DOI: 10.29141/2658-5081-2023-24-1-4 EDN: PSIBPS JEL classification: D22, J24, O14
Daria A. Starovatova HSE University, Saint Petersburg, Russia
The relationship between robots and labour productivity:
Does business scale matter?
Abstract. Scholarly literature on the economic consequences of robotisation at the microeconomic level often does not take into account the pronounced digital gap between small and medium-sized businesses and large ones. In this regard, theoretical and real estimates may differ for companies of different sizes. The article studies the relationship between robotisation and labour productivity in the Russian industry in the context of size groups of companies. Methodologically, the study relies on the theory of the firm and economic theories explaining the essence of labour productivity and methods for evaluating it. The research analyses the data about 725 Russian industrial enterprises for 2017 using the methods of regression modeling. The data was obtained in the course of the fundamental research programme at the HSE University. According to the results, only small and medium-sized enterprises have a significant and reliable relationship between the introduction of robots and labour productivity. Probably due to the complexity of business processes, large businesses need deeper and more elaborate robotisation to gain labour productivity benefits. The calculations also demonstrate a negative relationship between exports and labour productivity in large companies, which contradicts the 'classical' ideas about the impact of export activities on the efficiency indicators. This may indicate that the high labour productivity of a considerable part of large Russian enterprises proceeds from their monopoly position in domestic markets, while formally less productive companies, which do not occupy dominant positions, appear to be competitive and motivated enough to enter foreign markets. The findings can be useful for the leadership of enterprises, especially that of SMEs, for the managerial decision-making in terms of increasing productivity, in particular, through robotisation of production.
Keywords: labour productivity; growth drivers; robotics; automation; digital gap; robotisation of production.
Acknowledgements: The paper is prepared in the course of the research no. 22-00-065 as part of the HSE Academic Fund Programme in 2022.
For citation: Starovatova D. A. (2023). The relationship between robots and labour productivity: Does business scale matter? Journal of New Economy, vol. 24, no. 1, pp. 81-103. DOI: 10.29141/2658-5081-2023-24-1-4. EDN: PSIBPS.
Article info: received November 21, 2022; received in revised form December 19, 2022; accepted December 30, 2022
Introduction
Robots have long become an integral part of today's life, but there is still debate about the impact of the system-wide transition from traditional manual labour to machines. In some aspects, the superiority of automated technical systems is undeniable: they can operate in unsafe working conditions, ensure the stable quality of products, and do not feel tired or sick. It is little wonder that the year of 2020 witnessed one of the most widespread introductions of robots ever1. During the COV-ID-19 pandemic, when numerous enterprises around the world put their work on pause due to the forced social isolation and lockdown, those companies that had already used robots or been able to quickly introduce them into production continued working uninterruptedly and remained competitive.
Despite the significant interest of scientific community in digitalisation, there is lack of studies devoted to the analysis of factors, problems and effects of robotisation at the level of industries and firms. Russian researchers place special emphasis on the correlation between automation and employment. According to Urunov and Rodina [2018, pp. 138-142], massive job cuts are an inevitable consequence of the widespread introduction of robots: unemployment will increase to 15-20 % in Russia and up to 30 % in the world with one to three professions disappearing per year. Zemtsov [2017, pp. 142-157] argues that 44.78 % of employees in the Russian Federation may suffer from negative consequences of robotisation; however, these calculations assume "immediate" robotisation, which is unlikely due to economic, political and other restrictions. Tolkachev and Kulakov [2016, pp. 79-87] voice the opposite opinion that the automation-caused increase in productivity, on the contrary, will create new jobs thanks to additional economic activity.
Heterogeneity of views about the impact of robotisation on employment is also typical of foreign publications. It was found that robotisation only reshapes the demand for new labour force capable of ensuring the correct functioning of new technologies, as well as changes their distribution without detriment for the overall employment [Dottori, 2020, pp. 739-795]. The contrasting view is taken by Acemoglu and Restrepo [2020] who revealed that one additional robot per thousand employees reduced the total US employment-to-population ratio by 0.2 percentage points and total wages by 0.42 %. At the same time, wage cuts only occur in the short term [Berg, Buffie, Zanna, 2018]. There are also disagreements over the effect of automatisation
1 IFR. (2021). World Robotics 2021 Industrial Robots. https://ifr.org.
on the low-skilled workforce. De Vries et al. [2020] found a significant negative correlation between robotisation and the share of workers engaged in routine work, but it was only relevant for high-income countries. However, other researchers [Klenert, Fernandez-Macias, Anton Perez, 2020] find no strong evidence for a significant and reliable correlation between robots and low-skilled workers. Thus, the question of the relationship between robots and employment remains open to this day.
In recent studies, attention is being increasingly focused on the impact of robotisation on industry and country productivity. Researchers stress the meaningful contributions of modern industrial robots expressed in the annual labour productivity growth by 0.37 percentage points, which is just over one-tenth of the cumulative growth of the economy [Graetz, Michaels, 2018]. The significant influence of robotisation is also confirmed by Jungmittag and Pesole [2019], who find that one additional robot per 1 million euros of non-ICT capital investment increases labour productivity by 44 %. However, as stated by Cette, Devillard and Spiezia [2021], robotisation has not become a "source of revival" for the total factor productivity (TFP), and the average contribution of robots to the growth of the countries' TFP does not exceed 0.2 percentage points per year. Thus, despite holding contrasting views on the significance of the influence, all the authors highlight the positive effects exerted by robots on labour productivity in different countries.
The present study is of high relevance as it focuses on the relationship between robotisation and labour productivity at the microeconomic level using the in-house data of Russian enterprises and demonstrates the ambiguity of this relationship in terms of size groups of companies.
