2.1 Data and sample selection

For our empirical analyses, we use a novel data set that was developed by linking the “IAB-ZEW Labour Market 4.0” establishment survey with employment biographies from social security records and additional information from BERUFENET and the IAB Establishment Panel.

Establishment survey: The “IAB-ZEW Labour Market 4.0” establishment survey on the use and importance of new digital technologies is a representative survey of establishments in Germany. ^[3] Approximately 2,000 establishments participated in the survey in 2016. The sample was drawn from the establishment data file of the IAB. It was stratified by four establishment size categories, East and West Germany and five sector categories differentiating between 1. non-knowledge intensive manufacturing (e. g. furniture producers, building establishments), 2. knowledge intensive manufacturing (e. g. car manufacturers, machine manufacturers), 3. non-knowledge intensive services (e. g. wholesalers, restaurants, logistics), 4. knowledge intensive services (e. g. scientific services, banks, insurances) and 5. information and communication technologies (ICT) (e. g. producers of data processing equipment, consumer electronics or telecommunications equipment, enterprises that provide services in information technology, telecommunication or data processing). While the basic differentiation is between producers and service establishments, the ICT sector (where both producers and service establishments are included) is viewed separately because of its central role as a technology hub and core enabler of a digitalized economy. The technical managers and experts of the establishments were asked to categorize production technologies (PT), on the one hand, and office and communication technologies (OCT), on the other hand, into three different classes (see Table 1). The higher the class is, the higher the degree of digitization.

PT 3 represents new digital production technologies (in Germany also called “industry 4.0 technology”) and OCT 3 represents new digital office and communication technologies (in Germany also called “services 4.0 technology”). In 2016, the establishments were asked for their current status as well as for the status five years ago and the expected status in the future (in five years). Collating this status information, we are able to identify temporal changes. If the share of PT 3 and/or OCT 3 within the establishment increases over time, this is a clear indication of investments in new digital technologies. It is important to know that producers usually use both PT and OCT, while some service establishments only use OCT.

Figure 1 shows that the share of new digital technologies (both industry 4.0- and services 4.0-technologies) is still very limited. Looking at the values from 2016 (in the middle columns), an average of 5.1 % of PT and 7.8 % of OCT can be assigned to new digital technologies. The degree of IT-supported OCT (49.4 %) is also distinctly higher than the share of indirectly controlled PT (11.9 %). Due to the “natural” high volume of digital technologies in OCT, the share of non-IT-supported technologies is much lower (42.8 %) than the corresponding group of manually controlled PT (83.1 %). In both PT and OCT, there is a slight trend towards IT-supported and indirectly controlled technologies, but this trend seems to evolve rather slowly.

Figure 1:

Trends in the usage of digital technologies across German establishments.

Based on this categorization and further information provided by the survey, ^[4] we differentiate the groups of establishments as follows: Latecomers are defined as establishments who indicate that they do not use new digital technologies; accordingly, the share of new digital technologies in the year 2016 is 0 % (see Figure 2). Pioneer establishments in digital technologies (pioneers) already use new digital technologies and invested in new digital technologies between 2011 and 2016. For this group of establishments, the degree of IT-supported OCT increased from 12.5 % in 2011 to 25.1 % in the year 2016, and the degree of autonomously controlled PT increased from 7.2 % in 2011 to 13.7 % in 2016 (see Figure 4). The remaining establishments are gathered in the peloton group. The average degree of 4.0-technologies in the year 2016 is approximately 6 % for both OCT and PT (see Figure 3). This differentiation leads to 363 latecomers (18 % of the establishments in the sample), 1.172 peloton establishments (58 %) and 497 pioneers (25 %) in our data set.

Figure 2:

Trends in the usage of digital technologies across latecomer establishments.

Figure 3:

Trends in the usage of digital technologies across peloton establishments.

Figure 4:

Trends in the usage of digital technologies across pioneer establishments.

