Faculty of Economics and Business
Msc International Economics and Business
July 2015
Digital empowerment: A cross-country study
on the effect of digital technologies on the
labor market in Europe
Supervisor: R. Ortega Argiles
Co-assessor: S. Brakman
ABSTRACT
This study examines the impact of digital empowerment -determined by availability, frequency of use and ability to use digital technologies- on the quality and quantity of the labor market in Europe. The quality is measured by labor productivity and the quantity by employment growth. The empirical analysis employs a panel data set including 27 EU countries for the time period 2005-2013. Our findings show that digital technologies have a negative effect on labor productivity and a positive effect on employment growth, which suggests that they enable individuals to improve job searching and job matching procedures but hinder productive work behavior. Thus, only enhancing the quantity aspect of the labor market in Europe.
1. INTRODUCTION
We live in a digital world. In the last two decades digital technologies have changed our lives on an economic, social and cultural level. It has even been adopted in public policies. In 2010, the European Commission introduced the “Europe 2020 Initiative” to advance the economy of the European Union. Part of this strategy is the ‘Digital Agenda for Europe’ that proposes to better exploit the potential of Information and Communication Technologies (ICTs) in order to foster innovation, economic growth and progress.
There has been a lot of research on the relationship between technology, more specifically ICT, and the economy. A large part concentrates on the effect of ICT on output growth, establishing a link between usage and diffusion or production of ICT to an improved economic performance in the US (Colecchia and Schreyer 2002; Jalava and Pohjula 2002). Other literature focuses on production growth as a key variable for economic growth by applying ICT as a new input within a production function analysis (Koutroumpis 2009; OECD 2004; van Reenen, Draca and Sadun 2007; Ventorini 2009).
Recently, the view has changed and a more comprehensive approach has been adopted. The research on ICT goes beyond the previous linear and technology-based relationship and concentrates on economic as well as social effects on a macro-level (Guerrieri and Bentivegna 2011; Evangelista, Guerrieri & Meliciani 2014). Three digitalization indicators were identified that impact key economic variables, namely ICT infrastructure, actual usage and empowerment. It was found that ICT usage and especially empowerment have a positive economic influence, when measuring for labor productivity, employment growth and GDP growth, while ICT infrastructure is merely a necessary pre-condition.
Therefore, this paper aims at exploring the effect of digitalization on a macroeconomic level. More specifically, we will explore the relationship between digital empowerment and the labor market in Europe. Thus, the research question is:
Does digital empowerment, and thus digital technologies have an effect on the labor market in Europe?
technologies, as well as the impact it might have on the quality and quantity of the European labor market. Similarly to Evangelist et al. (2014), the study investigates digital empowerment along three dimensions: (a) digital access in terms of availability to digital technologies, (b) digital usage that refers to the frequency of use of digital technologies, and (c) digital skills that refers to the ability to use digital technologies. Furthermore, the labor market is determinant by quantity and quality, i.e. labor productivity and employment respectively.
One could argue that the elected time period of nine years is too short, however taking into account how fast technologies change and countries and its people adopt new technologies, it is reasonable to assume that the impact of digital technologies evolves greatly even in a short amount of time. Especially in the last few years there has been tremendous change with the emergence of social media and mobile Internet. Today, the younger workforce grew up with digital technologies and probably uses them more vigorously. Additionally, in Europe there has been great change when it comes to digital empowerment due to the introduction of the ‘Digital Agenda for Europe’ in 2010, as mentioned before.
2. DIGITAL EMPOWERMENT
Digital empowerment comprises three facets of digital technologies. The availability determined by access to the Internet, the frequency determined by frequent use of the Internet and ability determined by digital skills. Thus, it is about whether people have access to digital technologies, how often they use them and how good they are in using them. It is believed that with digital technologies people gain new abilities and ways to participate and express themselves (Mäkinen, 2006).