In recent years, the Russian Federation has shown high growth rates of robotisation (~40 % per year1), but the country is lagging behind the global average. According to 2019 data, robot density in Russia is only 6 robots per 10 thousand employees in industry, while the global average in this field is 1132. In the early 1990s, about 40 % of all robots in the world were used in the USSR. However, with its collapse, the development of robots at the state level was impeded: the mass production and introduction of robots ceased, and the existing robots were dismantled and sold off [Ermolov, 2019]. As a result, Russia was compelled to initiate the robotisation process virtually from scratch. In addition, while in the Russian Federation there is an abundance of cheap labour, entrepreneurs are not inclined to invest in expensive and long-term automated production projects [Gurlev, 2020]. It should also be noted that there are not enough qualified personnel in the country to solve high-tech problems [Arkhipova, Melnikova, 2022]. According to Komkov and Bondareva [2016],
1 The Russian market for industrial robots. https://www.tadviser.ru/index.php. (In Russ.)
2 IFR. (2020). World Robotics. https://ifr.org/ifr-press-releases/news/record.
modernisation of the education system, including the introduction of university courses on robotisation and the renewed interest in technical creativity in schools, can resolve this problem.
Lagging behind leading countries at robotisation reduces Russia's competitiveness in the international market. In 2020, the country was ranked 38th of 42 reporting nations in terms of GDP per hour worked1 and 6th in terms of annual average hours worked2. Thus, Russian enterprises should promote the more productive use of labour resources. In this regard, it is important to assess the impact of robotisation on the performance of enterprises as one of the factors in the growth of labour productivity.
Due to the limited data available at the level of firms, the study was conducted using the evidence for 2017. However, there were no critical changes in the level of robotisation and labour productivity in the Russian Federation during this period: robot density in 2017 was 4 robots per 10 thousand employees3, and the labour productivity level was 25 dollars per hour4, which is only 3 dollars less than in 2020. Hence, the findings of the study are still relevant today.
The research aims to analyse the relationship between the introduction of robots and labour productivity. The literature review presented above allows putting forward the following basic hypothesis.
HI. Ceteris paribus, labour productivity of robotised enterprises is higher than that of non-robotised ones.
To test the sustainability of the research results, additional models were built in the context of size groups of companies. Large firms significantly outstrip others in the development and implementation of digital solutions [Nazarenko, 2021]. Owning to the lower business digitalisation, small and medium-sized enterprises (SMEs) are at a much earlier stage of digital transformation. The main difference between digitalisa-tion and digital transformation is that the former refers to the process of implementing digital solutions, and the latter implies qualitative changes in business processes initiated by digitalisation and leading to significant socioeconomic effects [Gokhberg et al., 2021, pp. 11-16]. Thus, due to the significant lag in the process of SMEs' digital transformation, there is a digital divide between enterprises of different sizes.
With digital technologies becoming increasingly sophisticated, this gap is getting more pronounced, which is associated with a lack of competent personnel required to implement digitalisation in small and medium-sized businesses [Igoshina, 2021]. Additionally, SMEs are more limited in financial capabilities, which hampers the process
1 OECD. (2021). GDP per hour worked (indicator). https://www.oecd.org/.
2 OECD. (2021). Hours worked (indicator). https://www.oecd.org/.
3 IFR. (2018). World Robotics. https://ifr.org/ifr-press-releases/news/record.
4 OECD. (2018). GDP per hour worked (indicator). https://www.oecd.org/.
of their digital transformation. As a result, to measurably increase productivity of SMEs, a smaller critical mass of robots is likely to be required than at large enterprises.
To test the above statement, we formulate an additional hypothesis.
H2. The use of robots has a significant relationship with the level of labour productivity, especially for SMEs.
Theoretical foundations of the research
Robotisation and labour productivity at enterprises. Robotisation has a number of obvious benefits for businesses. First, when labour is more expensive than capital, replacing workers with machines reduces the marginal cost of wages, which increases labour productivity [Bonfiglioli et al., 2020; Deng, Plumpe, Stegmaier, 2021]. However, this is more applicable to large firms that obtain more benefits by paying fixed automation costs to save on variable costs. Secondly, the introduction of robots boosts production as their productivity remains constant, which, accordingly, expands the company's output and productivity. Thirdly, robotisation improves safety. While this does not have a direct impact on productivity, it reduces the risk of shutdowns due to human factors. Moreover, the probability of product defect decreases and the stable quality of products is ensured, which, according to Dixon, Hong and Wu [2021], is the primary reason for replacing manual labour with machines, even ahead of the desire to save on costs.
Graetz and Michaels [2018] note the growing importance of robots in explaining aggregate and sectoral productivity in the years preceding the world recession. After the global economic crisis, however, labour productivity in countries nosedived and demonstrated only a partial recovery before 2013, which was likely caused by financial frictions that hindered innovation, technology adoption and efficient capital allocation [Stiebale, Suedekum, Woessner, 2020]. This slowdown reignited the debate about the role of technology waves in labour productivity growth. The present research was carried out using data for 2017, i.e., the period when business recovered from the consequences of the global crisis, while the COVID-19 pandemic, which again undermined the financial stability of enterprises, had not yet arrived. This diminishes the likelihood of external economic shocks affecting the results of the study.
Despite the paucity of company-level data, there are scientific works on the influence of robotisation on the economic performance of enterprises. In the work [Ballestar et al., 2020], the authors conducted a research based on a survey of 1,800 Spanish small and medium-sized manufacturing firms for the period from 1990 to 2015 and proved that an increase in robot use resulted in a rise in the companies' labour productivity. It was found that robotised SMEs were more efficient (by 2 % and 5 % in 2008 and 2015, respectively), hired more employees, and paid higher salaries. At the same
time, the authors focused not only on the effects of robotisation, showing that the introduction of robots led to a significant increase in labour productivity (~ 20-25 % over four years), but also on what kind of firms were involved in this introduction.
It is noted that robots are more often used in larger and more productive companies, which is probably due to the digital transformation barriers faced by SMEs [Koch, Manalo, Smolka, 2019]. Bonfiglioli et al. [2020] come to the similar conclusion that the introduction of robots occurs after the firm grows in size, which is followed by an increase in efficiency and a drop in demand for labour. After robotisation, companies' labour productivity enhances again in common with the employment rate of highly qualified personnel. It is also revealed that robotisation affects overall sales much less strongly than labour productivity, which may be caused by the reduction of the workforce. This indicates that efficiency gains do not always lead to an equivalent price drop by offsetting part of consumer benefits by increasing the company's markups. Consequently, enterprises, while gaining in productivity, do not seek to reduce the cost of their products, which generates more profits for them. Thus, robotisation of production can not only increase the productivity of the enterprise, but also strengthen the market power of larger manufacturers.
