Innovation and Productivity in SMEs and Large Businesses in the United Kingdom

(1)

Innovation and Productivity in SMEs and

Large Businesses in the United Kingdom

Master Thesis

Name: Eva Svobodová Student number: S3690148 Track: MSc BA-SB&E

Supervisor: Dr Maria Kristalova Co-assessor: PD Dr Michael Wyrwich Date: 22 June 2020

Word count: 8354

(2)

1

Abstract

This research aims to examine the relationship of innovation and productivity within SMEs and large businesses in the United Kingdom, as this relationship is often studied within large businesses only. Furthermore, it aims to bring more clarity to investors and managers on whether investment into innovation increases the productivity of the company and, therefore, its further economic growth. After analysing the sample of 1299 SMEs and 711 large businesses from across manufacturing and service industries in the UK, the results show that innovation input is positively related to productivity with all companies in the sample. This result is likely driven by a significant relationship within SMEs. The relationship between innovation input and productivity in large businesses, as a separately studied subsample, is not significant. The results for the relationship between innovation output and productivity are also insignificant. In order to find a better comparison, the relationship between innovation and productivity in SMEs compared to large businesses has to be studied further. Key words: innovation, labour productivity, R&D spending, patents, innovation input,

(3)

2

Table of Figures and Tables

Figure 1: Conceptual Model ... 12

Table 1: Descriptive Statistics ... 18

Table 2: Frequencies ... 18

Table 3: Summary of Regression Analysis for Hypothesis 1 ... 20

Table 4: Summary of Regression Analysis for Hypothesis 2 ... 21

(4)

3

Introduction

Productivity and its determinants are often given plenty of attention by scholars and policy makers as it is generally related to higher profits and overall economic growth (Hall, 2011; Crowley and McCann, 2015; Tsvetkova et al., 2020). Productivity goes hand in hand with the innovation, as innovation leads to higher productivity and therefore enhances economic growth (Hall, 2011; Mohnen and Hall, 2013). Most of the studies focus only on the impact of innovation on productivity in overall economies (e.g. Bahar, 2016; Ahmad, 2018) or on this relationship within the large businesses (e.g. Bond and Guceri, 2016). However, there are only a few studies that focus on the innovation-productivity relationship in small and medium sized enterprises (SMEs) (e.g. Hall et al., 2009; Baumann and Kritikos 2016). It can be argued that larger firms are more innovative with their high investments in a range of various activities (Hall, 2011), where, in fact, SMEs are the engines of technological changes and innovative activities in certain industries (Hall et al., 2009). Ultimately, SMEs are drivers of the economy. They are the main sources of employment, accounting for about 70 per cent of jobs on average in all OECD countries. In addition, they are generating 50 to 60 per cent of value added on average, as they are major contributors to value creation (OECD, 2017). Despite playing a big role in economic development SMEs often face lack of financial resources from venture capitalists who are capable of valuing R&D projects. Therefore, they are often reliant on bank capital market systems and its credit rationing problems (Hall et al., 2009).

This research contributes to rather limited studies on the innovation-productivity relationship in SMEs. In addition, it aims to bring more clarity to investors and managers of SMEs about whether the investment into innovation in SMEs increases the productivity of the company and, therefore, its further economic growth. Moreover, this research looks into productivities of SMEs and large businesses separately, based on their innovation inputs and innovation outputs. Consequently, the objective of this research is to find the answer to following research question:

What is the impact of innovation on productivity in SMEs and large businesses?

(5)

4

by studying the relationship between innovation and labour productivity in four countries, one of which is United Kingdom (UK). Another study by Hall (2011) finds that product innovation has a significant effect on productivity, whereas process innovation does not. These are just a few ground-breaking studies, which are examined in more depth in the literature review.

In so far as SMEs drive the economies the UK economy is not an exception. A vast number of businesses are registered in the UK, thanks to the benefits provided by the country. Examples of such benefits include generous corporate tax rates and various investment schemes incentivised by the UK Government (Gov.uk, 2020). The benefits luring new businesses in helped to continuously grow the UK economy before Brexit (Badiani, 2019). A country with strong economic growth likely has productive companies and therefore provides value to this research. An additional reason is the availability of the data that comes with the high number of businesses mandatorily registered in the UK.