The three indicators were chosen from variety digitalization indicators that have been found in the literature. Czernich (2014) has investigated whether broadband, here named digital access, has an influence on the unemployment rate. Hargittai (2004) also investigated the effect of Internet access and use. And Evangelista et al. (2014) determined three digitalization indicator, which are ICT infrastructure, actual usage and empowerment (in the sense of the ability to use digital technologies). We chose to depict the years 2006 and 2013 in the graphs below, because they best show the evolution of the indicators.
2.1. Digital access
Figure 1. Digital access across the EU 27 countries in 2006 (in blue) and 2013 (in red).
2.2. Digital usage
Secondly, digital usage is measured by the frequency of use, more specifically by the percentage of individuals that use the Internet at least once a week (including every day). Figure 2 shows that increasingly more people use the Internet frequently, on average it increased from 45 percent to 72 percent across the European 27 countries. As expected and similar to Internet access, Internet use varies greatly across the European 27 countries where the Western European and Scandinavian countries are ranking high (Luxembourg 0.93, the Netherlands 0.92 and Sweden 0.92) and the Southern and Eastern European countries are ranking relatively low (Romania 0.45 and Bulgaria 0.52).
2.3. Digital skills
Lastly, digital skills refer to the percentage of individuals that have carried out 1 or 2, 3 or 4 and 5 or 6 of the 6 Internet related activities. The six Internet related activities that were used by Eurostat to group the respondents are: (a) use of a search engine to find information; (b) send an e-mail with attached files; (c) post messages to chat rooms, newsgroups or any online discussion forum; (d) use the internet to make telephone calls; (e) use peer-to-peer file sharing for exchanging movies, music etc.; (f) create a web page.
Figure 3 depicts an accumulation of the three categories per country for the years 2006 and 2013. Overall, digital skills have improved over the years and go as high as 94 percent of individuals (in Denmark) that can carried out some of the six Internet related activities. At a closer look it can be noticed that not all three indicators have the same growth rate. In the case of Denmark, individuals that can carry out 1 or 2 activities have sharply dropped from 0.47 to 0.23 (Appendix figure 4), individuals that can carry out 3 or 4 activities have doubled from 0.27 to 0.50 (Appendix figure 5), and individual that can carry out 5 or 6 activities even tripled from 0.07 to 0.21 (Appendix figure 6). It suggests that many people have improved their digital skills over the years and moved from low proficiency to medium or high proficiency, and that therefore the overall growth arises from changes in higher proficiency.
Figure 3. Digital skills across the EU 27 countries in 2006 (in blue) and 2013 (in red).
3. LITERATURE REVIEW
3.1. Hypotheses
The term digital divide first appeared in the late 1990s and referred to the gap between those who have access to new technology and those who do not (US department of commerce, 1999). In the past it has been viewed as a one-dimensional concept that focused on the impact of ICT infrastructure or access to the Internet on economic growth. Factors such as the use of digital technologies or their adoption within society have been thus far ignored. It was found that investment in and production of ICT positively contributes to economic performance (Colecchia and Schreyer 2002, Jalava and Pohjula 2002). More recent studies also explore the effect of broadband infrastructure, i.e. internet access, on economic growth finding that countries with an increase in broadband penetration experience annual per-capita growth (Koutroumpis 2009, Czerich 2011). Other studies, however, suggest that IT infrastructure has peaked in 2006 (at least in Europe) assuming that the majority of people have access to the internet and that further investment in IT infrastructure will not yield substantial advantages (Guerrieri and Bentivegna 2011).