Other factors in labour productivity growth. This section presents other factors in labour productivity growth used as control variables for building models.
Company size. This indicator should be taken into account, since larger companies are characterised by higher labour productivity. There are four categories of enterprises depending on the number of employees: micro (up to 15), small (16-100), medium (101-250), and large (over 250).
Company age. The fact that it is much more difficult to re-equip the existing production in accordance with new technological solutions than to initially utilize new equipment and technologies substantiates the company age parameter included in the model. Therefore, the younger the enterprise, the more advanced technologies it uses, which results in the increase in its productivity.
Type of ownership. Many of the company's business processes that in one way or another affect its efficiency depend on the owner's decisions, which are determined by personal goals. The model distinguishes between two types of owners: 1) foreign companies/individuals; 2) federal, regional and/or local authorities.
Industry. Industries differ significantly, primarily in terms of science intensity, and, therefore, are usually grouped according to the degree of manufacturability. Apart from differentiation in R&D costs, there are also differences in the robotisation level. Robot adopters are likely to be in industries with more notable advances in robotic technology and greater adjusted penetration of robots (APR) that measures the common increase in robot use in an industry among advanced economies (excluding
France) since 1993 and adjusts for the mechanical effect of industry growth on robot use [Acemoglu, LeLarge, Restrepo, 2020]. There are a number of manufacturing industries with high APR and, accordingly, a high degree of robotisation: pharmaceuticals, chemicals, plastics, food and beverages, metal products, primary metals, industrial machinery and automotive. Thus, the model should consider industry-specific effects.
Exports. Entering new markets is associated with great competitive pressure, which gives impetus to the development of the enterprise and generates new knowledge "flowing" into the firm. Ballestar et al. [2020] suggested that increasing sales in international markets brought about higher labour productivity. Under this assumption, the authors considered the growing ability of a manufacturing firm to enhance labour productivity by cutting training expenditures through expanding its presence in international markets. The research results confirmed the hypothesis, which gives us grounds for including exports in the model.
In comparison with the general characteristics of firms, the availability of firmlevel data allows controlling more significant factors.
Innovations. In today's fast-paced world, innovations are of immense importance for staying competitive in the market. In particular, research studies highlight the substantial role of innovations in boosting labour productivity [Garshina, 2011; Tra-chuk, Linder, 2017]. Improved labour results can be achieved through the introduction of innovative products, as well as innovative technologies; our model, therefore, includes both of these components of companies' innovation activity. In addition, it also considers R&D investments, since such inputs contribute to the modernisation of outdated production technologies, which helps reduce costs, increase production volumes, and introduce more high-quality or fundamentally new products.
Equipment depreciation. High depreciation of the company's fixed assets (FAs) leads to a decrease in the production efficiency, while timely replacement and modernisation of FAs improve or at least maintain the same level of labour productivity. Kosyakova and Popova [2017] state that the high degree of fixed assets depreciation has become one of the main reasons behind the low labour productivity of Russian companies. In their study, the degree of FAs depreciation is measured as a percentage of the enterprise's total production capacity generated by machines and equipment aged 11 years or above.
Materials and methods
Data description. The empirical basis of the study was the Competitiveness of the Russian Industry 2018 (CRI 2018) database formed in 2018 through a survey of enterprises as part of the project "The Factors of Competitiveness and Growth of Russian
Manufacturing Enterprises" of the HSE Basic Research Programme. In addition, exact data on the revenue and employment at these enterprises for 2017 was collected using the taxpayer identification number (TIN).
The survey involved 1,716 randomly selected Russian enterprises operating in twenty industries. The sample is representative in terms of industry sectors and size groups (see Appendices 1, 2). The high number of large enterprises covered is due to the purpose of the study. The key feature of the collected database is the availability of data on the use of robots in production. This provides a unique opportunity to explore the relationship between robot use and labour productivity at the level of firm rather than the industry level.
Most of the survey's questions could be answered by a Yes or No, therefore, the main part of the variables under consideration are represented by a binary type. However, there are two exceptions, namely the variables of labour productivity and the share of the enterprise's total production capacity attributable to equipment aged 10 years or above. They are presented in numerical form, with the last variable ranging from 0 to 1.
Of 1,670 enterprises that provided information on robot use, only 313 (~18.7 %) introduced robots in production, which confirms the weak level of robotisation at Russian enterprises.
When using the data collected with help of the TIN, gaps in the companies' staff numbers were found: the information either contained explicit outliers or was absent. After eliminating the observations, 1,091 firms remained in the sample, of which 232 enterprises used robots. When categorising enterprises into size groups (micro - up to 15 employees), small (16-100 employees), medium (101-250 employees) and large (251 or more employees)), it was confirmed that robotised firms are more common among larger and more productive enterprises (Tables 1, 2).
Table 1. The level of robotisation by size groups of enterprises
Size of enterprises N Use robots Do not use robots Level of robotisation, %
Micro 122 20 102 16
Small 383 68 315 18
Medium 196 41 155 21
Large 390 103 287 26
Total 1,091 232 859 21
Table 2. Descriptive statistics of labour productivity by robot use at enterprises,
thousand rubles/person
Robot use Min Quarter 1 Median Mean Quarter 3 Max
Use 19.9 1,679.0 3,361.1 8,421.2 8,115.9 217,161.6
Do not use 18.1 1,183.1 2,484.8 5,993.9 5,362.5 278,490.8
After eliminating gaps for all the variables and revenue outliers, there were 725 observations left in the database; however, the sample remained representative in terms of size groups. Thus, the analysis was carried out using the evidence from 459 small and medium-sized enterprises and 266 large enterprises.