Employment histories: Next, we link the survey data to employment biographies from the social security records (Beschäftigten-Historik, BeH, Version 10.02.01-171117) of all workers employed in the surveyed establishments between 2011 and 2016. The BeH covers the majority of the German workforce and is representative of dependent workers. ^[5] It contains important personal characteristics (sex, age, skill, nationality, job status, occupation) as well as information on the region, industry, and wage. Because the BeH is derived from mandatory employer notifications to the German social security system, the data are highly accurate and reliable.

Despite these strengths, the BeH suffers from some moderate limitations, as follows: first, earnings are top-coded in the data. For this reason, we estimate censored regressions for each year (we use age, skill, establishment size, occupation, establishments’ foreigner share, region and type of region as covariates), separated for male and female workers, and impute the censored wages. We follow the procedure described in Dustmann et al. (2009), but we include more covariates than they do in their baseline imputation model. The wages are then deflated to 2010 prices. Second, working time is only reported in three categories: full-time, part-time with at least 50 % of full-time working hours and part-time with less than 50 % of full-time working hours. To avoid bias due to imprecise information on working time, we restrict our analysis to men and women (16–65 years-old) working full-time, excluding apprentices, trainees and working students. Third, the data show some inconsistencies (or missing data) with regard to workers’ formal education, which we refer to in the following as “skill”. We apply a basic version of the approach proposed by Fitzenberger et al. (2006) and impute the information concerning education according to the information available in preceding or subsequent spells of the individuals’ employment history. Finally, we exclude observations with dubious wage information below a specific time-varying threshold. ^[6] Focusing on employment spells overlapping June 30th of a year, our sample selection comprises approximately 1.1 million worker-year observations.

The aim of our study is to investigate how workers’ wages are affected by the investments in new digital technologies by establishments. Because the survey provides results on the changes of technologies between 2011 and 2016, but do not include the exact dates of the technology investments, we focus on workers that are employed in both years 2011 and 2016. That is, we create a balanced panel of male and female establishment stayers. This allows us to measure the wage effects of establishment stayers in a meaningful way, but as a consequence, the paper does not discuss the wage effects for establishment leavers and new entrants to establishments (or the wage effects for part-time workers). We are aware that our sample might be a positive selection of workers. We discuss this issue in the last paragraph of Section 3. Altogether, we observe in each year 90,982 male and female full-time workers in 1,525 establishments. ^[7]

BERUFENET: The data source from which we identify digital tools is the BERUFENET, an online expert database of the Federal Employment Agency. ^[8] The BERUFENET offers detailed information about every single occupation, e. g. occupational and vocational training contents, tasks, tools, entrance requirements, earnings and employment perspectives. The occupations are based on the German classification of occupations (Klassifikation der Berufe 2010, KldB2010). The key section of the BERUFENET for the purpose of this paper is the section on work items/tools (Arbeitsgegenstaende). We use a unique BERUFENET data extract of the Federal employment agency. This extract facilitates analyses of tools for 2,963 occupations at 8-digit level of KldB2010.

The definition of tools in BERUFENET is very broad and covers approximately fourteen thousand tools. After selecting suitable tools (tools in the narrow sense, e. g. no documents, no production output/finished products), we use 5,919 of them. We also test a version with the full set of available work equipment for robustness checks. ^[9] However, the narrow set of work tools applied for the digital-tools index (dtox) presented in the remaining of this paper is the better choice to answer the research questions of this paper because it focuses only on those tools that are closely connected to the tasks performed. Another approach might be to count the digital tools as an absolute number and to rank the occupations in order of these numbers. Unfortunately, the documented tools differ widely between each individual occupation at BERUFENET. Hence, an occupation with many tools may also have many more digital tools than an occupation with less tools documented in BERUFENET. To avoid this “large-number-of-tools”-bias, we decided to apply the percentage of digital tools of the total number of tools.

Due to technical reasons, the tools data extract from BERUFENET is only available for 2016 so far (date of extract: June 13, 2016). Nevertheless, this cross-section data at the current edge facilitates calculating the share of digital tools—and even to differentiate between IT-aided (industry 2.0/3.0) and IT-integrated digital tools (industry 4.0) in 2016. However, because of the lack of data, we cannot state if this share has changed between 2011 and 2016. Therefore, the access to further years of tool data from BERUFENET is an important and promising target for future research. ^[10]

We divide the tools into three categories, as follows:

1. IT-aided tools are electronically based tools, such as computers, printers, and electronic machines (digital technology classes 1 and 2, see Table 1).

