CC BY
130
2019
ВЕСТНИК САНКТ-ПЕТЕРБУРГСКОГО УНИВЕРСИТЕТА ЭКОНОМИКА
Т. 35. Вып. 4
ИННОВАЦИИ И ЦИФРОВАЯ ЭКОНОМИКА
JEL: C61; E66; F68; H70
Comparative Analysis of the Development of the Digital Economy in Russia and EU Measured with DEA and Using Dimensions of DESI
Z. Banhidi, I. Dobos, A. Nemeslaki
Budapest University of Technology and Economics, 2, Magyar tudosok krt., Budapest, 1117, Hungary
For citation: Banhidi Z., Dobos I., Nemeslaki A. (2019) Comparative Analysis of the Development of the Digital Economy in Russia and EU Measured with DEA and Using Dimensions of DESI. St Petersburg University Journal of Economic Studies, vol. 35, iss. 4, pp. 588-605. https://doi.org/10.21638/spbu05.2019.405
The aim of this paper is to compare the development of the digital economy in Russia with that of the 28 countries of European Union (EU). Data were compiled from the European Commission's International Digital Economy and Society Index (I-DESI 2018) database. After providing a brief overview of various alternative ways to measure the impact of information and communications technologies (ICT), we examine the most important features, advantages, and drawbacks of this database. We then describe the structure of our dataset and proceed with the analysis of the digital competitiveness of Russia and the EU-28. Our main research questions are concerned with the robustness of the EU data supply and the stability of its ranking. For this, we use the data envelopment analysis (DEA) method and the one-dimensional version of multidimensional scaling, which can also be employed for ranking issues. In addition to the conventional DEA method, we also investigate the viability of common-weights DEA models. We compare the results obtained to answer our questions. In evaluating the results, we also discern how data from Russia matches EU data on the digital economy. The comparison suggests that methods used in our study provide a similar solution, but the ranking of a few countries (including Russia) show wider variation.
Keywords: DESI index, digital public administration, innovation, data envelopment analysis, multidimensional scaling, ranking.
© Санкт-Петербургский государственный университет, 2019
Introduction
The International Digital Economy and Society Index (I-DESI) was designed to provide "an overall assessment of where the EU stands, compared to non-EU economies, in its progress towards a digital society and economy"1. First published in 2016, it aims to "mirror and extend" the results of the European Commission's original (EU-only) Digital Economy and Society Index (DESI) by "finding indicators that measure similar variables for non-EU countries", including Russia. Both of these are composite indices that combine several individual indicators and use similar (but not identical) weighting systems to rank each country based on its digital performance with the aim to benchmarking the development of the digital economy and society. They measure performance in five principal dimensions or policy areas: connectivity, human capital (digital skills), use of Internet by citizens, integration of technology and digital public services.
The aim of this article is to compare the development of the digital economyof Russia with the 28 countries of European Union (EU). Data were compiled from the 2018 edition of the International Digital Economy and Society Index (I-DESI 2018) database2. We investigatethe robustness of the EU data supply, and the stability of its ranking. For this, we used the data envelopment analysis (DEA) method and the one-dimensional version of multidimensional scaling, which can also be used for ranking. We compare the results obtained to answer our questions. In evaluating the results, we can also find out whether Russia faresbetter or worsethan the EU in the digital economy.
The paper is organized as follows. In the second part, the measurement method of DESI data is supplied with the five dimensions of the scales. The next chapter presents the ranking of the countries involved in the examinations with the five dimensions. We outline six models for ranking. First, the countries involved in the study are sorted by a weighting method known as the scoring model. The resulting index is the DESI overall index. The following two models are closely related, as we use the classic data envelopment analysis model in both of them. However, we use two slightly different databases. The reason for this is that the input criteria in our case have the best values, but in the DEA model they have to be sorted for the worst values. This sorting can be achieved in two ways: either the reciprocal of the input data is taken, or our initial data is placed on a new scale with a linear transformation. Both approaches are followed and their results are compared in this article. A disadvantage of the basic DEA model is that we need to solve a linear programming problem for as many objects as we have in our dataset (29 in our case). In the next two models, while maintaining the assumption on inputs, we use the DEA common weights analysis method, i.e. we count all countries with the same weight as the scoring model. Our last ranking is linked to the multidimensional scaling method known from multivariate statistics. Namely, if we project our points from the multi-dimensional space to the straight line, that is to say, one-dimensionally, we get a sequence that we use. The next chapter compares these six types of ranking. The comparison suggests that the methods described provide a similar solution. The last, fifth chapter of the paper concludes the results.
1 I-DESI 2018: How digital is Europe compared to other major world economies? // European Commission 26.10.2018. URL: https://ec.europa.eu/digital-single-market/en/news/international-digital-econo-my-and-society-index-2018 (accessed: 04.06.2019).
2 International Digital Economy and Society Index 2018. URL: https://ec.europa.eu/digital-single-market/en/news/international-digital-economy-and-society-index-2018 (accessed: 05.06.2019).