A single-factor indicator of productivity, namely labour productivity, was used as a dependent variable. Due to data limitations, we cannot observe indicators of value added or production volumes, therefore, to measure output, revenue is used. This approach is acceptable, especially if prices show differences in product quality rather than manufacturers' market power, since in the latter case, productivity levels will reflect less about how efficient they are and more about the state of their local output market [Syverson, 2011, p. 331]. Thus, in this research, labour productivity shows how many thousand rubles one employee added to the company's revenue in 2017.
The distribution of labour productivity is far from normal (Figure 1). 0.00020 0.00015
£
a 0.00010
<u
0.00005 0.00000
OinONrnoOCNVO'—iifiO\ilOOmNHiiO^O*mSN»OOinO\r'l
ooovo in m n o o\ n in ^ in h o\ m >o in n h q oo N in ^ « o o\
m in oo h it n o" n in » h n a\ n in m h n on n in oo h 10
Productivity level, thousand rubles/person Fig. 1. Distribution of labour productivity
L
In order to approximate the distribution to normal, the variable was taken as a logarithm. The results are presented in Figure 2.
0.50
I jlillliilii.. —
o.oo —^^¿^M __
mvqooqNinNOvN^iqM^tninNON^NW^tniqoqqfn inininio«»ifiiONNNNoooooooooioio!o\oid66dHH
< 1-H T—I 1—I l i—I
Productivity level, In Fig. 2. Distribution of the natural logarithm of labour productivity
It is also noted that a small enterprise is not always less productive than a large one (Figure 3). However, the relationship between size and labour productivity is positive, and with an increase in scale, the dispersion of labour productivity decreases, and the general trend shows that the large companies with low labour productivity are less common than the small ones.
4.5
0 5,000 10,000 15,000 20,000
Number of employees
Fig. 3. Dependence of the logarithm of labour productivity on an enterprise's size
No obvious differences in the characteristics of robotised and non-robotised enterprises were discovered. It can be noted that robot adopters, as mentioned above, are larger and more productive. Besides, among such companies there are more foreign and state owners. Apparently, robotised enterprises are more attractive to such investors, or vice versa, companies headed by foreign entrepreneurs or the state are prone to introduce robots due to higher financial capabilities or other strategic goals. In addition, equipment depreciation at robotised enterprises is lower and the percentage of new or upgraded technologies is higher. Detailed descriptive statistics are provided in Appendix 3.
Table 3 presents descriptive statistics of the enterprises according to the size group. The large companies have a higher level of state/foreign ownership, and therefore, this group is typically more involved in export activity. Having overseas entrepreneurs/ firms as owners allows manufacturers to overcome difficulties often faced by SMEs, such as high costs of entering foreign markets, lack of information about foreign markets, and low availability of investment resources [Budkova, 2015]. As specified earlier, a mature company's machinery is more likely to be less modern. This is evidenced by the higher share of production capacity generated by equipment older than 10 years in large companies, as they tend to be founded much earlier than most small and medium-sized enterprises. Innovative activity, including investment in innovation, as well as modernisation of products and production technologies, is noticeably higher in such companies due to their substantial financial capabilities and focus on exports. The sectoral distribution of companies is rather similar in both groups.
According to correlation analysis, no multicollinearity was found. There is little correlation between the variables. An exception is categorical variables broken down into binary ones. Their correlation is higher, but not critical (no more than |0.6|). The description of the variables prepared for analysis is presented in Appendix 4.
Table 3. Descriptive statistics of the enterprises
Indicators Enterprises
small and medium-sized (459) large (266)
N % N %
Robot use
Do use 80 17 70 26
Do not use 379 83 196 74
Owners
Foreign 24 5 26 10
State 10 2 15 6
Age
Registered:
Before 1992 27 6 85 32
Between 1992 and 1998 132 29 95 36
Between 1999 and 2010 217 47 74 28
After 2011 83 18 12 5
Export
Exporting 124 27 118 44
Non-exporting 335 73 148 56
R&D investment
Investing 96 21 93 35
Non-investing 363 79 173 65
Introduction of new/modernised products
Introducing 213 46 165 62
Non-introducing 246 54 101 38
Introduction of new/modernised technologies
Introducing 70 15 118 44
Non-introducing 389 85 148 56
Manufacturing industries
Food 107 23 61 23
Textile 8 2 5 2
Clothing 19 4 5 2
Leather, leather goods, footwear 8 2 4 2
Woodworking, wood and cork production, excluding furniture, splint production, weaving materials 21 5 4 2
Cellulose, paper, cardboard and cardboard products 14 3 9 3
Production of coke, oil products 0 0 2 1
Chemicals 16 3 19 7
Medicines and medical materials 9 2 8 3
Rubber and rubber products 31 7 19 7
Table 3 (concluded)
Indicators Enterprises
small and medium-sized (459) large (266)
N % N %
Other non-metallic mineral products 38 8 26 10
Metals production 2 0 10 4
Finished metal products, excluding machinery and equipment 44 10 16 6
Computer technology, electronic and optical equipment 16 3 11 4
Electronic machinery and equipment 24 5 15 6
Machinery and equipment not included in other groups 33 7 18 7
Cars, trailers, semi-trailers 10 2 12 5
Other vehicles and equipment 5 1 14 5
Furniture 25 5 3 1
Repair and installation of machinery and equipment 29 6 5 2
Indicators Min Quarter 1 Median Mean Quarter 3 Max
Small and medium-sized enter prises
Labour productivity, ln 5.4 7.1 7.9 7.9 8.7 11.1
Share of production capacity generated by equipment older than 10 years 0.0 0.0 0.1 0.3 0.5 1.0
Large enter prises
Labour productivity, ln 6.3 7.4 8.0 8.1 8.6 10.8
Share of production capacity generated by equipment older than 10 years 0.0 0.1 0.4 0.4 0.7 1.0
Research method. To conduct the research, a multiple linear regression model was built, which was chosen according to the nature of the dependent variable. The analysis was based on cross-sectional data, it is impossible, therefore, to trace the change in labour productivity depending on the change in regressors over time. The paper examines the "robotisation premium", that is, the amount by which, all other things being equal, robotised companies' labour productivity is higher, as the year of introduction and the number of robots are unknown.