2. IT-integrated tools are electronically based AND are explicitly dedicated to an industry 4.0 or services 4.0 feature, such as 3D printers, machine learning software or mobile robot clusters (digital technology class 3, see Table 1).

3. Non-IT tools are not covered by categories 1 and 2. By definition, these tools comprise a very broad range of different tools.

Given the large number of potential tools, the identification of digital tools is based on a text mining procedure to identify digital tools. Adapting the text mining approach introduced by Janser (2018), we apply a comprehensive catalog of digital tool keywords and regular expression algorithms to identify those BERUFENET tools that are IT-aided or IT-integrated.

Table 2 shows the frequency of keywords and the results after the text mining with automatic coding. Overall, 279 key expressions were applied (IT-aided tools: 134, IT-integrated tools: 145), which led to 748 matches with tools of the BERUFENET tool catalog. Using these results in the occupations-tools matrix, we identified 2,402 occupations with (only) IT-aided tools, 370 occupations that have IT-integrated tools, whereas only remaining 191 occupations do not have any digital tools within their portfolio. The relatively small number of occupations with IT-integrated tools might be explained by the circumstance that, due to the editorial process of BERUFENET, there is some time lag between the emergence of the real labor market demand and the inclusion in the database. Another reason might be that, due to the flexibility of standard PC work places, some new digital tools are included in those tool descriptions referred to “PC work places” and, consequently, they are not marked as separate tools (e. g. cloud computing services, machine learning algorithms).

Based on the digital tools identified, we create an occupations-tools-matrix that allocates the number of digital tools to every single occupation and groups them by categories of IT-aided and IT-integrated tools. This matrix facilitates the calculation of the (unweighted) digital-tools index, dtox. The dtox describes the proportion of digital tools categories in the total sum of tools of a single occupation occ8d (8-digit level) in 2016.

where dtoxcocc8d is the digital tools index (category c) of individual occupation occ8d in 2016, ∑dtococc8d is the number of digital tools (category c) of occupation occ8d in 2016, and ∑toolsocc8d is the number of all tools of occupation occ8d in 2016. c represents the categories of digital tools: 1. IT-aided digital tools, 2. IT-integrated digital tools, and 0. Digital tools total (1+2). Occ8d stands for the 8-digit level of KldB2010. ^[11]

Administrative employment data are only available at higher aggregated levels, starting at the 5-digit level of the KldB2010. To link the dtox values to administrative employment data, we have to aggregate the dtox from the 8-digit level to the 5-digit level. For the transformation of dtoxcocc8d to dtoxcocc5d, we use a procedure similar to that applied by Dengler et al. (2014) and Janser (2018): the digital tools index of the 8-digit occupations is added up, and the total is divided by the number of 8-digit occupations within the 5-digit occupation, or as the following formula:

As the data of employees per occupation is only available at the 5-digit level, there is no way to apply an employment-weight at this stage of the process (see also Dengler et al. 2014). Therefore, we have to assume that the number of employees in individual 8-digit level occupations is equally distributed within the aggregate of the 5-digit level occupations (“equal distribution assumption”), as follows:

where empocc8d∈5d is the estimated number of employees per 8-digit level occupation within the occupational group (5-digit level), empocc5d denotes the total number of employees within the occupational group (5-digit level) and Nocc8d∈5d stands for the number of occupations at the 8-digit level within the occupational group (5-digit level). Using the occupational group of “Occupations in warehousing and logistics–skilled tasks” as an example, Table 3 shows how this aggregation procedure is implemented in the dtox data. ^[12]

Table 3:

Exemplary aggregation procedure for dtox index.

