Big Data Analysis for Management from Solow’s Paradox Perspective in Polish Industry

Piotr Jatowiecki

Warsaw University of Life Sciences—SCCW

Introduction

Technological progress is most often understood as a process that leads to an increase in productivity because the same volume of production can be produced using smaller inputs of other factors of production or with the same inputs it is possible to increase production. Increases in capital and labor in the modified Solow—Swan model lead only to a time increase, which lasts until the economy reaches a new state of steady-state growth, while permanent growth is ensured only by technological changes (Solow, 1957). With this approach, technological progress in the product of the production function served as the “rest” that aggregates all factors affecting product growth except for capital and labor as an additional factor shaping production growth. It was called “the rest of Solow”, but is better known as the Total Productivity Factor (TPF). It is defined as the product of the level of technological progress and the rate of technology growth. It should be emphasized that other ways to take into account technological progress within the production function were also sought, assuming that it can supply capital growth or increase of work.

In 1967, Denison distinguished three main components of TPF: (1) shifts of resources from agriculture to industry; (2) progress in knowledge and technical, organizational, and management skills; and (3) so-called “residual productivity”. Also, Solow suggested that technological progress is included only in new investments and that it has an impact primarily on improving the quality of work. Based on empirical studies for the US economy in the first half of the 20th century, he estimated the size of the rest of Solow at around 87%. This meant that the explanation of economic growth to a large extent does not result from the accumulation of capital and labor. Tire modified Solow—Swan model shows that the development of information and communication technologies (ICT) should lead to increased productivity. First of all, investments and implementation of modern ICT should lead to an increase of technical equipment of work, and thus to increase its efficiency. Secondly, technological progress is transferred, as it were, automatically from the ICT industry to other industries thanks to lower prices of goods, which should lead to lower production costs. Thirdly, ICT is a general-purpose universal technology and, therefore, generates significant non-financial external effects also known as spillover effects.

However, in the ‘70s and ‘80s of the last century, the results of empirical research on the state of the economy in the United States showed a decline in productivity despite very significant and growing investments in modern technologies of information processing and transmission. The obtained results did not confirm the existence of a positive relationship between the implementation of modern ICT and the increase in productivity. Also, it turned out that the greatest drop in productivity occurred in the sectors of the economy in which investments in ICT were the largest, and the share in the employment of people serving them was the highest (Brynjolfsson, 1993). Therefore, already in 1987, Robert Solow formulated his famous paradox of productivity, noting that investments in ever more modern ICT solutions being a catalyst for changes in almost all areas of life, do not translate into an increase in productivity both on the micro- and macroeconomic scales.

Since the formulation of the productivity paradox by Solow, many attempts have been made to explain it. They can be divided into two main trends. As part of the first one, not rejecting neoclassical theories of growth, in particular, the modified Solow-Swan model, various factors are being sought that affect the lack of translating investments into ICT solutions into the financial results of enterprises. The second trend is connected with new theories and models of product growth, in which the possibility of eliminating imperfections found in the Solow—Swan model itself is sought. Thus, in the mid-1990s, four main groups of arguments explaining the existence of the productivity paradox were identified: (1) problems resulting from the lack of accuracy in the measurement of enterprises’ outlays and results, (2) profit redistribution, (3) adjustment processes that the economy must undergo, and (4) lack or insufficient investment complementary to investment in modern ICT (Milgrom and Roberts, 1995; Jorgenson, 2001) (Brynjolfsson and Saunders, 2010; Cardona et al., 2013).

Research on the occurrence of the Solow productivity paradox in various countries in the 21st century led to the conclusion that it is much more visible in less developed countries, but also that the period of the economy’s adaptation to ICT, followed by an increase in productivity is shorter in more developed countries. Also, in the United States and EU, in 1990-2008, three factors that have the strongest influence on these differences were distinguished: faster development of the service sector, faster technological progress, and larger sizes of ICT solutions providers. Only in recent years, the differences in the impact of the use of advanced ICT technologies on the increase in productivity between the US economy and the economies of the leading EU countries have started to decrease significantly (Jorgenson and Vu, 2005; Dedrick et al., 2013; Niebel, 2014).

The purpose of the research, the results of which were presented in the work, was to identify sources and the frequency of using Big Data in Polish enterprises, taking into account their division into employment size groups, PKD sections, and voivodeships. The impact of using Big Data for business management purposes and the generation of PKD in the examined groups of business entities was also examined. Which discusses issues related to the Solow productivity paradox and attempts to explain it so far, this chapter includes a chapter containing the most important information about Big Data, a chapter which briefly describes the objectives and scope of research, source data, and applied research methodology, as well as chapter with research results and the last chapter with conclusions resulting from research. Because data on the use of Big Data in Polish enterprises and economic characteristics of enterprise groups are also available in aggregate form from previous years, further research will be conducted to identify development trends of both the use of Big Data as well as to assess changes in the strength of the relationship between them and selected characteristics economic.

Big Data

Nowadays, unprecedented progress can be observed in the field of digital technologies, primarily ICT, both in the implementation of ever new solutions and systems in the field of information management, as well as their dynamic development consisting in their rapid dissemination and, consequently, their use in virtually every field of science, economy as well as social and private life.