The categorical variables of the enterprise's size, age, and industry are included in the model by splitting into binary variables. Since the model includes one constant, the number of dummy variables should be less by one than the number of categories. The following categories were taken as basic variables: micro for size, the date of registration in the Soviet era for age, food for the manufacturing industry.
Two regression models were obtained for analysis. Model (1) is standard for analysing labour productivity of enterprises, since most studies on the topic under discussion use aggregated data. The use of firm-level data makes it possible to control both the general and internal characteristics of companies that can affect their labour
productivity. Therefore, Model (2) also includes variables that characterise the innovative activity of companies, as well as the enterprise's equipment depreciation.
The models in question take the following forms.
Model (1):
ln (labour productivity)- p0 + Pi x robot use + p2 x micro enterprises + p3 x small enterprises + p4 x medium-sized enterpises + p5 x reg.USSR + p6 x reg.1992-1998 + p7 x reg.after 2010 + p8 x foreign owner + p9 x state owner + p10 x export +
p11-30 x industry, + u.
Model (2):
ln (labour productivity)- p0 + p1 x robot use + p2 x micro enterprises + p3 x small enterprises + p4 x medium-sized enterpises + p5 x reg.USSR + p6 x reg.1992-1998 + p7 x reg.after 2010 + p8 x foreign owner + p9 x state owner + p10 x export + P11 x R&D investment + p12 x introduction of new/modernised products + p13 x introduction of new/modernised technologies + p14 x equipment depreciation + p15-34
x manufacturing industry, + u.
To test Hypothesis 2, the sample is divided into two parts: small and medium-sized enterprises and large enterprises. The models used for them are the same as for testing Hypothesis 1 with the exception of size control variables.
Results and discussion
Consider the results of the regression analysis (Table 4) while taking the critical significance level equal to 10.
Table 4. Determinants of the natural logarithm of labour productivity
Variable Model
all enterprises SMEs large enterprises
(1) (2) (3) (4) (5) (6)
Robot use 0.251*** (0.091) 0.162* (0.092) 0.307** (0.133) 0.216* (0.135) 0.179 (0.108) 0.119 (0.109)
Micro enterprises base base - - - -
Small enterprises 0.195 (0.147) 0.224 (0.140) - - - -
Medium-sized enterprises 0.371** (0.156) 0.374** (0.150) - - - -
Large enterprises 0.335** (0.150) 0.418*** (0.145) - - - -
Reg. USSR base base base base base base
Reg. 1992-1998 0.059 (0.098) 0.036 (0.099) 0.049 (0.200) 0.018 (0.205) 0.055 (0.106) 0.049 (0.104)
Reg. 1999-2010 0.325*** (0.108) 0.260** (0.107) 0.335 (0.204) 0.264 (0.203) 0.271** (0.129) 0.259** (0.129)
Table 4 (concluded)
Model
Variable all enterprises SMEs large enterprises
(1) (2) (3) (4) (5) (6)
Reg. after 2010 0.463*** 0.355** 0.473* 0.349** -0.018 -0.105
(0.151) (0.154) (0.225) (0.229) (0.271) (0.288)
Foreign owner 0.249* 0.252* 0.247 0.331* 0.322* 0.293*
(0.147) (0.139) (0.219) (0.193) (0.176) (0.176)
State owner -0.109 -0.044 0.350 0.560 -0.286 -0.327
(0.197) (0.205) (0.298) (0.363) (0.229) (0.217)
Export 0.019 0.069 0.245** 0.297*** -0.189** -0.183**
(0.074) (0.075) (0.104) (0.101) (0.088) (0.091)
R&D investment - 0.055 (0.093) - 0.017 (0.138) - 0.108 (0.113)
New/modernised products - -0.133 (0.083) - -0.184 (0.120) - -0.071 (0.101)
New/modernised 0.021 0.068 0.052
technologies (0.086) (0.127) (0.098)
Equipment depreciation - -0.720*** (0.112) - -0.871*** (0.142) - -0.400*** (0.161)
Constant 7.469*** 7.784*** 7.626*** 8.019*** 8.197*** 8.197***
(0.186) (0.187) (0.213) (0.222) (0.163) (0.163)
Industry effect + + + + + +
N 725 725 459 459 266 266
Adjusted R-square 0.10 0.15 0.09 0.16 0.21 0.22
Notes: *** p < 0.01, ** p < 0.05, * p < 0.1.
Robot use. Models (1) and (2) are built on a sample of all observations. The data in column 1 demonstrate a significant positive relationship between robot use in manufacturing and labour productivity. The variables of innovation activity and equipment depreciation included in Model (2) weakened the significance of the key regressor from 1 % of the critical significance level to 10 %, but led to an increase in the significance of enterprise size variables and in the quality of the model as a whole. Thus, other things being equal, the natural logarithm of the robotised enterprises' labour productivity level is 16.2 % higher, and the level of labour productivity itself is 17.6 % higher. Therefore, the basic hypothesis of the study has been confirmed.
At the same time, when analysing Models (3) - (6), which were built in the context of the enterprises' size groups, a significant positive relationship is found between the use of robots and labour productivity for SMEs. According to Models (3) and (4), the natural logarithm of the labour productivity level, caeteris paribus, is higher by 21.6 %. In other words, the average premium for robotisation at small and medium-sized enterprises amounted to 29.8 % to their labour productivity. However, as
indicated by Models (5) and (6), the relationship between robot use and labour productivity at large enterprises is insignificant. Similar results are received in the work by Ballestar et al. [2020], where the use of robots was also part of the binary variable model and appeared to be an insignificant factor in the labour productivity of large businesses. In reality, this may not be the case. As already mentioned, in the larger companies, the level of enterprise robotisation and digitalisation is generally higher than in others, and because of this digital divide, the threshold number of digital technologies, including robots, that can have a profound effect on labour productivity of large companies is greater. The availability of robots is, therefore, of no great matter for large enterprises in terms of their labour productivity: here, it is more important to address the depth of robotisation or even the complexity of robots used in production. Thus, Hypothesis 2 has been confirmed.