Since then the views on ICT and the digital divide have evolved and expanded to a more comprehensive approach that not only studies the economic effect, but spans to the field of sociology. Digital technologies play a significant role in social inclusion and exclusion, thus it cannot be studied on a mere technological or one-dimensional level. They contribute to cultural and social change, and in a larger sense to the development of societies. Therefore, careful examination is needed on a multi-dimensional level (Mossberger 2008). The first steps in that direction were taken by extending the measure of internet access by various indicators such as place of work, home, family or friends and investigate whether that leads to unequal opportunities. It was found that most people in developed countries, and especially the US, have access to the Internet, but that it was not a predictor for inequalities. It spans further than Internet access alone. Studies show that digital inequality is influenced by the frequency of use of digital technologies and that proficiency was mostly obtained though learning by doing. Suggesting that usage of digital technology might be advantageous to overcome digital exclusion. Thus, the key issue wasn’t just the unequal access but the frequency of use and unequal abilities to use digital technologies (Hargittai 2004, van Dijk 2005).
have high access to and use of ICT at school and at home also report higher confidence in their operational ICT skills, in their use of social media and in their ability to use the internet safely and responsibly, as well as more positive opinions about ICT’s impact on their learning. The same study also suggests that teachers are interested in gaining ICT skills on their own as 70 percent of students is taught by teachers who engage in personal learning in ICT in their own time.
Furthermore, evidence shows that digital exclusion can be viewed equally to social exclusion and thus has an impact on an individual level as well as social level and even further a micro as well as meso and macro level (Digital Inclusion Team 2007).
On the individual level, a trend has emerged in which digital technologies have formed an important part in people’s social life. From Wikipedia to Facebook to Twitter to Whatsapp, everybody is sharing information and exchanging knowledge with each other. However, these digital technologies are not only used in a personal context they also increasingly appear on a social level.
In terms of education, digital technologies are being increasingly incorporated in form of blended learning and MOOCs. The Economist (2014) even claims that the technical evolution that comes with the emergence of MOOCs challenges the business model of higher education institutions. However, there is a distinction to be made between technologies for education and for learners, the former creating an accountability culture meaning the collecting of data on learners for educators, the latter creating an participatory culture meaning increased social interactions by participation in new media. Learners therefore use, create and share content and strategies for engagement and thus use technology to create their own environment for learning. Some believe that digital technologies can reshape learning in and out of schools (Halverson, R. & Shapiro, R. B 2012).
Recent studies also started to focus on the employment level. In terms of productivity, the majority of the studies indicate a positive effect of ICT (Cardona, Kretschmer and Strobel, 2013). Other research suggest that the use of digital technologies effect labor productivity by encouraging part-time work arrangements through better job matching and job flexibility opportunities. Additionally, employment growth and employment participation rates are positively affected due to the inclusion of disadvantaged groups, e.g. women and long term unemployed, through better job searching measures (Evangelista et al., 2014).
convergence, deregulation and differentiation (Golding 2000, Golding and Murdock 2001). Generally, the view is for government intervention, the importance of digital literacy and providing meaningful content. One study also suggests that the digital skills of an individual heavily influence these inequalities. Digital skills are distinguished between lower proficiency (such as knowing how to use the computer and surf the net) and higher proficiency (such as being able to obtain results of various kinds) (Warschauer 2003).
Thus, the first hypothesis is concerned with the influence of digital empowerment on a macro economic level, more specifically with the access, use and skills needed for digital technologies to have an effect on the labor market.
H1a: An increase in digital empowerment determined by digital access will positively affect the labor market in terms of labor productivity and employment growth.
H1b: An increase in digital empowerment determined by digital usage will positively affect the labor market in terms of labor productivity and employment growth.
H1b: An increase in digital empowerment determined by digital skills will positively affect the labor market in terms of labor productivity and employment growth.
Furthermore, the cross effect is studied to see whether education has any implications when looking at digital empowerment. The second hypothesis thus checks the effect of education on digital access and digital usage.
H2a: The impact of digital access on the labor market is indirectly and positively affected by education.
H2b: The impact of digital use on the labor market is indirectly and positively affected by education.
Lastly, we will check for non-linearity by applying an additional variable to the estimations of digital access and digital use, as there seems to be in between variability as discussed in section 2.
H3a: The effect of digital access on the labor market is non linear.
3.2. Conceptual Model
In the following the conceptual model is depicted and explained. It reflects the hypotheses stated above and includes the dependent and independent variables, as well as the cross effect.
