of organised labour, reflected by trade union membership and wage-setting institutions (Parolin, Reference Parolin2021). Automation has, however, also chipped away at organised labour’s power (Meyer, Reference Meyer2019), and trade unions typically have difficulties gaining ground in the service sector (Brady, Reference Brady2007; Palier and Thelen, Reference Palier and Thelen2010). We control for the strength of organised labour by adding trade union density and the centralisation of wage bargaining. Fifth, we use government partisanship to control for the impact of left-wing governments. Left-wing parties express more criticism regarding various forms of non-standard employment (Picot and Menéndez, Reference Picot and Menéndez2019). Finally, we account for economic conditions by including GDP growth and the unemployment rate. Economic downturns are typically associated with an increase in involuntary part-time employment (Valletta et al., Reference Valletta, Bengali and Van der List2020).
Table 1a shows the estimation results from our partial adjustment model. The coefficients of the size of middle-skill employment show a negative and statistically significant relationship with the incidence of involuntary part-time employment at the bottom end of the labour market. In other words, a decline in the size of middle-skill employment is associated with an increase in the incidence of involuntary part-time employment. This implies that the automation of routine-intense labour indeed intensifies competition. Moreover, the long-run multiplier shows that a percentage point decrease in middle-skill employment is associated with a permanent increase in involuntary part-time employment of approximately 0.8 percentage points (see Table 1b).Footnote 8 These results confirm our hypothesis that technological change pushes workers to accept part-time jobs that have fewer than the desired number of hours.
Note: Panel corrected standard errors (in parentheses) and panel specific AR1 structure (estimated through Prais-Winsten transformation). Trend not shown. * p<0.1, ** p<0.05, *** p<0.01
Note: Panel corrected standard errors (in parentheses) and panel specific AR1 structure (estimated through Prais-Winsten transformation). * p<0.1, ** p<0.05, *** p<0.01
Turning to the social investment-oriented ALMPs (see Table 1a), effort on training is associated with a decrease in the incidence of involuntary part-time employment. The coefficient is significant, indicating that a one-unit increase in effort on training per unemployed as a share of GDP tends to decrease the incidence of involuntary part-time employment by 0.06 percentage points. This finding is in line with previous studies reporting positive effects regarding effort on training (Kluve, Reference Kluve2010; Card et al., Reference Card, Kluve and Weber2018). In contrast, effort on employment incentives is associated with an increase in involuntary part-time employment. This is in line with our expectation that employment incentives encourage unemployed individuals to accept jobs that involve few hours. Finally, effort on direct job creation is associated with a decrease in the incidence of involuntary part-time employment, confirming our expectation. This might provide support for the importance of the cushioning role of demand-side policies, like the “Melkertbanen” in the Netherlands, and “Nouveaux Services Emplois Jeunes” in France (Daguerre, Reference Daguerre2007; Huo, Reference Huo2009; Vlandas, Reference Vlandas2013).
With regard to the results of the institutional and economic factors, the coefficients of effort on PLMP and EPL are insignificant. Firm involvement in training seems to increase involuntary part-time employment, providing support for the argument that dual training systems limit a countries’ ability to adjust to the knowledge economy (Anderson and Hassel, Reference Anderson, Hassel and Wren2013). Furthermore, trade union density is associated with a decrease in the incidence of involuntary part-time employment, which is in line with previous studies (Parolin, Reference Parolin2021). The coefficients of bargaining centralisation and left-wing governments have the expected sign but are insignificant. Finally, GDP growth is associated with a decrease in the incidence of involuntary part-time employment, whereas the coefficient for unemployment is not significant.
Our second hypothesis pertains the potential cushioning role of ALMPs regarding the competition stemming from the squeezed demand for routine-intense labour. Figure 3 graphically plots the result of the interaction with effort on training (see Table 1a for the coefficients). The figure shows that the decline of middle-skill employment does not have a significant effect in countries with relatively high levels of effort on training. However, the average effort on training exceeds 16 percent per unemployed as a share of GDP per capita only in Denmark for the entire period. Besides, for some periods in Austria (between 2008 and 2010), Norway (between 1999 and 2004, and in 2007) and Sweden (between 1999 and 2002) the yearly level of effort on training exceeds the threshold. Overall, these findings support our hypothesis that effort on training cushions competition, and is thus associated with a decrease in involuntary part-time employment at the bottom end of the labour market. In this regard, this policy seems to live up to expectations.
Figure 4, which plots the interaction with effort on employment incentives, does not show a significant effect. Hence, employment incentives do not exacerbate competition for low- skill employment. Finally, Figure 5 plots the results of the interaction with effort on direct job creation. The plot reveals a similar pattern as described for effort on training. In countries with higher levels of effort on direct job creation, a decline in the size of middle-skill employment does not have a significant effect. The Netherlands is the only country in which effort on job creation exceeds the threshold of 10 percent per unemployed as a share of GDP per capita between 1999 and 2010. Furthermore, effort in Belgium (in 2000 and 2001), France (between 2000 and 2004, and in 2007), Ireland (between 1999-2008), and Luxembourg (in 2010) also exceeds the threshold during some periods. Although the coefficient of the interaction with job creation is only significant at the 10 percent level (also reflected in the confidence intervals), this finding underlines the previously suggested success of demand-side policies in cushioning competition stemming from the automation of routine-intense labour.