This research contributes further to the existing literature by analysing the sample of 1299 SMEs and 711 large businesses from across manufacturing and service industries in the UK. The results show a positive correlation between innovation input and productivity in all companies in the sample, especially SMEs. This result is likely driven by the significant relationship shown in the SMEs subsample. The relationship between innovation input and productivity in large businesses, as in the separately studied subsample, is not significant. The results for the relationship between innovation output and productivity are also insignificant. The following section of this research is a literature review, where the main definitions are described, hypotheses are formulated, and the conceptual model is presented. The theory section is followed by a methodology one, with a focus on data and their analysis. The last sections of this research engage in the description of results of analysis, discussion and conclusion.

Literature Review

Productivity

(6)

5

that is necessary to produce a certain level of output (Hall, 2011). According to Syverson (2011), productivity is quite literally a matter of business’ survival.

As each firm is a heterogeneous agent, it can be expected that their productivity levels differ based on their individual characteristics (Bahar, 2016), such as the level of technological knowledge, organization, size, and other environmental factors, as competition, that influence the productivity of companies (Hall, 2011). One of these factors is innovation, which is also one of the most commonly studied factors in regards to productivity (Álvarez et al. 2013; Chudkovsky et al., 2006; Dutch-Brown et al., 2018; Griffith et al., 2006; Hall et al., 2009). Innovation

Hall (2011) defines innovation, based on the Oslo Manual (OECD, 2005), as “the implementation of a new or significantly improved product (good or service), or process, a new marketing method, or a new organisational method in business practices, workplace organisation or external relations.” The subjective element of the definition is the level of ‘newness’ which is often interpreted differently. Based on the degree of newness, an innovation can be classified as radical or incremental (Schilling, 2017). However, radical innovations may be more important than incremental (Hall, 2011). In addition to the degree of newness there are also different dimensions of innovation: technological innovation, which includes product and process innovations, and non-technological innovation, which includes organizational and marketing innovations. Each of these brings different value to the productivity. However, the results of different studies are often contradictory.

Furthermore, according to Hall (2011) there are only two reasons for innovation: 1) to increase demand for better goods, and 2) to reduce the number of inefficient competitors. The competition is the biggest driver of innovation. The presence of international competitors is often strongly and positively related to R&D efforts. Additionally, engaging in exporting activities implies higher investment in R&D (Hall et al., 2009).

(7)

6

firms are more likely to carry out R&D, but smaller firms allocate more resources to R&D (Duch-Brown et al., 2018). A study in Catalonia concludes that R&D intensity has, in fact, an effect on labour productivity in both service and manufacturing firms (Segarra-Blasco, 2010). Many studies that study relationship between innovation and productivity use the model of Crépon et al. (1998), better known as the CDM model. Instead of studying the impacts of R&D on the productivity of companies directly, the CDM model has four stages: 1) the decision by a firm to engage in innovative activities, 2) the amount of resources (innovation input) that are allocated to the innovative activities once it decides to engage in them, 3) the effect of innovation input on innovation output, and 4) the effect of innovation output on productivity (Crépon et al, 1998). The studies that use the CDM model find strong relationship between R&D expenditure, innovation output and firm performance (Teplykh, 2014).

Innovation Input

Productivity is assumed to be dependent on innovation, and innovation is dependent on investment choices (Hall et al., 2013). This sums up the close relationship between innovation input and output. Investment into R&D helps drive innovation which in turn has a large impact on productivity (Hall et al., 2013). In another study, Hall (2011) concludes that innovative activity generally increases a firm’s ability to derive revenue from its inputs (Hall, 2011). Consequently, it means that innovative input has a positive effect on revenue productivity.

Furthermore, Hall et al. (2009) studied Italian SMEs and compared them to the study of Griffith et al. (2006), who studied firms in France, Germany, Spain, and the UK. The results of both studies show that investment into innovation has a positive impact on productivity. Interestingly, the productivity of the Italian firms is higher compared to those studied by Griffith et al. (2006). Especially for smaller SMEs that tend to invest more in R&D than their larger counterparts, which invest into a range of R&D activities (Hall et al., 2009).

(8)

7

evidence from industrialised countries. Small firms follow different innovation strategies compared to large firms that can carry out research and innovate with external support. SMEs, and companies in the service sector, depend on interaction with customers rather than on formal R&D. They conclude that innovation expenditure has significant and positive impact on labour productivity in both service and manufacturing industries, with the impact being larger in small firms (Aboal and Garda, 2015).