As can be seen in the model above and represented by hypotheses 1a-c, digital empowerment is divided into three independent variables, namely digital access, digital usage and digital skills. All variables are assumed to have a positive effect on the dependent variable labor market. Not depicted is the fact that the dependent variable consists of two parts, which are labor productivity for the quality of the labor market and employment growth for the quantity of the labor market.
The cross effect that is represented by hypotheses 2a-b is also depicted in the conceptual model as the level of education that is assumed to indirectly and positively influence the effect of digital access and digital usage on the labor market. Not included are the hypotheses on non-linearity.
H1a!(+)!
H1b!(+)!
H1c!(+)!
Digital Access
Digital Usage
Labor
Market
Digital Skills
Level of Education H2a!!
4. EMPIRICAL ANALYSIS
4.1. Sample
This analysis is an attempt to assess whether and how digital access, digital usage and digital skills are influencing the labor market according to macro economic performance variables, namely labor productivity and employment growth.
All data used in this study is from Eurostat and comprise the EU 27 countries including a EU27 average. The EU27 countries comprise all countries that joined the European Union by 2005 including Bulgaria and Romania that only joined in 2007 but participated in the Eurostat data collection early on. The data was collected over a time period of 2005-2013 for a total of 252 observations. The panel data is unbalanced due to missing data in the set. Particularly for the variables skill, skill2 and skill3 there is no data available for the years 2008, 2009 and 2012 and the variable access is missing data for the year 2012. We recognize that the missing data can lead to a misrepresentation of the sample and distort the conclusions drawn from the population, especially for the digital skill variables in our case. However, we will take full account of all information available and introduce control variables in the tested equations to support the reliability of the model.
4.2. Variables
4.2.1. Labor market (dependent variable)
In this study labor market is defined by two variables that are included as a dependent variable in the regressions. The first variable is labor productivity, which is calculated as the real output per unit of labor input. It provides a picture of productivity developments in the economy across countries and years and represents the quality of the labor market.
The second variable is employment growth calculated as a year-on-year comparison in the total number of persons employed across the EU27 countries. It depicts the increase and decrease of total workforce across countries and years and therefore, represents the quantity of the labor market.
4.2.2. Digital empowerment (independent variables)
Internet at home. Digital usage stands for the frequency of use of digital technologies and is calculated by the percentage of individuals that use the Internet at least once per week. And digital skills represents the ability to use digital technologies. It is split up into three indicators: percentage of individuals that can carry out 1 or 2 of the 6 Internet related activities, percentage of individuals that can carry out 3 or 4 of the 6 Internet related activities, and percentage of individuals that can carry out 5 or 6 of the 6 Internet related activities. The Internet related activities that are measured are as follows: (a) use of a search engine to find information; (b) send an e-mail with attached files; (c) post messages to chat rooms, newsgroups or any online discussion forum; (d) use the internet to make telephone calls; (e) use peer-to-peer file sharing for exchanging movies, music etc.; (f) create a web page (Eurostat, 2015). The evolution of these variables was further explored in section 2.
4.2.3. Control variables
Throughout the literature there is a large number of determinants discussed for labor productivity and employment growth. However, there is no consensus reached as to which variables are most appropriate. Therefore, we will apply control variables for robustness that have been applied in similar studies (Evangelista, Guerrieri and Meliciani 2014, Czernich 2011, Guerrieri and Bentivegna 2011). They include human capital, investment, labor cost, demand and population growth.
Human capital can be measured by a variety of indicators, however, similar to previous studies in this field we measure it by education, i.e. percentage of individuals with a tertiary education and percentage of individuals with at least a secondary education. Investment refers to the share of gross fixed capital formation over GDP. Labor cost is the rate of growth of labor costs (measured as total remuneration per employee in constant prices). Demand refers to the rate of growth of demand (measured as the rate of growth of GDP at constant prices). And population stands for the rate of growth of total population.