1. Short Literature Review
The literature on measuring the development and impact of the digital economy and society is very diverse and we only attempt to provide a short overview of some of the relevant themes that have been explored in this context. A recent study by a joint Czech-Latvian teamanalyzed ICT-relatedhuman capital elements and government policies in the Czech Republic and Latvia, finding that there was no statistically significant difference in adults' readiness to study online between the two countries [Mirke, Kasparova, Cakula, 2019]. A paper of Götzanalyzed the impact of Industry 4.0 on German-Polish economic relations. The author concludes in herwork that the digital economy can have a positive effect on German-Polish relationships [Götz, 2017]. Another recent study by Silvaggi and Pesce looked at how digitalization and the digital economy "can win" in Portugal, Italy and Greece. Their research focused on the impact of digitization on museums, including the redemption of workplace skills [Silvaggi, Pesce, 2018].
Russian and Ukrainian scholars have also been fairly active in the field. Grytsulenko and Umanets evaluated the spread of the digital economy in an international context. The comparison was carried out with the involvement of the European Union, the Commonwealth of Independent States and Ukraine. Their analysis was mainly done by processing the available statistical data [Grytsulenko, Umanets, 2018]. Another recent article by Belanova and co-authors sought to identify the main directions and indicators for the development of the digital economy. The authors carry out a comparative analysis of international indices related to Information and Communication Technologies (ICT) and digitalization, including the I-DESI [Belanova, Kornilova, Sultanova, 2020]3. A paper of Afonasova, Panfilova and Galichkina analyzed indicators that characterize the level of digital sector development with a view to developing measures stimulating the digitali-zation process[Afonasova, Panfilova, Galichkina, 2018]. A recent study by Dobrulyova, Alexandrov and Yefremov aims to benchmark Russian ICT development with that in the EU countries and identify some important preconditions for the digital transformation. The authors conclude that Russia's lag in terms of connectivity, digital skills, and business adoption of digital technology is significant and is likely to further increase [Dobrolyubo-va, Alexandrov, Yefremov, 2017]. Finally, Petrenko and co-authors analyzed sub-indices of the international Networked Readiness Index (NRI) in order to understand the problems of transition to the digital economy in Russia and determine the ways to resolve them [Petrenko et al., 2017].
2. Measurement of the Digital Economy
Due to the pervasiveness of ICT, data about its application and impact is generated in unprecedented magnitudes. There are several indices, scores, indicators, measurement units that describe the status of the digital economy, society, public administration and used as descriptors of digital transformation.
Firstly, there are the scoring systems describing and comparing global impacts and situation in digitization. These are for instance the UN, OECD, World Bank or ITU re-
3 Although the study is due to be published in 2020 as a book chapter, it is already available online from February 2019 at the publisher.
ports serving similar objectives as some major consulting firms' regular research projects such as Forrester, IDC, Gartner or McKinsey surveys.
The second category of these measures are the ones that focus on regional or well-defined country clusters belonging to a geopolitical area. Typical surveys of this kind are the EU scoreboards: the Digital Economy and Society Index (DESI)4, Digital Skills Indicator (DSI)5 or the Consumer Conditions Scoreboard (CCS)6.
Finally, the third set of data that is collected for describing the ICT impacts are country specific collections usually carried out by National Statistical Offices or domestic research firms.
Although DESI is being debated by experts, and as we will show there are several problems of its method and collection system, it is still the most robust, unavoidable and arguably the best choice for describing European progress on digitalization.
The DESI reports track the progress made by Member States in terms of their digitization. They are structured around five chapters (Table 1).
Table 1. Dimensions of DESI
DESI Dimensions Relevant policy areas and indicators
Connectivity Fixed broadband, mobile broadband and prices
Human Capital Internet use, basic and advanced digital skills
Use of Internet Services Citizens' use of content, communication and online transactions
Integration of Digital Technology Business digitization and e-commerce
Digital Public Services eGovernment and eHealth
Based on: The Digital Economy and Society Index (DESI) // European Commission. URL: https:// ec.europa.eu/digital-single-market/en/desi (accessed: 04.06.2019).
It is a widely used and quoted measurement system by the experts and policy makers but it certainly has its advantages and serious limitations. Its main advantage is that it is measured in 28 countries, and by doing so allows comparison, it is accepted by the European Union and allows compliance, and it provides the big picture of the digital ecosystem in the Union and the member countries.A separate dataset (International Digital Economy and Society, I-DESI) aims to mirror and extend the results of DESI to all 28 EU and 17 non-EU countries for benchmarking purposes.
Disadvantages are rooted from similar sources as advantages: the fact that measurements are collected in 28 different countries entails that the methodology is determined to be general and applicable in all. Therefore, the results are also fairly general and not suitable for deep analysis and explanation of certain phenomena. Specifically, major drawbacks are that measurement factors often have the impression of improvised choice in a
4 The Digital Economy and Society Index (DESI) // European Commission. URL: https://ec.europa. eu/digital-single-market/en/desi (accessed: 04.06.2019).
5 A new comprehensive Digital Skills Indicator // European Commission. URL: https://ec.europa.eu/ digital-single-market/en/news/new-comprehensive-digital-skills-indicator (accessed: 04.06.2019).
6 Consumer Scoreboards // European Commission. URL: https://ec.europa.eu/info/policies/consumers/ consumer-protection/evidence-based-consumer-policy/consumer-scoreboards_en (accessed: 06.06.2019).
given year and they often change. It often seems biased by industry lobbies, the time between the data collection and publication is very long — resulting frequently in outdated assessments. Indicators and sub-indicators change year by year which makes it difficult to compare time series performances because these corrections are not emphasized enough. There are also significant differences between the statistical offices and data collection methods between countries and these problems are only exacerbated for the extended database7.