Appendix 2 presents a number of additional estimations to examine the robustness of our results. The first row in this table presents the standardised beta coefficient and LRM of our baseline estimation of the association between the size of middle-skill employment and the incidence of involuntary part-time employment (see Table 1, first column). First, the findings hold if we extend the period to 2018 by applying a crosswalk to link the ISCO-88 and ISCO-08 classifications. Second, our findings are robust if we extend our analysis to the total employment by including all jobs (low-, middle- and high-skill). Note that the magnitude of the coefficients is relatively small compared to the original results. This seems to confirm that the competitive pressures mainly affect those at the bottom end of the labour market. Next, we limit our sample to individuals of prime working age (between 25-54). In this way, we rule out the possibility that our findings are mainly driven by workers just entering the labour market or approaching retirement. Indeed, the presented coefficients are fairly similar. Third, we restrict our sample to either men or women. There are various studies showing that (involuntary) part-time employment is especially relevant regarding women’s labour market position (Insarauto, Reference Insarauto2021). Although our results confirm that women are more affected by the changing labour market structure, the coefficient for the estimation only including men is also highly significant.
Subsequently, we test the robustness of our results for different methodological specifications. First, we include the initial share of low-skill employment to control for differences between countries to the extent they already relied on low-skill jobs. Second, our preferred specification does not include country and/or year fixed-effects as this might introduce bias into the model (Nickell, Reference Nickell1981) or amplify bias (Plümper and Troeger, Reference Plümper and Troeger2019). Nevertheless, our results are largely unaltered by the introduction of country fixed effects or a combination of country and year fixed effects. Finally, we test whether our results hold up using a general error correction model.Footnote 9 Again, the coefficients remain highly significant and comparable in magnitude.
Technological change has transformed Western economies’ labour markets substantially since the early 1990s. As artificial intelligence, computers, and robotics proved to be a low-cost substitute for routine-intense labour, medium-educated workers suffered a fall in demand. As a result, a sizeable proportion of medium-educated workers is forced to shift into low-skill employment. We argue that the inflow of these workers in the competitive pool for low-skill employment worsens the employment possibilities of those dependent on this type of employment. Workers in low-skill employment have a high replacement risk: they typically perform tasks that require little investment in training. Their bargaining position corroded as the automation of routine-intense labour increased their potential number of substitutes. This pushed these individuals to accept part-time jobs that involve fewer than the desired number of hours.
Our empirical analysis provides support for the argument that the automation of routine-intense labour is associated with an increase in involuntary part-time employment at the bottom end of the labour market. Analysing 16 European countries between 1999 and 2010, we show that involuntary part-time employment in this segment of the labour market increased at the macro-level. Accounting for the cross-country variation in (labour market) institutions, we show that the decrease in the size of middle-skill employment is associated with an increase in the incidence of involuntary part-time employment, both in the short and the long run. This finding fits within previous studies that showed how high replacement risks impact job trajectories (Reichelt, Reference Reichelt2015; Mattijssen et al., Reference Mattijssen, Pavlopoulos and Smits2020), translates into wage pressure in the context of flexibilization (Bellani and Bosio, Reference Bellani and Bosio2019; Weisstanner, Reference Weisstanner2021), and affects job quality (Eichhorst and Marx, Reference Eichhorst, Marx, Eichhorst and Marx2015). Moreover, the results add to the descriptive evidence that the transition to the knowledge economy and the squeezed demand for routine-intense labour are linked to a rise in non-standard employment (Green and Livanos, Reference Green and Livanos2017; Greve, Reference Greve2017; Peugny, Reference Peugny2019).
Furthermore, the results confirm that effort on training cushions competition for low-skill employment. Previous studies have already shown that training is associated with an increase in the employment rate (Kluve, Reference Kluve2010; Card et al., Reference Card, Kluve and Weber2018). We add to this that training helps individuals, whose skills have become obsolete due to automation, acquiring the skills necessary to shift into high-skill jobs. As a result, these programmes relieve pressure on the bottom end of the labour market. We also presented evidence that effort on direct job creation has a similar effect. Encompassing job creation schemes mitigate the adverse effect of the automation of routine-intense labour at the bottom end of the labour market. However, note that our analysis of the effectiveness of ALMPs comes with two limitations. First, our measure does not include benefit conditions and eligibility rules (Knotz, Reference Knotz2020). Besides, ALMPs can complement each other: the success of an individual policy might hinge on such complementarities (Benda et al., Reference Benda, Koster and Van der Veen2019; Bakker and Van Vliet, Reference Bakker and Van Vliet2021).
To conclude, the transition to the knowledge economy goes hand in hand with increased competition at the bottom end of the labour market, increasing involuntary part-time employment. However, training and direct job creation can cushion this competition. These findings are relevant considering predictions that more jobs will disappear in the next 15 to 20 years (Frey and Osborne, Reference Frey and Osborne2013). In this regard, they provide support for the prominence of skills in the European Pillar of Social Rights; an initiative of the European Commission to reform European labour markets. Even so, governments face obstacles in expanding such policies, as recent studies show that these policies have not found their way to workers’ hearts yet (Bremer and Bürgisser, Reference Bremer and Bürgisser2022; Busemeyer and Sham, Reference Busemeyer and Sham2021).
To view supplementary material for this article, please visit https://doi.org/10.1017/S0047279422000629
This study is part of the research project Flexibilisation, globalisation and technological change: consequences for labour markets and social security. Financial support from Institute Gak is gratefully acknowledged. Earlier versions of this paper were presented at the SASE Conference (July 2020), the 78th Annual Midwest Political Science Association Conference (April 2021), and the 27th International Conference of Europeanists (June 2021). We thank Despina Alexiadou, David Weisstanner and all participants as well as two anonymous reviewers for their helpful comments and suggestions.
The authors declare none.