4.3. The model
The following models will be used to test our first hypothesis:
(1) Labor Marketit = a0 + ß1 accessit-1 + ß2 useit-1 + ß3 skillit-1 +
ß4 educit + ß5 invit + ß6 lcit + ß7 gdpit + ß8 popit + eit
- where Labor Market refers to the rate of growth of labor productivity (lprod) defined by the change in percentage of real labor productivity per hour worked in country i at time t, and the rate of growth of employment (emp);
- educ is the education that consists of the percentage of population with tertiary education; - inv accounts for investment, which is the share of gross fixed capital formation over GDP; - access, use and skill are the yearly changes of digital empowerment that are measured by
percentage of individuals over total population with internet access at home, percentage of individuals over total population that use the Internet at least once a week, and the percentage over total population of individuals that can carry out 1 or 2, 3 or 4 and 5 or 6 Internet related activities, respectively.
Equations (1) reflects that the dependent variable is in fact determined by two variables, namely labor productivity to test for the quality and employment to test for the quantity of the labor market. The model includes all other variables discussed previously in section 4.2. Furthermore, the equation shows that all independent variables are lagged by one period to account for endogeneity as potential problems may occur when independent variables correlate with the error term (Hill, Griffith and Lim, 2012). The lags are further justified by theoretical arguments. Earlier it is mentioned that there appears to be a substantial change in the digital empowerment variables due to policy interventions, however this change may take some time to come into effect. By using lags we ensure that some time can elapse before digital empowerment shows its full effect.
In order to test Hypothesis 2, the following equations will be estimated: (2) Labour Marketit = a0 + ß1 accessit-1 + ß2 (access*educ)it-1 +
ß3 educit + ß4 invit + ß5 lcit + ß6 gdpit + ß7 popit + eit
ß3 educit + ß4 invit + ß5 lcit + ß6 gdpit + ß7 popit + eit
As depicted, in equation (2) and (3) two interaction terms are introduced. The cross effect is measured between access and tertiary education denoted by access*educ, as well as use and tertiary education denoted by use*educ. With this model we assess how the results of equation (1) will generalize to an independent data set and thus serves as a robustness check. For Hypothesis 3 the following will be estimated:
(4) Labour Marketit = a0 + ß1 accessit-1 + ß2 (access2)it-1 +
ß3 educit + ß4 invit + ß5 lcit + ß6 gdpit + ß7 popit + eit
(5) Labour Marketit = a0 + ß1 useit-1 + ß2 (use2)it-1 +
ß3 educit + ß4 invit + ß5 lcit + ß6 gdpit + ß7 popit + eit
To perform another robustness and sensitivity check, I will check for non-linearity in access and use. These two variables were chosen because there seems to be in between variability as discussed in section 2.
4.4. Data
The summary statistics (table 1) shows a great variance between the observations, especially for skill and access, this is due to the missing data that is mentioned above. Standard deviation is within the min and max, and does not deviate to far from the mean, which suggests that there are no substantial outliers that have to be considered. The correlation matrix, however, shows potential multicollinearity between the independent variables access, use and skill. I will account for that by estimating the independent variables in separate specifications.
Table 1. Descriptive Statistics
Dependent Variable Number of obs. Mean Std. Deviation Min Max
Labor Productivity (lprod) 252 1.1663 2.9452 -8.6 15.3
Employment (emp) 252 0.2437 2.5658 -14.3 6.5
Independent Variable
Digital access (access) 221 54.7421 20.3729 11 93
Digital usage (use) 249 59.5100 17.8829 18 93
Digital skills (skill2) 160 26.3938 9.8217 5 50
Digital skills (skill3) 160 9.8688 5.8880 1 32
Control Variable
Tertiary education (educ) 252 22.5254 6.9395 9.1 36.3
Secondary education (educ2) 252 47.1441 12.9166 16.3 72.2 Investment (inv) 251 20.9649 4.3009 10.6 36.1 Labor cost (lc) 234 4.1517 5.3657 -9.5 30.8 Demand (gdp) 252 1.4675 4.0626 -17.7 11 Population (pop) 252 47.1440 12.9166 16.3 72.2
Note: Data are of lagged independent variables.