Size and age of enterprise. Models (1) and (2) show that the larger the company, the greater its productivity, which can be explained by the market power of larger companies, as well as their wider financial capabilities, technological superiority, and depth of automation. However, the older the enterprise, the less productive it is. Companies registered in the USSR and the post-Soviet period probably still bear the imprint of Soviet inefficiency, which manifested itself in the new economic environment emerged after 1991. As for the small and medium-sized enterprises founded before the 2000s, their lack of scaling over such an extended period indicates their initial inefficiency, which prevents them from transforming from small and medium-sized companies to large businesses. In addition, more "adult" enterprises are characterised by a higher degree of equipment depreciation, and, as confirmed by Models (2), (4) and (6), the higher the percentage of an enterprise's production capacity attributable to equipment older than 10 years, the lower labour productivity. Therefore, in order to boost productivity and maintain competitiveness, companies, regardless of their size, should pay scrupulous attention to the timely renewal of fixed capital.
Owner of enterprise. For enterprises of all sizes, there was found a positive relationship between labour productivity and the presence of foreign individuals or firms among the owners, while there was no significant relationship with the state owners. The presence of a foreign owner provides additional financial opportunities, accelerates entry into new markets, and facilitates the exchange of best practices and new technologies, which ultimately can enhance the competitiveness and performance of the company.
Export. The relationship between exports and labour productivity in the general sampling models is not significant, which is probably due to the overlapping export effects in SMEs and large companies. As indicated by Models (3) and (4), exports are positive and significant, while in Models (5) and (6) the relationship is negative.
The positive correlation between exports and labour productivity is usually contingent on either firms' training through export activities, which allows them to gain in competitiveness, or the production of a premium product that is in demand both in the domestic and foreign markets, or the fact that exporting companies are innately more productive, which allows them to enter the international market. With small and medium-sized businesses, classic export schemes are workable, which is not the case for large enterprises. Probably, their high labour efficiency is mainly based not on competitive advantages, but a monopoly position in the Russian market. Being in a dominant role, such companies have no intention of penetrating foreign markets. Nevertheless, enterprises that do not enjoy such advantages are formally less productive, since they do not have influence on pricing, but at the same time they offer products that are in demand in foreign markets. Moreover, large companies are under certain pressure from the state. In the light of saving jobs, they are constrained in dismissals even if they do not need a numerous staff. At that, high export costs and hiring an excessive number of employees taken together can cause a fall in enterprises' productivity.
Innovation. The innovation activity of the enterprises is insignificant in all the models. On the one hand, this may result from the specificity of business in the Russian Federation, where the growth in enterprises' labour productivity is largely attained not through the introduction of new products or technologies, but through cost cutting via industrial automation. On the other hand, innovations produce long-term effects. In its 2014 Transition Report1, which examined the relationship between innovation and productivity at enterprises, the European Bank for Reconstruction and Development indicated that the correlation between the two factors was weaker than the real influence of innovation on productivity. It takes time for this impact to be fully translated into higher productivity, and the more complex the innovation, the longer it will take. With regard to R&D investment, it is intensity that plays an important role. Based on the analysis of labour productivity growth factors, Simachev et al. [2020] conclude that the positive dynamics of this indicator can be traced for the companies with significant R&D expenditures. In our case, we recorded only the fact of financing, but not the depth of investment.
Conclusion
Using firm-level data from a survey of Russian manufacturing enterprises and linear regression modelling methods, we investigated the relationship between
1 The European Bank for Reconstruction and Development. (2014). Transition Report, pp. 47-69. https://www. ebrd.com/downloads/research/transition/tr14r.pdf. (In Russ.)
the adoption of robots and the level of labour productivity of these enterprises depending on their size.
As hypothesised, a significant and positive relationship between robotisation and labour productivity was found only for small and medium-sized enterprises. The research findings showed that the premium for SMEs robotisation amounted to 29.8 %. In other words, in robotised companies, the average share of revenue (in thousands of rubles) earned by one employee was by 29.8 % more than in companies that did not introduce robots.
The fact that there is no correlation between robot use and labour productivity at large enterprises is probably the result of their higher digitalisation compared to SMEs. Lagging behind in digital transformation allows small and medium-sized businesses to notably increase their labour productivity even if the number of robots adopted is small, while large companies have to automate more business processes to reach this goal due to the scale of their business activities. Consequently, their digitalisa-tion is slower, and if inappropriate technologies are utilised or they are introduced in an incorrect order, automation may fail to bring the desired effect or even weaken efficiency, which will require restructuring other business processes. In this regard, when examining the relationship between robots and labour productivity at large companies, it is necessary to consider not just the fact of robot use, but more comprehensive data, for example, the number of robots in production or the complexity of robotisation technologies.
Thus, when studying the effect of digitalisation on economic performance indicators, the scale of enterprises should be taken into account and different methods applied. For instance, for smaller enterprises, it is sufficient to use binary variables for the use/non-use of robots (including other digital technologies) in production, while for large companies this variable should be more in-depth and contain more qualitative data.
The results of the study can be used when formulating managerial decisions in the field of increasing the efficiency of production through automation. The clear demonstration of the significant and positive relationship between robots and labour productivity in SMEs suggests that small and medium-sized businesses operating in the manufacturing sector need to use robots in order to remain competitive and efficient when competing for a market share.