Bond and Guceri (2016) find that R&D spending tends to be followed by an increase in productivity in larger British manufacturing establishments. Teplykh (2014) studies the relationship between innovation and productivity in Western European manufacturing firms in Germany, Italy, France, Spain, and the UK during 2008 crisis. He finds that involvement in R&D is a significant factor for productivity in pre- and post-crisis periods.

Additionally, Álvarez et al. (2013) find that innovation input and output bring improvements to productivity in manufacturing and service sectors. Surprisingly, the service sector has a higher level of expenditure than manufacturing. However, innovation expenditure is still a more important determinant of labour productivity in the manufacturing sector (Álvarez et al., 2013). Baumann and Kritikos (2016) also find that manufacturing micro firms in Germany have a similar link between R&D, innovation, and productivity when compared to their larger counterparts. In addition, they report that every second micro firm is trying to innovate, and half of these firm do so without formal R&D expenditure. Finally, Tsvetkova et al. (2020) summarize that firms’ R&D investment explains a large part of their differences in productivity.

(9)

8

Furthermore, local conditions play an important role in the innovation and productivity relationship (Griffith et al., 2006; Syverson, 2011; Crowley and McCann, 2015; Duch-Brown et al., 2018; Tsvetkova et al., 2020). Depending on local conditions, various scientific and technological opportunities exist which could influence this relationship even further (Crespi and Pianta, 2008; Teplykh, 2014). Such opportunities can include knowledge spill-overs (Crespi and Pianta, 2008; Castellacci 2009) or even close proximity to a university, which can help SMEs with conducting R&D (Tsvetkova et al., 2020).

As can be seen earlier, it goes without saying that competition that is a driver of innovation on its own (Crespi and Pianta, 2008; Hall et al., 2009) as well as industry play an important role in the innovation input and productivity relationship (Aboal and Garda, 2015; Crowley and McCann, 2015). The last underlying factors influencing this relationship are labour quality (Syverson, 2011; Teplykh, 2014; Tsvetkova et al., 2020), access to funds (Crowley and McCann, 2015; Baumann and Kritikos, 2016), demand (Teplykh, 2014), and a firm’s willingness to invest in innovation (Teplykh, 2014).

Based on the evidence from the different studies that spread over various sectors and countries, the first hypothesis is formulated.

H1: Innovation input increases productivity.

H1a: Innovation input increases productivity in SMEs.

H1b: Innovation input increases productivity in large businesses.

Innovation Output

As it was previously recognised, innovation input is not always measured as R&D because of insufficient reporting by the companies. Therefore, some studies of the innovation-productivity relationship moved to an output approach instead of an input approach (Hall et al., 2009), while some use both approaches (e.g. Álvarez et al., 2013; Baumann and Kritikos, 2016). They compare different innovation types (i.e., product, process, marketing, and organizational) considered as innovation output and their effects on productivity across various industries.

(10)

9

innovation is rather unclear. Crespi and Pianta (2008) find an increase in productivity through product innovation when technological competitiveness is present, especially in Germany, the Netherlands, and the UK. Furthermore, the presence of active price competitiveness in process innovation increases productivity too, however, the fit is better for countries such as France, Italy, and Portugal. However, only small differences were found in productivity growth between manufacturing and service industries (Crespi and Pianta, 2008).

The study of Baumann and Kritikos (2016), into German manufacturing micro firms, found that micro firms engaging in product innovation have a stronger effect on their productivity levels. The process innovation is found to have a negative effect on productivity in micro firms (Baumann and Kritikos, 2016).

Hall and Sena (2017) look at the impact of protection of intellectual property (IP) on the productivity of firms. They found that companies that formally protect their IP are more productive, as the firms that opt for formal IP methods may signal their profitability and long-term viability to investors. Formal protection is also more common in large firms. However, SMEs have greater need to access external inputs and, therefore, they need to protect their knowledge more formally (Hall and Sena, 2017). On a related note, Balasubramanian and Sivadasan (2011) found that patent grants can increase the productivity of a company; however, the supporting evidence is weaker.

(11)

10

Furthermore, they found that all types of innovations have a positive effect on productivity levels of non-innovating firms (Crowley and McCann, 2015).