3. Ranking of Countries Russia and EU-28
Our dataset (Table 2, fig. 1) was compiled from the I-DESI website8. The original dataset contains data from 45 countries: data from the EU-28 and data from 17 non-EU countries, including Russia. From this dataset, we collected data from the 28 EU Member States and supplemented with Russia's sub-indicators to obtain a dataset with 29 coun-tries.The five indicators/variables were used for ranking analysis. We were looking for answers to the following questions with data envelopment analysis (DEA):
(a) what is the ranking with scoring model under known weights used in EU materials;
(b) are the results changed with basic DEA method; is DEA/CWA a robust method;
(c) is DEA/CWA a robust method;
(d) are the results with multidimensional scaling significant?
Since the xi scores for the dimensions are calculated from a weighted sum of normalized individual indicators, the numbersin Table 2 "have little meaning as quantities in themselves"9, but they should allow us to compare the relative performance of our 29 countries in each dimension and evaluate their overall digital competitiveness. Russia ranks 10th in the Human Capital dimension, 18-19th (tied with Poland) in Digital Public Services, 23rd in the Use of Internet, 28th in the Integration of Digital Technology and 29th in Connectivity.
The European Commission uses a weighted sum of these dimensions to calculate the DESI overall index (and their own ranking), but data envelopment analysis (DEA) and multidimensional scaling (MDS) offer viable alternative solutions to the aggregation/ ranking problem, allowing us to test the robustness of their ranking. Six analyses are pre-
7 The authors ofthe I-DESI 2018 report them selves note that although "the match-up between I-DESI and EU28 DESI indicators is generally good", "[p]erfection could only be achieved if the sample sizes and data collection methods used by national statistical agencies inEU28 Member States was replicated in other countries" (p. 30). They also add that "[g]iven a reliance on secondary data to build the 2018 I-DESI it was necessary tomake estimations to compensate for missing and incomplete data" (p. 33). International Digital Economy and Society Index 2018 // SMART 2017/0052 — Final Report. A study prepared for the European Commission DG Communications Networks, Content & Technology by Tech4i2 (Paul Foley, David Sutton, Ian Wiseman, Lawrence Green, Jake Moore). URL: https://ec.europa.eu/digital-single-market/en/news/ international-digital-economy-and-society-index-2018 (accessed: 05.06.2019).
8 International Digital Economy and Society Index 2018. URL: https://ec.europa.eu/digital-single-market/en/news/international-digital-economy-and-society-index-2018 (accessed: 05.06.2019).
9 International Digital Economy and Society Index 2018 // SMART 2017/0052 — Final Report (p. 10). A study prepared for the European Commission DG Communications Networks, Content & Technology by Tech4i2 (Paul Foley, David Sutton, Ian Wiseman, Lawrence Green, Jake Moore). URL: https://ec.europa.eu/digital-single-market/en/news/international-digital-economy-and-society-index-2018 (accessed: 05.06.2019).
sentedin this chapter. First, we determine the classical DESI overall index with the weights suggestedby the Commission. This investigation is known in the decision theory as a scoring model. The values in Table 3 are used for this.
Table 2. The basic data (x)
Country Code Connectivity Human Capital Use of Internet Integration of Digital Technology Digital Public Services
Austria AT 0.63 0.59 0.60 0.59 0.72
Belgium BE 0.68 0.60 0.62 0.61 0.61
Bulgaria BG 0.61 0.47 0.42 0.36 0.45
Croatia HR 0.54 0.45 0.49 0.46 0.56
Cyprus CY 0.54 0.45 0.54 0.39 0.49
Czech Republic CZ 0.67 0.58 0.58 0.39 0.43
Denmark DK 0.77 0.80 0.79 0.71 0.71
Estonia EE 0.62 0.66 0.70 0.53 0.85
Finland FI 0.72 0.73 0.78 0.67 0.83
France FR 0.59 0.62 0.59 0.53 0.82
Germany DE 0.64 0.62 0.66 0.59 0.69
Greece EL 0.50 0.48 0.46 0.45 0.48
Hungary HU 0.60 0.62 0.55 0.51 0.46
Ireland IE 0.63 0.77 0.56 0.51 0.66
Italy IT 0.51 0.50 0.42 0.47 0.68
Latvia LV 0.65 0.47 0.58 0.32 0.56
Lithuania LT 0.61 0.53 0.58 0.46 0.63
Luxembourg LU 0.65 0.67 0.79 0.77 0.64
Malta MT 0.64 0.48 0.57 0.57 0.66
Netherlands NL 0.75 0.69 0.76 0.75 0.76
Poland PL 0.53 0.53 0.51 0.33 0.57
Portugal PT 0.60 0.43 0.47 0.39 0.55
Romania RO 0.61 0.43 0.48 0.27 0.39
Russia RU 0.39 0.64 0.49 0.30 0.57
Slovakia SK 0.57 0.65 0.59 0.40 0.38
End of Table 2
Country Code Connectivity Human Capital Use of Internet Integration of Digital Technology Digital Public Services
Slovenia SI 0.60 0.44 0.53 0.43 0.67
Spain ES 0.64 0.62 0.58 0.55 0.82
Sweden SE 0.75 0.69 0.78 0.65 0.73
United Kingdom UK 0.74 0.65 0.72 0.68 0.90
Based on: International Digital Economy and Society Index 2018. URL: https://ec.europa.eu/digi-tal-single-market/en/news/international-digital-economy-and-society-index-2018 (accessed: 05.06.2019).