Appendix 1. Representativeness of the sample by industry groups, %
Clothing
Leather, leather goods, footwear
Woodworking, wood and cork production, excluding furniture, splint production, weaving materials
Cellulose, paper, cardboard and cardboard products Production of coke, oil products Chemicals
Medicines and medical materials Rubber and rubber products Other non-metallic mineral products
Metals production
Finished metal products, excluding machinery and equipment
Computer technology, electronic and optical equipment Electronic machinery and equipment Machinery and equipment not included in other groups
Cars, trailers, semi-trailers Other vehicles and equipment Furniture
Repair and installation of machinery and equipment
0.0
5.0
10.0 15.0 20.0 25.0
All Russian enterprises
Enterprises that participated in the survey
Enterprises from the sample
Appendix 2. Representativeness of the sample by size groups for the industries under consideration, %
100.0
Small (incl. micro) enterprises
Medium-sized enterprises
Large enterprises
All Russian enterprises Enterprises that participated in the survey
Enterprises from the sample
Appendix 3. Descriptive statistics of the companies by robot use
Indicators Enterprises
robotised (150) non-robotised (575)
N % N %
Size of enterprise
Micro 12 8 65 11
Small 41 27 205 36
Medium-sized 27 18 109 19
Large 70 47 196 34
Owner
Foreign 17 11 33 6
State 9 6 16 3
Age
Registered:
before 1992 30 20 82 14
in 1992-1998 48 32 179 31
in 1999-2010 60 40 231 40
after 2011 12 8 83 14
Export
Exporting 51 34 191 33
Non-exporting 99 66 384 67
R&D investment
Investing 44 29 145 25
Non-investing 106 71 430 75
Introduction of new/modernised products
Introducing 83 55 295 51
Non-introducing 67 45 280 49
Appendix 3 (concluded)
Indicators Enterprises
robotised (150) non-robotised (575)
N % N %
Introduction of new/modernised technologies
Introducing 70 47 190 33
Non-introducing 80 53 385 67
Manufacturing industries
Food 19 13 149 26
Textile 4 3 9 2
Clothing 4 3 20 3
Leather, leather goods, footwear 2 1 10 2
Woodworking, wood and cork production, excluding furniture, splint production, weaving materials 2 1 23 4
Cellulose, paper, cardboard and cardboard products 10 7 13 2
Production of coke, oil products 0 0 2 0
Chemicals 4 3 31 5
Medicines and medical materials 4 3 13 2
Rubber and rubber products 11 7 39 7
Other non-metallic mineral products 12 8 52 9
Metals production 2 1 10 2
Finished metal products, excluding machinery and equipment 16 11 44 8
Computer technology, electronic and optical equipment 9 6 18 3
Electronic machinery and equipment 14 9 25 4
Machinery and equipment not included in other groups 10 7 41 7
Cars, trailers, semi-trailers 5 3 17 3
Other vehicles and equipment 9 6 10 2
Furniture 7 5 21 4
Repair and installation of machinery and equipment 6 4 28 5
Indicators Min Quarter 1 Median Mean Quarter 3 Max
Robotised enterprises
Labour productivity, ln 5.7 7.5 8.1 8.2 8.9 10.4
Share of production capacity generated by equipment older than 10 years 0.0 0.0 0.2 0.2 0.4 1.0
Non-robotised enterprises
Labour productivity, ln 5.4 7.2 7.9 7.9 8.6 11.1
Share of production capacity generated by equipment older than 10 years 0.0 0.0 0.2 0.3 0.7 1.0
Appendix 4. List of variables
Indicators Description
Ln (labour productivity) Dependent variable. The ratio of revenue (thousand rubles) and the number of employees (ln)
Robot use Key regressor. Using robots in production (1 - yes, 0 - no)
Micro enterprises Less than 16 employees (1 - yes, 0 - no)
Small enterprises Between 16 and 100 employees (1 - yes, 0 - no)
Medium-sized enterprises Between 101 and 250 employees (1 - yes, 0 - no)
Large enterprises Basic variable in category. More than 250 employees (1 - yes, 0 - no)
Reg. USSR Registered before 1992 (1 - yes, 0 - no)
Reg. 1992-1998 Registered in 1992-1998 (1 - yes, 0 - no)
Reg. 1999-2010 Basic variable in category. Registered in 1999-2010 (1 - yes, 0 - no)
Reg. after 2010 Registered after 2011 (1 - yes, 0 - no)
Foreign owner Among owners there are foreign individuals and companies (1 - yes, 0 - no)
State owner Among owners there are federal, regional and/or local authorities (1 - yes, 0 - no)
Export Performing export activities in 2017 (1 - yes, 0 - no)
R&D investment Investing in R&D in 2017 (1 - yes, 0 - no)
New/modernised products Introduction of new/modernised products (1 - yes, 0 - no)
New/modernised technologies Introduction of new/modernised technologies (1 - yes, 0 - no)
Equipment depreciation Share of total production capacity generated by machinery and equipment aged 11 years and older
Manufacturing industry Including all the manufacturing industries in the model using the binary method; the basis is the food industry
References
Arkhipova L. S., Melnikova D. M. (2022). Assessment of modern barriers affecting digitalization Russian labor market. Regionalnaya ekonomika i upravlenie: elektronnyy nauchnyy zhurnal = Regional Economics and Management: Electronic Scientific Journal, no. 2 (70), 7002. DOI: 10.24412/19992645-2022-270-2. (In Russ.)
Budkova S. V. (2015). Factors defining development of the export focused small and medium-sized enterprises. Sotsialno-ekonomicheskie yavleniya i protsessy = Social-Economic Phenomena and Processes, no. 6, pp. 24-29. (In Russ.)
Garshina O. P. (2011). The impact of innovations on labour productivity growth at industrial enterprises. Vestnik Samarskogo gosudarstvennogo universiteta = Vestnik of Samara State University, no. 6 (87), pp. 19-25. (In Russ.)
Gokhberg L. M., Rudnik P. B., Vishnevskiy K. O., Zinina T. S. (eds.). (2021). Digital transformation of industries: Starting conditions and priorities. HSE University Report. Moscow: HSE University. 239 p. (In Russ.)
Gurlev I. V. (2020). Digitalization of the Russian economy and problems of robotics. Vestnik Evrazi-yskoy nauki = The Eurasian Scientific Journal, no. 4 (12). https://esj.today/PDF/08ECVN420.pdf. (In Russ.)
Ermolov I. L. (2019). Role of industrial robots in perspectives of Russian economy. Innovatsii = Innovations, no. 10, pp. 127-129. DOI: 10.26310/2071-3010.2019.252.10.015. (In Russ.)
Zemtsov S. P. (2017). Robots and potential technological unemployment in the Russian regions: Review and preliminary results. Voprosy ekonomiki = The Issues of Economics, no. 7, pp. 142-157. DOI: 10.32609/0042-8736-2017-7-142-157. (In Russ.)