Table 2. Correlations of all variables
Dependent Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 1. Labor Productivity (lprod) - 0.05 -0.32 -0.28 -0.16 -0.27 -0.11 0.03 0.37 0.47 0.52 0.77 -0.46 2. Employment (emp) 0.05 - 0.03 0.06 0.17 -0.04 0.05 0.08 0.08 0.31 0.24 0.61 0.37 Independent Variable
3. Digital access (access) -0.32 0.03 - 0.99 0.51 0.91 0.55 -0.01 -0.04 -0.44 -0.37 -0.26 0.14
4. Digital usage (use) -0.28 0.06 0.99 - 0.53 0.92 0.56 0.04 -0.00 -0.40 -0.30 -0.21 0.12
5. Digital skills (skill) -0.16 0.17 0.51 0.53 - 0.27 -0.25 0.07 0.03 -0.23 -0.26 -0.05 0.30
6. Digital skills (skill2) -0.27 -0.04 0.91 0.92 0.27 - 0.58 -0.05 0.01 -0.37 -0.28 -0.26 0.03
7. Digital skills (skill3) -0.11 0.05 0.55 0.56 -0.25 0.58 - 0.02 0.07 -0.11 -0.05 -0.05 -0.12
Control Variable 8. Tertiary education (educ) 0.03 0.08 -0.01 0.04 0.07 -0.05 0.02 - -0.10 0.11 0.19 0.06 0.04 9. Secondary education (educ2) 0.37 0.08 -0.04 -0.00 0.03 0.01 0.07 -0.10 - 0.29 0.38 0.36 -0.40 10. Investment (inv) 0.47 0.31 -0.44 -0.40 -0.23 -0.37 -0.11 0.11 0.29 - 0.76 0.59 -0.20 11. Labor cost (lc) 0.52 0.24 -0.37 -0.30 -0.26 -0.28 -0.05 0.19 0.38 0.76 - 0.56 -0.37 12. Demand (gdp) 0.77 0.61 -0.26 -0.21 -0.05 -0.26 -0.05 0.06 0.36 0.59 0.56 - -0.16 13. Population (pop) -0.46 0.37 0.14 0.12 0.30 0.03 -0.12 0.04 -0.40 -0.20 -0.37 -0.16 -
Note: Data are of lagged independent variables.
distributed. However, even with non-normality the coefficients estimated using POLS and random effects could be the best linear unbiased estimators as long as the sample is large enough. Thus, we need to be cautious about the results and conclusions drawn, as we will proceed to use the sample discussed above to test our hypotheses.
Furthermore, the Breusch-Pagan test is applied to check for heterogeneity. It shows that the variances for all observations are not the same meaning that heteroskedasticity exists for both dependent variables. This could implicate that our model is no longer best or that the least squares standard errors are incorrect, which invalidates interval estimates and hypothesis test (Hill, Griffiths and Lim, 2012). Using cluster robust standard errors in the estimations can solve this problem.
Additionally, dummy variables for each year (except one) were included in the estimation and the explanatory variables were lagged to reflect their effect in the previous time period, so that a change in digital empowerment now will have an impact on the labor market now and in future periods as it takes time for the effect of a change in digital empowerment to fully work its way through the economy.
5. EMPIRICAL RESULTS
Table 3 and 4 report the results of the estimations of equation (1) using pooled OLS with cluster robust standard errors and the random effects model to account for random individual differences. The results depict the impact of digital access, digital use and digital skills on the labor market determined by labor productivity and employment.
5.1. Labor Productivity
Table 3 depicts all results for labor productivity. It shows that pooled OLS seem to be the best model for predicting the dependent variable. The explanatory power of the model lies at 76.7% for access, 75.3% for use, and 73.7-76.6% for skill. The F-test further suggests that it is explained significantly in all cases.