0.90 IT 0.85
0 0.80 c 0.75
0.70 S 0.65
1 0.60
~ 0.55
œ
= 0.50
ra
> 0.45 I 0.40 I 0.35 i 0.30 ■n 0.25 ™ 0.20 I 0.15
I 0.10
1 0.05 0.00
A
/T--X L A
PUBN / /Tl
Ellin / \ Bin /\ /
H ^P m ■SI m w 11 PB m
IS w ■ lfli 1 ^p UV
mm ^ r I: ▼] lfv W U
mwtmmmm \rm
CP
V
* of
r ^ w
Fig. 1. Spread of the basic data Note: see Table 2.
Based on: International Digital Economy and Society Index 2018. URL: https://ec.europa.eu/digital-single-market/en/news/international-digital-economy-and-society-index-2018 (accessed: 05.06.2019).
Table 3. The weights of the variables for DESI overall index (vector w)
Connectivity Human Capital Use of Internet Integration of Digital Technology Digital Public Services
0.25 0.25 0.15 0.2 0.15
Based on: International Digital Economy and Society Index 2018. URL: https://ec.europa.eu/dig-ital-single-market/en/news/international-digital-economy-and-society-index-2018 (accessed: 05.06.2019).
We then place the DEA model at the center of the analysis. In the DEA model, the sub-indicators (criteria) are divided into two groups: input and output criteria. The input
criteria are connectivity and human capital, while the output criteria are use of Internet, integration of digital technology and digital public services10. However, we need to transform our data, because in the case of our two input criteria, we have to convert the best maximum value to the minimum. This can be achieved in two ways: by reciprocating the criteria values or by linear transformation. Both methods are used to analyze whether they give significantly differingresults.
Similarly, we perform the data envelopment analysis/common weights analysis (DEA/CWA) with two different data sets. The advantage of this method is that we do not have to solve 29 linear programming problems in our case, only one, and we take the data of all countries into account with the same weight.
Finally, multidimensional scaling is projected to a one-dimensional one, giving us a ranking.
3.1. DESI Overall Indices for the Given 29 Countries
with Scoring Model
In decision theory [Parmigiani, Inoue, 2009], scoring models assign value to decision making units (DMU) to multiply the given criteria with a predetermined weight vector. Suppose that for weight vector w the ith DMU values along the criteria are vector x,. Then we assign a w-Xi value to ith DMU:
m
Fi = w ■ Xj = Ywj'xji (i = 1 2, •••, n).
j=i
where the number of criteria is m and the number of DMU's is n. The values F, are then the DESI overall indices.
The indices are contained in Table 4. The countries with the top rankingsare Denmark, Netherlands, and Finland. Russia ranks 26th, outperforming Greece, Bulgaria, and Romania, which are the countries with the least favorable rankings.
3.2. Basic DEA Model with Reciprocal Values of Input Criteria
The DEA method is a general framework to evaluate countries in the absence of weights of the criteria. The basic method was initiated by Charnes with co-authors to determine the efficiency of decision-making units (DMU) [Charnes, Cooper, Rhodes, 1978; Charnes et al., 2013]. The model offered by them is a hyperbolic programming model under linear conditions. A general solution method for this kind of model was first investigated by Martos, who examined the problem as a special case of linear programming models [Martos, 1964]. The aim of the DEA model is to construct the weights for the input and output criteria. The weights are vectors v and u for the input and output criteria. Let
10 The delineation of input and output criteria was based on the characteristics of their sub-dimensions and individual indicators. The DESI 2018 methodological note also suggests that Connectivity and Human Capital "represent the infrastructure of the digital economy and society", while the other dimensions "are enabled by the infrastructure and their contribution is strengthened by the quality of such infrastructure" (p. 18). DESI 2018 Digital Economy and Society Index. Methodological note // European Commission. URL: http://ec.europa.eu/information_society/newsroom/image/document/2018-20/desi-2018-methodol-ogy_E886EDCA-B32A-AEFB-07F5911DE975477B_52297.pdf (accessed: 04.06.2019).
us formulate the DEA model in the next form, assuming that we examine the efficiency of the 1st decision making unit:
u • yi / v • xi ^ max (1)
s.t.
u • yj / v • Xj < 1; j = 1, 2, ..., 29. (2)
u > 0, v > 0. (3)
— (3) is the basic DEA method, which can be reformulated in a linear programming model (LP) in the following form:
u • yi ^ max (4)
s.t.
v • xi = 1, (5)
u • yj -v • Xj<0; j = 1, 2, ..., 29. (6)
u > 0, v > 0. (7)
(4)-(7) can be solved with commercial software, e. g., with Microsoft Excel Solver. Throughout the paper, we apply this software to construct our calculations.
The input criteria/variables of the evaluation are Connectivity and Human Capital, while the outputs are Use of Internet Services, Integration of Digital Technology, and Digital Public Services. To determine the efficiencies of countries, 29 linear programming (LP) problems must be solved.
First, let us transform the values of the input criteria. The new input values are equal to x'ji = 1 / Xji. The new transformed values are shown in Appendix (Table 1).
After obtaining the results of 29 LP problems, the DEA efficiencies are presentedin
Table 4.