Another study that links types of innovation to productivity is the study of Tevdovski et al. (2017). The study compares the productivity of Bulgarian and Romanian companies to those in Germany. The results of the study show that process innovation is associated with an increase of labour productivity in Germany and Bulgaria, but not in Romania. Product innovation leads to higher productivity in all three countries. The introduction of organizational innovation seems to be important in all three countries, where marketing innovation is less important to the growth of productivity (Tevdovski et al., 2017).

In their study of French manufacturing and service firms between 1998-2000 and 2002-2004, Mairesse and Robin (2010) find that product innovation has a significant effect on productivity. The process innovation has no effect. Another study comparing manufacturing and service sectors by Criscuolo (2009) found that product innovation has a larger effect on productivity in the manufacturing sector than in the service sector. Álvarez et al. (2013) and Aboal and Garda (2015) found that productivity in the service sector is positively related to technological and non-technological innovations, where non-technological innovations have more impact. In the manufacturing sector, technological innovation is more relevant for productivity growth (Álvarez et al., 2013; Aboal and Garda, 2015). In another study, Aboal and Tacsir (2018) confirm these findings. Huergo and Jaumandreu (2004) found that new firms in the Spanish manufacturing market experience high productivity growth, which has progressively weakened over the years. At some point, process innovation leads to extra productivity growth, which tends to persist for a number of years. The relationship between process innovation and productivity of these firms is strong to the point that if process innovation stops, the productivity growth will come to an end (Huergo and Jaumandreu, 2004).

(12)

11

found that the relationship is even stronger in more competitive, rather than oligopolistic, industries.

The relationship between innovation output and productivity is also influenced by several underlying factors. The first factor, firm size, positively influences this relationship, as Crowley and McCann (2015) argue. Baumann and Kritikos (2016) again argue that micro firms between 15-35 years of age are less likely to be successful in innovation and that other larger SMEs, which are older than 35 years old are more likely to succeed with their innovations and thus be more productive. Furthermore, complementarities between the innovation types mentioned by Mohnen and Hall (2013) play a significant role in this relationship too.

Additionally, innovation output is driven by demand (Duch-Brown et al., 2018) that then drives competition. As previously mentioned competition drives innovation and furthers the productivity. This is especially important for process innovation (Crespi and Pianta, 2008; Tsvetkova et al., 2020). The relationship between innovation output and productivity is dependent on the industry as well. Álvarez et al. (2013) prove that innovation output increases productivity depending on the sector. Lastly, Huergo and Jaumandreu (2004) argue that bringing continuous innovation output helps to increase the productivity of companies, as stated earlier.

A significant amount of evidence points out that the effects of innovation output on productivity differs across industries, depending on the innovation type. However, most of the relationships are concluded to be positive. Regardless of the differences in the evidence, the second hypothesis is formulated as followed.

H2: Innovation output increases productivity.

H2a: Innovation output increases productivity in SMEs.

H2b: Innovation output increases productivity in large businesses.

Conceptual Model

(13)

12

Methodology

Data Collection

This research has used quantitative data collected from ORBIS database. The database, provided by Bureau van Dijk, includes detail information from financial statement and balance sheets of public and private firms collected from business registers (Kalemli-Özcan et al., 2015). With its various variables and wide range of collected data, this database is suitable for firm-level analysis of productivity (Gal, 2013). The database that has been used for this research has been downloaded in June 2020.

Since part of the goal of this research is to compare productivity between SMEs and large businesses, the sample has been constructed of 1299 SMEs and 711 large businesses from manufacturing and service industries in the UK. The country of analysis, the UK, has been chosen because of the continuous growth of its economy before Brexit, thus likely presence of productivity. The secondary reason is the large body of data available in ORBIS. SMEs have been defined by the European Commission (2012) as companies with fewer than 250 employees and a turnover of €50 million or less. Thus, for the purposes of this research, large businesses have been defined as firms with 250 or more employees and a turnover higher than €50 million. The sectors of the industries have been identified using 2-digit Nomenclature of Economic Activities (NACE) code in ORBIS, based on the study of Ahmad et al. (2018). The NACE codes for manufacturing industries are 10-33 and for service

Innovation Input Innovation Output

Firm Productivity Control Variables: Firm Size Firm Age Return on Assets Profit/Loss Foreign Ownership Part of Corporate Group

Industry Fixed Effects H1 +

H2 +

(14)

13

industries are 41-93, excluding codes 64-66 and 84. The excluded codes represent banking and insurance activities as well as public administration and defence. These excluded sectors do not use output per worker as a productivity measure but return on assets. Each company from the dataset has been observed during a three-year period falling between 2014 and 2019, depending on the last available year of the company’s data.