Contrary to hypothesis 1a, the results show that digital empowerment by all means has a negative effect on labor productivity at a significance level of 1%, except of skill that does not seems to be a significant indicator, which could mean that being able to carry out 1 or 2 of the 6 Internet related activities is too simple or easy a task to have an impact labor productivity. The regressions show that if access and use increase by 1% then labor productivity is reduced by 2.7%. For skill2 and skill3 the decrease in labor productivity is even higher with 6.6% and 8.0% respectively. So far, this seems to be the first empirical study to find statistically significant negative associations between digital empowerment and labor productivity. It seems that contrary to previous studies, opportunities for job matching and job flexibility do not arise in the investigated countries. A possible explanation could be that even though we assume that individuals with the availability to digital technologies and frequent use of them would lead to familiarity to new technologies and make them more productive, it in fact distracts them from their work and hinders productivity. Similarly, the negative correlation of digital skills might suggest that individuals with high proficiency in digital technologies do not apply them in their work or might not be able to exploit them sufficiently at their job and therefore reduce their productivity by applying their skills for personal, not-job related matters.
access and use, and suggests that people with a higher degree have a positive influence on labor productivity.
Thus, we can conclude that a highly developed country with a growing economy and highly educated people does not appear to be in need of an increase in digital empowerment to improve the labor productivity, it rather hinders the quality of the labor market.
5.2. Employment
From table 4 it can be concluded that the pooled OLS is the best model for access and use with an explanatory power of 75.2% and 74.6% respectively and indicated significantly by the F-test. The regressions for the three skill variables do not show a strong model were the explanatory power of the R2 is 63.2% for skill, 69.0% for skill2 and 67.1% for skill3.
In accordance with Hypothesis 1b, digital empowerment has a positive effect the quantity aspect of the labor market and is highly significant. Equal to the results of labor productivity, the regression for skill does not appear to be significant, suggesting that basic digital skills are not advantageous to employment as they might be too simple. However, for digital access and use the results show that a 1% increase can lead to higher employment by 2.9%. In the case of skill2 and skill3, employment would increase by 7.9% and 8.7% respectively. Similar to previous studies this effect could be explained by better job searching procedures, suggesting that having access to digital technologies and being able to use them can help individuals in their finding employment.
The results for tertiary education show that they are only significant in the case of access and use, similar to labor productivity, but with a negative correlation. It appears that higher educated individuals could have 4.6% less chance to find employment. A possible explanation could be that there are not enough jobs for highly educated people. Also different to labor productivity is the effect of population on employment, which is positive and highly significant and indicates that a larger population leads to a higher rate of growth of employment. The last significant indicator, GDP, expectedly also has a positive correlation to employment.
5.3. Robustness checks
Robustness checks were performed to examine how certain coefficient estimates of the model are when the regression specification is modified by adding variables that account for cross effects and non-linearity.
Table 4 present the results for the regressions including the cross effects, i.e. access multiplied by tertiary education and use multiplied by tertiary education. It shows that the results are equal to the previous results in table 3 and 4, except of a slightly lower significance at a 5% level for digital access and tertiary education. The cross effect is not significant. Thus, the coefficients seem to be plausible and robust meaning that the model shows structural validity. Additionally, I have checked non-linearity in access and use but did not find any significant effect as depicted in table 5. Similar to the cross effect, the results are the same as the ones from equation (1). While the effect can be found in the case of the variable skill, as can be seen in tables 3 and 4, there is no significant effect for the lower levels.
5.3.1. Sensitivity Checks
6. CONCLUSION
This study was an attempt to investigate the impact of digital technologies on the labor market for a panel of 27 EU countries in the time period 2005-2013. It was argued that digitalization measured by Internet access, Internet use and digital skills have a positive impact on macro economic labor variables. These variables where labor productivity and employment growth. By employing a panel data regression analysis we showed that digital technologies decreases labor productivity and increase employment growth. All results are robust for the investigated countries.