Igoshina D. R. (2021). Features of digitalization of business processes in companies of small and medium entrepreneurship. Industrialnaya ekonomika = Industrial Economics, vol. 11, no. 5, pp. 1092-1097. DOI: 10.47576/2712-7559_2021_5_11_1092. (In Russ.)
Komkov N. I., Bondareva N. N. (2016). The perspectives and the conditions of the robotics development in Russia. Modernizatsiya. Innovatsii. Razvitie = MIR (Modernization. Innovation. Research), vol. 7, no. 2, pp. 8-21. DOI: 10.18184/2079-4665.2016.7.2.8.21. (In Russ.)
Kosyakova L. N., Popova A. L. (2017). Tasks of improving labour productivity in Russia and ways to fulfill them. Izvestiya Sankt-Peterburgskogo gosudarstvennogo agrarnogo universiteta = Bulletin of the Saint Petersburg State Agrarian University, no. 3 (48), pp. 153-157. (In Russ.)
Nazarenko A. A. (2021). The spread of digital technologies in small and medium-sized enterprises in the Russian Federation. Voprosy innovatsionnoy ekonomiki = Russian Journal of Innovation Economics, vol. 11, no. 4, pp. 1439-1450. DOI: 10.18334/vinec.11.4.113671. (In Russ.)
Simachev Yu. V., Kuzyk M. G., Fedyunina A. A., Yurevich M. A. (2020). Labor productivity in Russian companies: How to foster sustainable growth. Zhurnal Novoy Ekonomicheskoy Assotsiatsii = The Journal of the New Economic Association, no. 4 (48), pp. 205-217. DOI: 10.31737/2221-22642020-48-4-10. (In Russ.)
Tolkachev S. A., Kulakov A. D. (2016). Robotization as the direction of neoindustrialization (on the example of the USA). Mir novoy ekonomiki = The World of New Economy, no. 2, pp. 79-87. (In Russ.)
Trachuk A. V., Linder N. V. (2017). Innovations and productivity of the Russian industrial companies. Innovatsii = Innovations, no. 4 (222), pp. 53-65. (In Russ.)
Urunov A. A., Rodina I. B. (2018). Influence of artificial intelligence and the Internet - technologies on national labor market. Fundamentalnye issledovaniya = Fundamental Research, no. 1, pp. 138-142. (In Russ.)
Acemoglu D., Lelarge C., Restrepo P. (2020). Competing with robots: Firm-level evidence from France. AEA Papers and Proceedings, vol. 110, pp. 383-388. DOI: 10.1257/pandp.20201003.
Acemoglu D., Restrepo P. (2020). Robots and jobs: Evidence from US labor markets. Journal of Political Economy, vol. 128, no. 6, pp. 2188-2244. DOI: 10.1086/705716.
Ballestar M. T., Diaz-Chaob A., Sainzb J., Torrent-Sellens J. (2020). Knowledge, robots and productivity in SMEs: Explaining the second digital wave. Journal of Business Research, vol. 108, pp. 119-131. https://doi.org/10.1016/j.jbusres.2019.11.017.
Berg A., Buffie E. F., Zanna L.-F. (2018). Should we fear the robot revolution? (The correct answer is yes). Journal of Monetary Economics, vol. 97, pp. 117-148. DOI: 10.1016/j.jmoneco.2018.05.014.
Bonfiglioli A., Crino R., Fadinger H., Gancia G. (2020). Robot imports and firm-level outcomes (CESifo Working Paper no. 8741). Munich Society for the Promotion of Economic Research. 48 p.
Cette G., Devillard A., Spiezia V. (2021). The contribution of robots to productivity growth in 30 OECD countries over 1975-2019. Economics Letters, vol. 200, 109762. DOI: 10.1016/j.econ-let.2021.109762.
Deng L., Plümpe V., Stegmaier J. (2021). Robot adoption at German plants (IWH Discussion Papers no. 19/2020). Leibniz-Institut für Wirtschaftsforschung Halle. 35 p.
Dixon J., Hong B., Wu L. (2021). The robot revolution: Managerial and employment consequences for firms. Management Science, vol. 67, no. 9, pp. 5586-5605. DOI: 10.1287/mnsc.2020.3812.
Dottori D. (2020). Robots and employment: Evidence from Italy (Bank of Italy Occasional Paper no. 572). Banca d'Italia. 61 p. http://dx.doi.org/10.2139/ssrn.3680743.
Graetz G., Michaels G. (2018). Robots at work. The Review of Economics and Statistics, vol. 100, no. 5, pp. 753-768. D0I:10.1162/rest_a_00754.
Jungmittag A., Pesole A. (2019). The impact of robots on labour productivity: A panel data approach covering 9 industries and 12 countries (JRC Working Papers on Labour, Education and Technology no. 2019-08). Joint Research Centre (JRC), European Commission. 27 p.
Klenert D., Fernandez-Macias E., Anton Perez J. I. (2020). Do robots really destroy jobs? Evidence from Europe (JRC Working Papers Series on Labour, Education and Technology, no. 2020/01). Joint Research Centre (JRC), European Commission. 51 p.
Koch M., Manalo I., Smolka M. (2019). Robots and firms (CESifo Working Paper no. 7608). Munich Society for the Promotion of Economic Research. 47 p.
Stiebale J., Suedekum J., Woessner N. (2020). Robots and the rise of European superstar firms (DICE Discussion Paper no. 347). Heinrich Heine University Düsseldorf, Düsseldorf Institute for Competition Economics (DICE). 60 p.
Syverson C. (2011). What determines productivity? Journal of Economic Literature, vol. 49, no. 2, pp. 326-365. DOI: 10.1257/jel.49.2.326.
Vries G. J., de, Gentile E., Miroudot S., Wacker K. M. (2020). The rise of robots and the fall of routine jobs. Labour Economics, vol. 66, 101885. DOI: 10.1016/j.labeco.2020.101885.
Information about the author
Daria A. Starovatova, Research Intern of the research group "Economics of Robotisation of Industries and Firms" in the Center for Structural Policy Research. HSE University, Saint Petersburg, Russia. E-mail: darastarovatova@gmail.com
© Starovatova D. A., 2023