The dynamics of the increase in computing power of computers are best described by Moore’s Law formulated in 1965 and modified by the author 10 years later. According to it, the number of transistors placed in microprocessors doubles every 18 months, and now, after Moore corrected the law in 1975, every 24 months. However, as early as 1999, Moore corrected his law again, predicting that in the recent past the rate of doubling the number of transistors in chips would slow down to every 4-5 years (Sienkiewicz, 2005). Even later Moore’s law was slightly reformulated, and instead of the number of transistors, the computing power of computers became an element of doubling. By analogy with the original version of Moore’s law, similar rules apply to the amount of RAM, hard disk capacity, bandwidth in computer networks, and above all the ratio of computing power to the cost of obtaining it.

Regardless of how Moore’s law is formulated, the inevitable extension of the doubling period of the number of transistors placed in integrated circuits should be expected because its course depends on the technological possibilities in the field of miniaturization of transistors. According to International Technology Roadmap for Semiconductors, in 2009, the average transistor size was 32 nm; in 2012, 22 nm; in 2015, 14 nm; in 2018, 10 nm; in 2019, 7 nm; and in 2020, it will be approx. 5 nm, while in 2022, about 3 nm. However, the further development of this miniaturization trend is under question due to technological limitations.

As a consequence, the focus on the development of ICT has moved to two other areas. The first of them, certainly very promising, are the modern technologies of the future: quantum computers, biocomputers, recording information on crystal media, etc., which does not change the fact that it must probably take a long time before they are brought into a form enabling their production, implementation, and widespread use. The second direction is already successfully implemented and is a derivative of two relatively well-known and widespread technologies: high-speed and ubiquitous Internet and multiprocessor systems. Their widespread use has led to the convergence sociosphere, infosphere, and technosphere, and consequently to the formation of the so-called cyberspace, which Castells (1990) originally referred to as the “Internet Galaxy”. Within this cyberspace, different secondary technologies, for example, e-commerce, e-business, e-administration, social media, or just a Big Data are formed and rapidly disseminate.

Nowadays, the understanding of this term has been extended by two additional aspects: variability and complexity. Especially the latter feature is very characteristic for modern Big Data sets, and consequently for their management methods and their analysis. Nowadays, Big Data collections are created in a completely automatic way based on detailed source data from very many, very diverse sources.

There are quite a large number of Big Data classifications, but the most frequently used in practice are classifications based on their origin. The most general Big Data classification covers five categories: (1) public data, (2) private data, (3) data exhaust, (4) community data, and (5) self-quantification data. The source of public data is government and self-governing institutions and organizations. In turn, private data comes from business, non-profit organizations, and individuals. A specific category of Big Data is data exhaust, that is, for those who generate them have no meaning or is it negligible, they are collected almost automatically, and the gain value only in conjunction with other types of data. Community or social data is a category of unstructured data, most often text data from social media, and primarily concerns user preferences. Self-quantification is data collected automatically from various Internet and mobile applications regarding user behavior. They often arise as a result of monitoring, e.g., its location (Kennedy, 2008; George et al., 2014).

Big Data sets are used in many different areas, among others, to prepare management information, i.e., information that can be used directly in decision-making processes. Of course, depending on the specific area of human activity, decisionmaking processes will be very diverse, as a consequence, the form of management information will be varied, as well as the sources and form of Big Data used to prepare it. Different areas of business activity can be a good example of this. It should be noted that depending on the specific area of different profiles are the source of their origin. Enterprises operating in various areas of the economy and using Big Data sets for business analysis can be a good example. Each area of the economy has a different profile of sources from which Big Data is obtained (see Table 13-1). In this study, the division of primary sources of Big Data were categorized into three most popular main groups: I—sensors and smart Al (Artificial Intelligence) devices, e.g., using communication between devices (M2M machines), digital sensors, radio-frequency identification (RFID) labels, etc.; II—geolocation data obtained from mobile devices, e.g., from mobile devices using telecommunications networks, wireless connections, or GPS; and III—data generated by social media, e.g., social networking sites, blogs, and websites used to exchange multimedia information.

The other primary sources of Big Data were grouped in the last fourth category (IV). Of course, a large number of companies executing Big Data analysis used more than one category of its primary data (see Table 13-1). Another important aspect of using Big Data for business analysis is performing them using the company’s resources or outsourcing them to specialized external entities. Also, in this respect, different solutions are preferred in different areas of the economy (see Table 13-1).

Hie use of Big Data for business analyses partly or in the vast majority automated way is a combination of modern ICT technologies and methods of managing, above all, customer relationship management (CRM) area. Thanks to the use of Big Data, it is possible to personalize the approach to the customer very effectively and virtually fully automated, as well as to adapt the services and products offered to his individual needs and expectations (Romika Yadav and Tarun Kumar, 2015; Anshari et al., 2019). Big Data most often in aggregate form are also used in macroeconomic analyses, enabling overall tracking of preferences, demand trends, and consequently modification of the product basket, effective product basket management, conducting effective marketing policy, as well as the use of business intelligence technology (defining key performance index, creating dashboards, etc.) (Chen et al., 2012; Elgendy and Elragal, 2014; Hu, 2018).

Table 13.1 Percentages of Enterprises Using Big Data for Business Analyses by Employment Size Groups and PKD 2007 Sections

Source: Own preparation based on of GUS data.

182 ■ Management in the Era of Big Data