Proving the Value of Digital Skills Training

Huawei recently wrapped up the grand final of its 9th Global ICT Competition, here in Shenzhen. The tournament attracted teams from over 100 countries across more than 2,000 colleges and universities. A huge total of 210,000 teachers and students took part in the competition at some point. After national and regional competitions, 179 teams from 48 countries successfully advanced to the global finale.

The Huawei ICT Competition is open to students that attend Huawei’s ICT academies through higher education institutions around the world. The aspiration of our ICT Competition is to provide students with an additional opportunity to test and advance their academic and vocational ICT knowledge and promote solutions to real-world problems. As well as improve upon their abilities to innovate and invent, using established and advanced digital technologies. 

The need to improve digital skills across the labour force is of paramount importance. It may for example be one of the main (and easier) ways that countries can revive what has been an almost uniform secular decline in labour productivity growth.

The OECD for example, have undertaken a lot of activity investigating this topic. They stress that “a lack of basic digital skills and access can have a huge negative impact on individuals’ lives, leading to poorer health outcomes and a lower life expectancy, increased loneliness and social isolation, and less access to jobs and education.”

The ITU suggest that job seekers with digital skills generally have greater success in finding employment. In Europe, the ITU assert that nine out of ten future jobs will require digital skills. In sub-Saharan Africa, over 200 million jobs will require digital skills by 2030, creating the need for almost 600 million training opportunities.
In support of low- and-middle-income country governments in Africa, the World Bank published their Methodological Guidebook V 2.0 for Preparing Digital Skills Country Action Plans in 2021. The guidebook provides detailed suggestions, plans and actions that can set digital skills attainment goals and measurable targets for ten-year national roadmaps.

The ability to better measure progress on digital skills and to better benchmark against country peers, has also been an aim of the G20. In 2022, the Indonesia G20 released their toolkit for governments to estimate and record changes in digital skills and digital literacy.

Empirical study

Given then the critical importance of digital skills training and development, it is somewhat surprising to find that there are few cross-country empirical studies that try to measure the economic impact of an improvement in digital skills. This is likely due to a lack of cross-country comparable data and the usual problems of measuring educational attainment (an issue the G20 is trying to address). 

Fortunately the World Economic Forum’s (WEF) Global Competitiveness Index does include a metric as a component of its overall index that is labelled as “Digital skills among (the) active population”. The metric is standardized across many countries for comparability and has a long timeline.

In the spirit of investigation, I set out to test whether changes in digital skills levels across countries has had a statistically significant impact on changes in labour productivity. 

The econometric model in this article follows the neoclassical production function. Assuming an augmented Cobb-Douglas production function gives the following equation:

Yi,t = TFPi,t KβKi,tLβLi,t    (1)                            

where Yi,t is real GDP, Ki,t is capital, Li,t is labor input (employment/persons engaged) and TFPi,t is Hicks-neutral total factor productivity, all for country i at time t.

By taking natural logarithms of equation (1) the following equation is derived:

lnYi,t = βK  lnKi,t  + βL lnLi,t + lnTFPi,t      (2)

where β represents the output elasticity of each input. Dividing by labor (L) we get:

ln(Y/L)i,t = βK  ln(K/L)i,t  + lnTFPi,t       (3)

Thus, labor productivity (Y/L) is a function of the capital labor ratio (K/L) and TFP which is measured as a residual. In the econometric specification used here, I also add Digital Skills as our variable of interest. I also include two control variables; a measure of the overall quality of human capital and a measure of digital infrastructure development, through the national adoption rate of fixed broadband.

ln(Y/L)i,t = β1 + β2 ln(Digital_Skills)i,t + β3ln(K/L)i,t + β4ln(HK)i,t + β5ln(FBB100)i,t + δTt + (αi, + εi,t)    (4)

Tt is a set of year dummy variables that help us control for common shocks across all countries over time.  αi is a set of unobserved country-specific effects that help us control for country specific factors.  εi,t  is the error term.

Data

Data on GDP, employment and capital are taken from the Penn World Tables 10.1 (Feenstra et al. 2015). Real GDP and capital stock are at constant 2017 national prices (million US$ 2017). Data on employment is measured as number of persons engaged. Labour productivity is thus measured as GDP per person engaged. The measure of human capital was retrieved from the human capital index as provided in the Penn World Tables. The index is based on the average years of schooling and an assumed rate of returns to education around the world. Data on digital development are taken from the ITU as fixed broadband subscriptions per 100 people.

Table 1: Descriptive statistics (2014-2019)

Variables	Mean	St. Dev.	Min	Max	No. obs
Labour productivity (Real GDP at constant 2017 national prices (in mil. 2017 US$ per person engaged)	23,554	17,390	570	79,853	540
Digital skills among active population	47.4	9.6	20.1	73.1	540
Capital labour ratio (in 2017 US$ per person engaged)	87,156	76,191	2,816	399,324	540
Human capital index	2.3	0.6	1.2	3.6	540
Fixed broadband subscriptions (per 100 people)	5.6	6.9	0.001	38.0	540

In total, a balanced panel of 90 low-and-middle-income countries for the period 2014-2019 are included in the regression analysis.

Results

Table 2: Model results Digital Skills and Labour Productivity (2014-2019)

Dependent variable: log of labour productivity (real GDP per person engaged)
	Time fixed effects	Country fixed effects	Time and country fixed effects
Log of digital skills index	0.386 (0.292)	0.143 (0.076)*	0.129 (0.077)*
Log of real capital per person engaged	0.628 (0.058)***	0.242 (0.048)***	0.242 (0.051)***
Log of human capital index	0.079 (0.280)	0.432 (0.242)*	-0.098 (0.354)
Log of fixed broadband subscriptions (per 100 people)	0.051 (0.036)	0.020 (0.008)**	0.009 (009)
Time fixed effects?	Yes	No	Yes
Country fixed effects?	No	Yes	Yes
Observations	540	540	540
Countries	90	90	90
Time periods	6	6	6
R2	0.80	0.36	0.32

Note: The estimates are based on pooled OLS and fixed effects. Cluster robust standard errors are presented in parenthesis. ***, **, * indicate statistical significance at the 1%, 5% and 10% levels, respectively.

When we control for both time and country effects, our measure of digital skills has a positive coefficient and is statistically significant at the 10% level. This is across a large group of quite diverse countries and does seem to hint that improvements in digital skills has had a positive impact on increases in labour productivity, ceteris paribus.

Some caution here. More robustness tests are needed to better validate the results (which is beyond the scope of this blog article). We are also limited by only having the WEF data set on digital skills (which in itself is based on the executive survey results). The economic data from the Penn World Tables only go to 2019 too and it would be useful to bring the time series up to more recent times.

But the initial results do appear promising and fit what feels intuitively right. Further research is definitely called for, so that we can better estimate the actual benefit-to-cost impacts of funding more extensive and better training in digital skills across low-and-middle income countries.

Learn more about Huawei's commimtment to expanding digital skills through its TECH4ALL initiative.

 

---

Disclaimer: Any views and/or opinions expressed in this post by individual authors or contributors are their personal views and/or opinions and do not necessarily reflect the views and/or opinions of Huawei Technologies.