Individual Occupation 8-digit level of KldB 2010 (ID+Title)		Occupational type 5-digit level of KldB 2010 (ID+Title)	No. of employees 5-digit level	No. of employees 8-digit level	dtoxIT-aid 8-digit level		dtoxIT-aid 5-digit level (aggregate)	dtoxIT-int 8-digit level		dtoxIT-int 5-digit level (aggregate)
51312–100 Order Picker		51312 Occupations in warehousing and logistics–skilled tasks	384,142	unobserved ↓ equal distribution assumption ↓ 64,023.7 per occupation at 8-digit level	0.17		0.21	0.08		0.09
51312–107 Receiving Department Clerk					0.17			0.17
51312–109 Specialist - Logistics/ Materials Management					0.20			0.07
51312–110 Dispatcher - Warehouse					0.60			0.00
51,312–111 Specialist - Warehouse Logistics					0.07			0.14
51312–112 Warehouse Clerk					0.05			0.05

1. Note: Aggregation of the individual occupations at the 8-digit level to 5-digit levels (KldB 2010) and the respective aggregation of the dtox index exemplarily for “Occupations in warehousing and logistics–skilled tasks”. The Appendix (Table 11) contains a comprehensive list of occupational segments sorted by their aggregated dtox value. Moreover, Table 12 provides an overview of the dtox index for different requirement levels.

2. Source: BERUFENET 2016, employment statistics of the Federal Employment Agency, own calculations.

Furthermore, the BERUFENET is also the initial source of the tasks index introduced by Dengler et al. (2014). We use this index to identify, e. g. the share of routine- and non-routine jobs.

3.1 Empirical approach

As described above, our analyses focus on full-time working males and females staying within their establishment during the observation period. The aim of the analyses is to estimate the effects of investments by establishments in new digital technologies on the wages of workers. We specifically investigate which groups of workers are positively or negatively affected by the digital transformation occurring in recent years. In addition to qualification, sex, age or sector affiliation, we also consider the role of the tasks that workers perform in their jobs and the role of the work equipment (namely, the degree of digital work tools in occupations) they use during their work.

To estimate the effects of investments in new digital technologies, we classify the establishments as shown above into pioneers (these establishments invest in new digital technologies between 2011 and 2016), peloton establishments (these establishments invest in digital technologies to a small extent) and latecomers (these establishments do not invest in new digital technologies and do not use them in 2016). This information is captured by dummy variables for the peloton and for pioneers, which we include in Mincer-type wage growth regressions, and latecomers are used as the reference group. We address time varying establishment- and worker-characteristics by including a battery of control variables; all time invariant characteristics are removed through differencing. Formally, the estimated model is as follows:

where yiet denotes the log wage of individual i in establishment e in year t. Xiet contains time-varying individual and establishment-related (individual) characteristics, such as individual ages, the digital tools index, the tasks index, (log) establishment size, (log) establishments’ gross per-capita output and per-capita investments, and the shares of intermediate consumption, ^[16] foreigners, female workers, highly skilled workers, temporary workers, etc. at the establishment level. All time constant individual characteristics, such as unobserved ability, ambition, and motivation, are contained in μi. They are removed by our approach, as are the time constant establishment characteristics ϑe (such as the location of the establishment or sector affiliation). δt captures general time shocks, and εiet  represents erratic shocks. The effects of investments in new digital technologies at the establishment level are captured by the coefficient β.β1 of the dummy variable DPelotoniet and β2 of the dummy variable DPioneeriet capture the effects for being employed in a peloton or pioneer establishment relative to being employed in a latecomer establishment.

We estimate this wage equation for the aggregate as well as for different subgroups of workers (by sex, age, skill, sector, main tasks groups, digital tools categories and by interactions of sector and skill, sector and tasks, etc.) The results of these estimates give us an idea of which workers suffer or benefit from the digital transformation in terms of wages.