The best countries are still Denmark, Finland, and theNetherlands. The worst countries on the field are Croatia, Bulgaria, and Greece. In this case, Romania and Russia perform considerably better, with the latter ranking 20th.
Table 4. The calculated rankings
Country DESI overall index (scoring) Efficiencies with DEA (reciprocal) Efficiencies with DEA (on a scale) Efficiencies with DEA/ CWA (reciprocal) Efficiencies with DEA/ CWA (on a scale) MDS values
Value Rank Value Rank Value Rank Value Rank Value Rank Value Rank
Austria 0.621 12 0.727 13 0.128 12 0.697 11 0.128 10 -0.462 11
Belgium 0.627 11 0.738 11 0.143 10 0.717 10 0.131 8 -0.422 13
Bulgaria 0.473 28 0.451 28 0.030 28 0.451 27 0.030 26 1.241 28
Croatia 0.497 22 0.479 27 0.046 25 0.478 24 0.046 21 0.789 20
Country DESI overall index (scoring) Efficiencies with DEA (reciprocal) Efficiencies with DEA (on a scale) Efficiencies with DEA/ CWA (reciprocal) Efficiencies with DEA/ CWA (on a scale) MDS values
Value Rank Value Rank Value Rank Value Rank Value Rank Value Rank
Cyprus 0.480 25 0.481 26 0.031 26 0.481 23 0.031 24 0.999 25
Czech Rep. 0.542 17 0.639 15 0.077 18 0.598 18 0.036 22 0.748 19
Denmark 0.760 1 1.000 1 1.000 1 1.000 1 1.000 1 -1.660 1
Estonia 0.659 7 0.926 6 0.170 8 0.804 7 0.164 7 -0.959 7
Finland 0.738 3 1.000 1 0.382 5 0.985 3 0.382 5 -1.545 2
France 0.620 13 0.839 9 0.131 11 0.686 12 0.131 9 -0.582 9
Germany 0.636 8 0.737 12 0.126 13 0.735 9 0.126 11 -0.568 10
Greece 0.476 27 0.414 29 0.031 27 0.400 28 0.025 27 1.045 26
Hungary 0.559 15 0.559 22 0.060 21 0.529 19 0.032 23 0.431 16
Ireland 0.635 9 0.839 10 0.339 6 0.645 14 0.108 14 -0.461 12
Italy 0.512 21 0.566 21 0.065 20 0.454 26 0.065 18 0.705 18
Latvia 0.515 20 0.636 16 0.083 17 0.636 15 0.083 17 0.823 21
Lithuania 0.559 16 0.626 17 0.086 16 0.626 16 0.086 16 0.208 15
Luxembourg 0.699 6 0.914 7 0.161 9 0.814 6 0.115 12 -1.341 5
Malta 0.579 14 0.684 14 0.115 14 0.662 13 0.115 13 -0.065 14
Netherlands 0.738 2 1.000 1 0.570 3 0.975 5 0.570 3 -1.536 4
Poland 0.493 23 0.500 24 0.047 24 0.484 22 0.047 20 0.917 23
Portugal 0.489 24 0.514 23 0.058 22 0.514 20 0.058 19 0.962 24
Romania 0.445 29 0.481 25 0.023 29 0.465 25 0.012 28 1.572 29
Russia 0.477 26 0.602 20 0.066 19 0.348 29 0.030 25 1.222 27
Slovakia 0.531 18 0.607 18 0.056 23 0.496 21 0.007 29 0.912 22
Slovenia 0.526 19 0.604 19 0.093 15 0.600 17 0.093 15 0.516 17
Spain 0.635 10 0.849 8 0.174 7 0.737 8 0.174 6 -0.659 8
Sweden 0.717 5 0.976 5 0.528 4 0.976 4 0.528 4 -1.294 6
UK 0.727 4 1.000 1 0.613 2 1.000 1 0.613 2 -1.537 3
Based on: International Digital Economy and Society Index 2018. URL: https://ec.europa.eu/dig-ital-single-market/en/news/international-digital-economy-and-society-index-2018 (accessed: 05.06.2019).
3.3. Basic DEA Model with Linearly Transformed Values of Criteria
The transformation of the basic data is based on a utility function. The utility functions of criteria have a range between 1 and 20. For the input, data we have chosen the
function . _ xmax
19 x j
Uij =-- • x, —19--J-- -1,
iJ max min iJ max min
X ■ — x■ x■ — x ■
J J J J
where value xja is the most preferable value of criterion j, and value xj is the worst value of this criterion. For the output data we have developed
19 xmax
U a =-- • xij —19--J-- + 20,
iJ max min iJ max min
x ■ — x■ x ■ — x ■
J J J J
where value xja is the most preferable value of criterion j, and value xj is the worst value of this criterion. The used transformation is an affine one, as analysed by Fare and Grosskopf [Fare, Grosskopf, 2013]. (See Appendix, Table 2 for the transformed values.)
After obtaining the results of 29 LP problems, the DEA efficiencies are presented in Table 4.
Denmark and the Netherlands retain their place in the top three, but in this case, they are joined by the United Kingdom instead of Finland. Greece, Bulgaria and Romania are at the bottom, and Russia ranks 19th, outperforming several Eastern and Southern European EU countries.