During the empirical analysis, some interesting results have emerged. Contrary to the suggested hypothesis, digital technologies have a negative effect on labor productivity. A reason might be that digital skills are not being used at work or that employees are not sufficiently equipped at work to exploit opportunities from digital technology. For employment growth, the empirical analysis shows that digital technologies have a positive and significant impact in terms of Internet access, Internet use and digital skills (excluding one of the skill variables). It is suggested that due to digital technologies, there are more education and job searching opportunities as well as a higher job flexibility possible.
This study contributes the existing literature on digitalization and its impact on macroeconomic performance variables, particularly on labor productivity and employment. Some recent previous studies found evidence on the positive effects of various ICT dimensions on labor productivity and employment (Cardona, Kretschmar and Strobel, 2013; Evangelista et al., 2014). The added value of our study is the multi-dimensional view on digital empowerment and the use of digital technology. Furthermore, it found a negative effect on labor productivity contrary to previous literature.
REFERENCES
Cardona, M., Kretschmer, T. and Strobel, T. (2013). ICT and productivity: conclusions from the empirical literature. Information Economics and Policy, 25(3), pp.109-125.
Colecchia, A. and Schreyer, P. (2002). ICT Investment and Economic Growth in the 1990s: Is the United States a Unique Case?. Review of Economic Dynamics, 5(2), pp.408-442. Czernich, N. (2014). Does broadband internet reduce the unemployment rate? Evidence for Germany. Information Economics and Policy, 29, pp.32-45.
Czernich, N., Falck, O., Kretschmer, T. and Woessmann, L. (2011). Broadband Infrastructure and Economic Growth. The Economic Journal, 121(552), pp.505-532.
Digital Inclusion Team (2007). The digital inclusion landscape in England. Delivering social impact through information and communication technology. [online] Available at:
http://digitalinclusion.pbworks.com/w/file/38362339/Delivery+Innovation+Team+Final+Rep ort.pdf
Dijk, J. (2005). The deepening divide. Thousand Oaks: Sage Publications.
European Commission, (2013). Survey of Schools: ICT in Education. Belgium: European Union.
Evangelista, R., Guerrieri, P. and Meliciani, V. (2014). The economic impact of digital technologies in Europe. Economics of Innovation and New Technology, 23(8), pp.802-824. Guerrieri, P. and Bentivegna, S. (2011). The Economic Impact of Digital Technologies. Cheltenham: Edward Elgar Pub.
Halverson, R., and Shapiro, R. B.(2012). Technologies for Education and Technologies for Learners: How Information Technologies Are (and Should Be) Changing Schools. [online] Available at:
Hargittai, E. (2004). Internet Access and Use in Context. New Media & Society, 6(1), pp.137-143.
Hill, R., Griffiths, W. and Lim, G. (2012). Principles of econometrics. Hoboken, NJ: Wiley. Jalava, J. and Pohjola, M. (2002). Economic growth in the New Economy: evidence from advanced economies. Information Economics and Policy, 14(2), pp.189-210.
Koutroumpis, P. (2009). The economic impact of broadband on growth: A simultaneous approach. Telecommunications Policy, 33(9), pp.471-485.
Mäkinen, M. (2006). Digital Empowerment as a Process for Enhancing Citizens' Participation. elea, 3(3), p.381.
Mossberger, K., Tolbert, C. and McNeal, R. (2008). Digital citizenship. Cambridge, Mass.: MIT Press.
OECD, (2004). The Economic Impact of ICT: Measurement, Evidence and Implications. Paris: OECD.
van Reenen, J., Draca, M. and Sadun, R. (2007). Productivity and ICTs: A Review of the Evidence. The Oxford Handbook of Information and Communication Technologies, pp.100– 147. Oxford: Oxford University Press.
The Economist, (2014). The digital degree. [online] Available at:
http://www.economist.com/news/briefing/21605899-staid-higher-education-business-about-experience-welcome-earthquake-digital [Accessed 19 May 2015].
APPENDIX
Figure 4. Internet skills: 1 or 2 of the 6 Internet related activities.
Figure 5. Internet skills: 3 or 4 of the 6 Internet related activities.