3.2 Estimation results

Table 5 shows the results of the individual fixed effects estimates. ^[17] Column 1 contains the results for the sample of all workers. The wage growth differential of working for a pioneer instead of a latecomer is 0.8 percentage points between the years 2011 and 2016. This effect is moderate but positive and is significantly different from zero at a 1 percent level. Hence, our result contradicts the literature that suggests negative wage effects of new technologies (for instance, Acemoglu/Restrepo 2017) and supports those papers that suggest positive effects on wages (for instance, Graetz/Michaels 2015). Note, however, that the mentioned studies investigate the effects of industrial robots on wages. In our view, this technology is not new because many of these industrial robots were introduced in the 1980s and 1990s. As a consequence, our results neither directly support nor contradict the literature because—to our knowledge—our study is the first that analyzes the wage effects of new digital technologies, such as big data, cloud computing systems, internet platforms, cyber-physical/embedded systems or the internet of things. Moreover, our study focuses on a specific group of directly affected workers, i. e. establishment stayers. We do not investigate the effects of new digital technologies on employment in this paper ^[18]; hence, possible selection effects could explain a part of the positive wage growth effects. This would be the case if pioneers lay off low-performance workers more often than latecomers. A glance into our selection process reveals, however, that the construction of our balanced panel of establishment stayers affects latecomers and pioneers in comparably the same manner, i. e. 67.3 % of pioneer workers and 68.4 % of latecomer workers survive this selection step. That could be understood as a hint that selection effects do not largely bias the presented results. Digging deeper, we investigate which groups of workers are predominantly affected by our selection. Columns 1 and 2 of Appendix Table 14 show mean wages and observation numbers of different skill groups in latecomer, peloton and pioneer establishments. For all 735 low-skilled latecomer workers (column 2), the mean wage in 2011 is €81.78 (column 1). The balancing of the sample decreases the number of workers to 475 workers (column 4) and increases the mean wage by 6 % to €86.82 (column 3). Here, the impact of the selection on wages and the number of low-skilled workers is slightly lower than in peloton and pioneer establishments (column 5 and column 6). For the other skill groups, however, the balancing has comparably the same effects. ^[19] Column 1 of Table 5 demonstrates that the wage growth effect is higher in peloton establishments compared with latecomer establishments. The effect amounts to 0.6 percentage points. Note that both effects—for peloton establishments as well as for pioneers—are highly robust with regard to inclusion or exclusion of further control variables (for instance, we additionally included controls for occupational changes on the individual as well as the establishment level).

Results by sex, skill and age: The remainder of Table 5 depicts the estimation results for specific worker groups. Columns 2 and 3 show that the wage growth effect of investments into new digital technologies is more pronounced for male workers than for female workers. For male workers, it amounts to 1.1 percentage points and is statistically highly significant. For female workers, it is −0.2 percentage points and is not statistically different from zero. It should be noted that the sample size is distinctly larger for men. Before we present our findings on the impact of digitization on wages for different skill groups, let us first summarize the previous results of other studies. According to Akerman et al. (2015), the access to broadband internet improves (worsens) the labor market outcomes and productivity of skilled (unskilled) workers. Dauth et al. (2017) find a negative impact of robots on individual earnings arising mainly for medium-skilled workers in machine-operating occupations, while high-skilled managers have a gain in earnings. Interestingly, we find the largest positive effect for low-skilled workers (3.6 percentage points, see column 4). For skilled workers, it is 1 percentage point (see column 5), and for high-skilled workers, it amounts to −0.7 percentage point but is not statistically significant (see column 6). ^[20] We interpret these results in such a way that it benefits low-skilled and skilled establishment stayers when establishments invest in new digital technologies.

For this analysis, however, we compare low-skilled stayers in pioneers with low-skilled stayers in latecomers and skilled stayers in pioneers with skilled stayers in latecomers. Hence, it does not necessarily mean that low-skilled and skilled workers benefit more from investments into new digital technologies than their high-skilled colleagues within the establishment. ^[21]

Regarding age effects, it can be seen from Table 6 that younger workers especially benefit from being employed in a pioneer compared to being employed in a latecomer. ^[22] The wage growth effect amounts to 2.6 percentage points. For middle age and older workers, the effect is only 0.6 percentage points to 0.7 percentage points. Since accumulation of establishment-specific and general human capital is especially important during the first years of the employment biography, this could be a hint that the accumulation of human capital benefits from the use of new technologies in establishments.