3.4. The DEA Common Weights Analysis (DEA/CWA) Model with Reciprocal Values of Input Criteria
Regarding the basic model of DEA, the question arises as to why each decision making unit (DMU) should be evaluated with different weights. This means that as many linear programming problems must be solved as the number of DMUs. In contrast, the DEA/CWA model is based on the assumption that it is sufficient to solve only a single LP problem with which we evaluate each DMU with the same weights. The purpose of LP is then to minimize the sum of differences between the outputs and the inputs for all DMUs.
Let us use the linear programming problem (4)-(7) for the case, when the sum of inequalities (6) is maximized. The problem (4)-(7) can be reformulated in the following form (4')-(7'):
u ■ Y • 1 - v ■ X • 1 ^ max (4')
s.t.
v ■ 1 = 1, (5')
u ■ Y- v ■ X < 0, (6')
u > 0, v > 0. (7')
In problem (4')-(7') vectors 1 are the summation vectors with elements one, matrices Y and X are the input and output matrices of the decision making units in the following form
Y = [yi, y2, yp], X = [xU X2, ..., Xp].
Equality (5') guarantees the boundedness of the set of the weights. Inequalities (6') subsume the efficiency indices. Goal function (4') summarizes the deviations from the maximal efficiency. The solution of problem (4')-(7') are the common weights for our problem. The next, second phase determines the efficiency of the decision making units. The optimal solution and the efficiencies are presented in Table 4.
The country with the best ranking is still Denmark, joined by the UK at the top. Russia ranks 29th, below Greece and Bulgaria.
3.5. The DEA Common Weights Analysis (DEA/CWA) Model
with Linearly Transformed Data
Solve problem (4')-(7') now with numbers in Appendix 2 (Table 2). The optimum efficiencies are in Table 4.
Denmark is first in this ranking as well, while now Slovakia is at the bottomwith Russia ranking 25th.
3.6. Ranking with Multidimensional Scaling (MDS)
Multidimensional Scaling is a well-known multivariate statistical method. The essence of the method is to map points from a higher dimensional space to a lower dimensional space so that the distances are kept as high as possible. If the MDS method is mapped into one-dimensional space, that is to say the line, then we get a sequence if the distances in the two spaces are well correlated.
Table 4 shows the distances received. The method's stress is 0.24235, which can be called good. Correlation between the distances of the two spaces, i.e. the R square, is 0.902, which is strong enough to be regarded as a sequence at the same time. In the ranking, Denmark is still on top, Russia is in 27th place above Bulgaria and Romania.
4. Comparison of the Results
The rankings obtained with DEA and multidimensional scaling are very similar to each other andtheranking using the original DESI weights (as evidenced by the fairly strong correlations between them), indicating their robustness. The rankings according to the DEA efficiencies, MDS values and DESI overall indices are presented below in Table 4, fig. 2 while the correlations between the ranking methodsused in our study areshown in Appendix (Table 3).
For most countries, the rankingis fairlystable regardless of which method is used, with Denmark ranking first in all of them. For Russia, however, it exhibitswider variation, asthe country ranks as high as 19th if the basic DEA model is used with linearly transformed data, but only 29th according to the DEA/CWA model with reciprocal data.
Conclusions
The paperdescribes the structure of the Digital Economy and Society Index with its five principal dimensions. The aim was to compare the indices of Russia and the 28 member states of European Union with the available data. We created six indices: the DESI
Fig. 2. Spread of the extreme rankings Note: see Table 4.
Based on: International Digital Economy and Society Index 2018. URL: https://ec.europa.eu/digital-single-market/en/news/international-digital-economy-and-society-index-2018 (accessed: 05.06.2019).
overall index, two efficiency indicators that can be determined by the DEA method, two DEA/CWA indicators, and finally an index of the multidimensional scaling of multivari-ate statistics.
Comparing the six indicators shows that the sequences exhibitvery similar results. This may also mean that weights for DESI do not significantly affect the order of countries. In our calculations, Russia is part of the last third of EU countries in digital development, although their ranking shows marked variation. Where Russia is considered to be strong is the dimension of Human Capital. This is the reserve that the country can draw onin the digital economy.
Future research should answer the question of how the results can contribute to the formulation of policy recommendations. To do this, the five dimensions of DESI should be examined in terms of how to improve the coherence of dimensions. It is also advisable to examine additional methods for conducting the ranking because the scoring model does not differentiate countries sufficiently if there is redundancy between the data.
Acknowledgment
The authors thank for the support of NKFIH K 132160. References
Afonasova M. A., Panfilova E. E., Galichkina M. A. (2018) Social and Economic Background of Digital Economy: Conditions for Transition. European Research Studies, Special Issue 3, vol. 21, pp. 292-302. Belanova N. N., Kornilova A. D., Sultanova A. V. Target Indicators and Directions for the Development of the Digital Economy in Russia (2020) In: Ashmarina S., Mesquita A., Vochozka M. (eds) Digital Transformation of the Economy: Challenges, Trends and New Opportunities. Advances in Intelligent Systems and Computing, 2020, vol. 908. Springer, Cham. URL: https://link.springer.com/chapt er/10.1007/978-3-030-11367-4_11 (accessed: 06.07.2019).