Tools and tasks: To investigate this hypothesis more deeply, we categorize our sample by the information on the work equipment (tools) and tasks of occupations. As described above, we are able to differentiate between IT-aided tools (dtox_IT-AID), i. e. tools such as computers, printers, and electronic machines, that are not explicitly dedicated to an industry 4.0 feature and IT-integrated tools (dtox_IT-INT), i. e. tools that are explicitly dedicated to an industry 4.0 or services 4.0 feature, such as 3D printers, machine learning software or mobile robots. We now assign the workforce to three categories using the dtox_IT-AID distribution in 2011. The average share of IT-aided work equipment is 29.4 % in 2011. The median is somewhat lower at 25.6 %. The category “low” comprises workers with a below the median share of dtox_IT-AID . The category “middle” comprises workers with a dtox_IT-AID share between the median and the 75th percentile of the distribution (this is at 49.8 %) and the category “high” comprises workers with a dtox_IT-AID share of the 75th percentile or higher.

The same categorization is done for dtox_IT-INT. Here, the median in 2011 is at 0.5 % and the 75th percentile of the distribution is at 3.5 percentage. We see that new digital work tools are still barely used. As discussed above, this might be explained by the circumstance that, due to the editorial process of BERUFENET, there is some time lag between the emergence of the real labor market demand and the inclusion of new working tools in the database. The left panel of Table 7 presents the results for the three dtox_IT-AID groups. It is at the first glance surprising that the wage growth effect is highest for individuals typically working with non-IT-aided tools (1.7 percentage points, see column 1). The wage growth effect is even negative (but only significant at a 10 percent level) for workers with a high share of IT-aided tools. This corresponds with estimates for different tasks groups (here, workers are classified with regard to the main task of the worker’s occupation) where the wage growth effect is negative (but not significant) for workers with a high share of analytical and interactive tasks and statistically significant and positive for workers often performing routine cognitive tasks (see Appendix Tables16).

Before shedding light on these unexpected results, we turn to the three dtox_IT-INT groups on the right panel of Table 7. Here, we see the largest wage growth effect for workers of the medium category (2.3 percentage points, see column 5). Although the effect is not significant for the highest category, it is highly significant for the intermediate one. Hence, it seems that the usage of 4.0 work tools has some beneficial effect on the wage growth of workers.

Sector-specific results: One could argue, that comparing, e. g. workers primarily using non-digital working tools between latecomers and pioneers could be misleading when pioneers are typically (e. g.) large, modern IT-establishments and latecomers are typically (e. g.) small construction establishments. In the former, it would be quite unusual that workers primarily use non-digital working tools, while this is entirely normal in the latter. Having this battery of control variables included in the fixed effects approach would mean, in the worst case, comparing apples with oranges. Therefore, we think that sector-specific estimates are better suited to understand the effects of new digital technology investments on wages. Table 8 shows that the wage growth effects for working for a pioneer vs. a latecomer are significantly positive in the sector aggregates knowledge intensive manufacturing (2.2 percentage points, see column 2; e. g. car manufacturers or machine manufacturers) and non-knowledge intensive services (4.3 percentage points, see column 3; e. g. wholesalers, logistics, restaurants). Among ICT establishments, the effect is also positive but is smaller and significantly different from zero at the 10 percentage level only (see column 5). By contrast, it is negative in knowledge intensive services (−1.8 percentage points, see column 4; e. g. banks, insurances, scientific services).

After further analyses, Table 9 shows the estimated coefficients of the treatment variable for the different skill groups within these sector aggregates. In both knowledge intensive manufacturing and non-knowledge intensive services, the wage growth effect is most pronounced for low-skilled and skilled workers and is not significant for high-skilled workers. This points to the fact that the positive effect detected for both sectors is actually driven by the effects for low-skilled and skilled persons. In knowledge intensive services, we observe that the negative wage growth effect can mostly be explained by the effect for high-skilled workers, i. e. it is negative (−4.8 percentage points) and statistically significant at the 1 percent level.

Repeating the dtox_IT-INT groups analyses separated by sectors (see Table 10), we observe significant effects only for the low and intermediate IT-integrated tools category in knowledge intensive manufacturing and the highest category in knowledge intensive services. Our interpretation is that the usage of these new digital working tools helps to explain the higher wage growth of the pioneers’ workers, especially within knowledge intensive services. By contrast, in knowledge intensive manufacturing, the positive effect is detected for workers who do not use IT-integrated working tools or who only use it to a small extent. Similar results are obtained for the usage of IT-aided working tools (by repeating the dtox_IT-AID groups analyses separated for sectors ^[23]). In terms of wages, it seems that the usage of work-equipment is polarized in the sense that people using non-digitalized and high-digitalized work equipment benefit from the establishments’ digital transformation, while this is not the case for workers using equipment with an intermediate share of digitalization. To recheck this finding, we change our point of view and focus on the employees in pioneers.

Who benefits within pioneers? To measure wage growth effects for workers within pioneers we interact the dummy variable that indicates the affiliation to a specific group (for instance skilled or high-skilled) with the year dummy for 2016. ^[24] Starting with differences between skill groups, it is worth noting that all effects are insignificant at a 5 percent level (in non-knowledge intensive manufacturing establishments, there is a positive effect being significantly different from zero at the 10 percent level: 1.4 percentage points for skilled workers relative to low-skilled workers; in knowledge intensive manufacturing and knowledge intensive services there are two negative effects for skilled workers relative to low-skilled workers that amounts to −1.9 percentage points and −2.5 percentage points, respectively). That means that low-skilled, skilled and high-skilled workers within pioneers have comparably the same wage growth over the observation period. The above-documented positive effect for low-skilled and skilled workers relative to the latecomer reference group is therefore less attributed to differing effects for skill groups within pioneers but is more attributed to differing wage growth rates in establishments without new digital technology investments (regarding latecomers, we can actually observe that the wage growth rates increase with the skill level of the employees).

Turning to the work tools again, we observe slightly positive effects in the amount of 1 percentage point (but statistically significantly different from zero at a ten percent level, only) for workers in pioneers if they operate with IT-aided tools and a smaller amount of IT-integrated tools. Our results show that working with many IT-integrated tools has no positive impact on employee wage growth. This could be due to learning effects that only increase productivity with a time delay when new tools are used. We will take a closer look at this effect in our future research.

How important are selection effects? As discussed above, one could argue that possible selection effects could explain a part of the positive wage growth effects. To take this into account, we abandon the restriction of the sample to employees who stayed with the same employer, which means that we include establishment leavers in our sample. ^[25] This increases our sample size from approximately 90,000 to 115,000 individuals. The disadvantage of this approach is that we do not know whether the person is switching to a pioneer, peloton or latecomer establishment. Nor do we know the reason for leaving the establishment. Due to this unobservable heterogeneity, we think that the restriction on establishment stayers provides a clearer picture. However, repeating our analyses with the extended sample, Appendix Table17 shows that for the aggregate as well as for specific subgroups, the positive effects are increasing. The wage growth differential of working or having worked, respectively, for a pioneer instead of a latecomer (see column 1) is now 1.7 percentage points between the years 2011 and 2016 (instead of 0.8 percentage point for establishment stayers). The effect for male workers (column 2) is now 2.1 percentage points (instead of 1.1 percentage points), and for female workers (column 3), it is 0.7 percentage points (instead of - 0.2 percentage points). Turning to the skill groups (columns 4–6), the estimates corroborate our finding that digitalization is beneficial, especially for low-skilled and skilled workers. The wage growth differential is 4.1 percentage points for low-skilled, 1.7 percentage points for skilled and - 0.1 percentage points (but statistically not significant) for high-skilled workers. Without wanting to overinterpret the results, there seems to be a positive effect that can be transferred from pioneers to other establishments. The reasons for this will be investigated in further analyses. The repetition of the other specifications documented above confirms the robustness of our results (these results are not included in the paper due to lack of space but are available from the authors upon request).