Charnes A., Cooper W. W., Rhodes E. (eds) (1978) Measuring the efficiency of decision making units. Eur J
Oper Resno. 2, pp. 429-444. Charnes A., Cooper W. W., Lewin A. Y., Seiford L. M. (eds) (2013) Data envelopment analysis: Theory,
methodology, and applications. Springer Science & Business Media. 513 p. Dobrolyubova E., Alexandrov O., Yefremov A. Is Russia Ready for Digital Transformation? (2017) Alexandrov D., Boukhanovsky A., Chugunov A., Kabanov Y., Koltsova O. (eds). Digital Transformation and Global Society. DTGS 2017. Communications in Computer and Information Science, vol. 745. Springer, Cham, pp. 431-444. URL: https://link.springer.com/chapter/10.1007/978-3-319-69784-0_36 (accessed: 06.07.2019).
Fare R., Grosskopf S. (2013) DEA, directional distance functions and positive, affine data transformation. Omega, no. 41, pp. 28-30.
Grytsulenko S. I., Umanets O. Y. ( 2018) The Infocommunications Development ofUkraine under Conditions
of the Transition to the Digital Economy. Problemy Ekonomiky, no. 4, 4960, pp. 49-60. (In Russian) Gotz M. (2017) Industry 4.0 — the perspective of international economics. The case of Polish-German
relationships. Przeglqd Zachodni, iss. 365, no. 04, pp. 169-185. Martos B. (1964) Hyperbolic programming. Naval Res Logist Q, iss. 2, no. 11, pp. 135-155. Mirke E., Kasparová E., Cakula, S. (2019) Adults' readiness for online learning in the Czech Republic and Latvia (digital competence as a result of ICT education policy and information society development strategy). Periodicals of Engineering and Natural Sciences, iss. 1, no. 7, pp. 205-215. Parmigiani G., InoueL. (2009) Decision theory: Principles and approaches, vol. 812. John Wiley & Sons,
Padstow, Cornwall, United Kingdom. 404 p. Petrenko S. A., Makoveichuk K. A., Chetyrbok P. V., Petrenko A. S. (2017) About readiness for digital economy. 2017 IEEE II International Conference on Control in Technical Systems (CTS), 2017, St. Petersburg, Russia, pp. 96-99. URL: https://ieeexplore.ieee.org/document/8109498 (accessed: 07.07.2019).
Silvaggi A., Pesce F. (2018) Job profiles for museums in the digital era: research conducted in Portugal, Italy and Greece within the Mu. SA project. Journal of Cultural Management and Policy, iss. 1, no. 8, pp. 56-69.
Received: 15.07.2019 Accepted: 11.09.2019
Authors' information:
Zoltán Bánhidi — MA in Economics; banhidiz@kgt.bme.hu Imre Dobos — Dr. Sci. in Economics, Professor; dobos@kgt.bme.hu
András Nemeslaki — PhD in Mechanical Engineering; Professor; nemeslaki@finance.bme.hu
Appendix
Table 1. The reciprocally transformed data
Country Code Connectivity Human Capital Use of Internet Integration of Digital Technology Digital Public Services
Austria AT 1.59 1.69 0.60 0.59 0.72
Belgium BE 1.47 1.67 0.62 0.61 0.61
Bulgaria BG 1.64 2.13 0.42 0.36 0.45
Croatia HR 1.85 2.22 0.49 0.46 0.56
Cyprus CY 1.85 2.22 0.54 0.39 0.49
Czech Republic CZ 1.49 1.72 0.58 0.39 0.43
Denmark DK 1.30 1.25 0.79 0.71 0.71
Estonia EE 1.61 1.52 0.70 0.53 0.85
Finland FI 1.39 1.37 0.78 0.67 0.83
France FR 1.69 1.61 0.59 0.53 0.82
Germany DE 1.56 1.61 0.66 0.59 0.69
Greece EL 2.00 2.08 0.46 0.45 0.48
Hungary HU 1.67 1.61 0.55 0.51 0.46
Ireland IE 1.59 1.30 0.56 0.51 0.66
Italy IT 1.96 2.00 0.42 0.47 0.68
Latvia LV 1.54 2.13 0.58 0.32 0.56
Lithuania LT 1.64 1.89 0.58 0.46 0.63
Luxembourg LU 1.54 1.49 0.79 0.77 0.64
Malta MT 1.56 2.08 0.57 0.57 0.66
Netherlands NL 1.33 1.45 0.76 0.75 0.76
Poland PL 1.89 1.89 0.51 0.33 0.57
Portugal PT 1.67 2.33 0.47 0.39 0.55
Romania RO 1.64 2.33 0.48 0.27 0.39
Russia RU 2.56 1.54 0.49 0.30 0.57
Slovakia SK 1.67 2.27 0.59 0.40 0.38
Slovenia SI 1.56 1.61 0.53 0.43 0.67
Spain ES 1.33 1.45 0.58 0.55 0.82
Sweden SE 1.35 1.54 0.78 0.65 0.73
United Kingdom UK 1.59 1.69 0.72 0.68 0.90
Based on: International Digital Economy and Society Index 2018. URL: https://ec.europa.eu/digital-single-market/en/news/international-digital-economy-and-society-index-2018 (accessed: 05.06.2019).
Table 2.The linearly transformed data
Country Code Connectivity Human Capital Use of Internet Integration of Digital Technology Digital Public Services
Austria AT -8.00 -11.78 10.24 13.16 13.42
Belgium BE -5.50 -11.27 11.27 13.92 9.40
Bulgaria BG -9.00 -17.95 1.00 4.42 3.56
Croatia HR -12.50 -18.97 4.59 8.22 7.58
Cyprus CY -12.50 -18.97 7.16 5.56 5.02
Czech Republic CZ -6.00 -12.30 9.22 5.56 2.83
Denmark DK -1.00 -1.00 20.00 17.72 13.06
Estonia EE -8.50 -8.19 15.38 10.88 18.17
Finland FI -3.50 -4.59 19.49 16.20 17.44
France FR -10.00 -10.24 9.73 10.88 17.08
Germany DE -7.50 -10.24 13.32 13.16 12.33
Greece EL -14.50 -17.43 3.05 7.84 4.65
Hungary HU -9.50 -10.24 7.68 10.12 3.92
Ireland IE -8.00 -2.54 8.19 10.12 11.23
Italy IT -14.00 -16.41 1.00 8.60 11.96
Latvia LV -7.00 -17.95 9.22 2.90 7.58
Lithuania LT -9.00 -14.86 9.22 8.22 10.13
Luxembourg LU -7.00 -7.68 20.00 20.00 10.50
Malta MT -7.50 -17.43 8.70 12.40 11.23
Netherlands NL -2.00 -6.65 18.46 19.24 14.88
Poland PL -13.00 -14.86 5.62 3.28 7.94
Portugal PT -9.50 -20.00 3.57 5.56 7.21
Romania RO -9.00 -20.00 4.08 1.00 1.37
Russia RU -20.00 -9.22 4.59 2.14 7.94
Slovakia SK -11.00 -8.70 9.73 5.94 1.00
Slovenia SI -9.50 -19.49 6.65 7.08 11.60
Spain ES -7.50 -10.24 9.22 11.64 17.08
Sweden SE -2.00 -6.65 19.49 15.44 13.79
United Kingdom UK -2.50 -8.70 16.41 16.58 20.00
Based on: International Digital Economy and Society Index 2018. URL: https://ec.europa.eu/digi-tal-single-market/en/news/international-digital-economy-and-society-index-2018 (accessed: 05.06.2019).
Table 3. Correlations between the DESI, DEA and MDS scores
DEA (reciprocal) DEA (on a scale) DEA/CWA (reciprocal) DEA/CWA (on a scale) MDS values
DESI overall index Pearson Correlation .968** .810** .961** .791** -.991**
Sig. (2-tailed) .000 .000 .000 .000 .000
DEA (reciprocal) Pearson Correlation .785** .940** .757** -.959**
Sig. (2-tailed) .000 .000 .000 .000
DEA (on a scale) Pearson Correlation .821** .981** -.773**
Sig. (2-tailed) .000 .000 .000
DEA/CWA (reciprocal) Pearson Correlation .829** -.952**
Sig. (2-tailed) .000 .000
DEA/CWA (on a scale) Pearson Correlation -.761**
Sig. (2-tailed) .000
Based on: International Digital Economy and Society Index 2018. URL: https://ec.europa.eu/digi-tal-single-market/en/news/international-digital-economy-and-society-index-2018 (accessed: 05.06.2019).
Сравнительный анализ развития цифровой экономики в России и ЕС: приложение метода DEA к данным индекса DESI
З. Банхиди, И. Добош, А. Немешлаки
Будапештский университет технологии и экономики, Венгрия, 1117, Будашепт, бул. Венгерских Ученых, 2
Для цитирования: Banhidi Z., Dobos I., Nemeslaki A. (2019) Comparative Analysis of the Development of the Digital Economy in Russia and EU Measured with DEA and Using Dimensions of DESI. Вестник Санкт-Петербургского университета. Экономика. Т. 35. Вып. 4. С. 588-605. https://doi.org/10.21638/spbu05.2019.405
Целью работы является сравнение развития цифровой экономики в России и в 28 странах Европейского союза. Данные были собраны из базы данных Международного индекса цифровой экономики и общества (I-DESI 2018) Европейской комиссии. В статье после краткого обзора различных альтернативных способов измерения воздействия информационных и коммуникационных технологий рассмотренынаиболее важные особенности, преимущества и недостатки этой базы данных. Затем описана структу-раисследуемого набора данных и проведен анализ цифровой конкурентоспособности России и ЕС-28. Основные вопросы исследования касаются надежности данных ЕС и стабильности их рейтинга. Для этого использован метод анализа охвата данных (DEA) и одномерная версия многомерного масштабирования, которая также может применяться для ранжирования вопросов. В дополнение к обычному методу DEA исследуется жизнеспособность моделей DEA с общим весом. Для ответа на поставленные в работе вопросы полученные результаты сравниваются. Их оценка показывает, насколько данные из России соответствуют данным ЕС в цифровой экономике. Сравнение демонстрирует, что методы, использованные в нашем исследовании, дают аналогичное решение, но для рейтинга нескольких стран (включая Россию) характерен более широкий разброс.
Ключевые слова: индекс DESI, цифровое государственное управление, инновации, анализ охвата данных, многомерное масштабирование, ранжирование.
Статья поступила в редакцию 15.07.2019 Статья рекомендована в печать 11.09.2019
Контактная информация:
Банхиди Золтан — магистр экономики; banhidiz@kgt.bme.hu
Добош Имре — д-р экон. наук, проф.; dobos@kgt.bme.hu
Немешлаки Андраш — канд. техн. наук, проф.; nemeslaki@finance.bme.hu