Measurements

The research has two independent variables, one dependent variable, and several control variables. The first independent variable is innovation input and it is represented by R&D expense. Hall (2011) identifies it as a suitable measure of innovative activity. It represents the decision to innovate by the company as it invests in innovation (Hall, 2011).

The second independent variable is innovation output, represented by the number of patents. Patents are measures of invention success and can be considered as partial measure of innovation output (Hall, 2011; Crespi and Pianta, 2008; Pakes and Griliches, 1984). The advantage of this measure is the availability of the data for longer periods of time. However, it has to be pointed out that in some fields it is more difficult to patent, and some prefer not to patent, as some companies may find it costly to patent (Mohnen, 2019). Patents represent innovation output in an entirety, therefore there is no distinction between innovation types in this research. All patents considered in this research have been granted. The count of patents per year has been determined by the application year instead of the publication year, because it better reflects the innovation input to innovation output in the given year.

(15)

14

(Ahmad et al., 2018). Therefore, the dependent variable, labour productivity, is represented by using labour productivity which can be defined as the turnover of the firm divided by the number of employees.

The first control variable, firm size, is defined as the number of employees. The relationship of firm size and productivity is significant, and its strength depends on the company size and type of innovation (Álvarez et al., 2013; Baumann and Kritikos, 2016; Hall and Sena, 2017). It is important to control for firm size because of the nature of this research.

The second control variable is firm age. Baumann and Kritikos (2016) found that firm age increases labour productivity, particularly among micro firms. It is important to control for age as there is a big variation in this variable across the sample. Since ORBIS provides the year of incorporation as the only information regarding the firm age, this control variable is calculated as the year of observationminus the year of incorporation.

The following two control variables represent profitability. Both return on assets (ROA) and profit have a significant relationship with productivity (Syverson, 2006; Teplykh, 2014). However, these relationships can work both ways in connection with innovation. First, productivity influences profitability, which can further influence productivity if a company decides to invest its profit into innovation again. Because of the nature of their relationship, it is important to control for them. Thus, the third control variable is represented by ROA and the fourth control variable, profit, is represented by profit/loss before taxes.

(16)

15

entities to be foreign-owned at 10 per cent. The results may be influenced and therefore it is considered to be a limitation to the research.

The sixth control variable is corporate group. It is important to control for corporate group as the literature indicates that companies that engage in innovation and are part of a corporate group have higher productivity than those that are not (Bond and Guceri, 2016; Castellacci, 2009). Corporate group is a dummy variable, where 1 represents the company being a part of a corporate group.

The last, seventh, control variable is industry fixed effects. They are included to capture unobserved differences across various sectors.

Method of Analysis

The first step of the analysis has been preparation of the dataset, starting with the calculation of variables, including the dependent variable, labour productivity, and the control variable, firm age. This has been followed by transformation of non-numerical variables, foreign ownership and corporate group, into dummy variables. Additionally, firm size variable has been adjusted for inconsistencies. Since the data are observed over a three-year period, the dataset has been reshaped in order to be able to use panel regression.

Furthermore, it has been necessary to calculate descriptive statistics to check the distribution of the data. However, due to the non-normal distribution of the data, the following step of the analysis has been logarithmic transformation of continuous variables. This way, the distribution of the data is normalised, and the results can be represented as elasticities.

Afterwards, the correlation analysis has been performed, followed by calculation of Cronbach’s alpha using reliability analysis. In order to conduct the regression analysis, the Hausman test has been performed to determine whether a fixed effect or random effects estimator should be used.

(17)

16 Descriptive Statistics

The ORBIS database, provided by Bureau van Dijk, consists of 2010 company cases. Overall, there is 1299 SMEs and 711 large businesses in the sample. The sample has 522 companies from service industries, where 67 per cent are SMEs. The manufacturing industries are represented by 1488 firms, where SMEs make up for 64 per cent. All descriptive statistics and frequencies are summarised in Table 1 and Table 2, respectively. The selection of the companies in the sample comes from various sectors and sizes of the companies. This diversity causes high values in descriptive statistics, especially because of the large companies subsample that includes big successful enterprises. Operating revenue is the first variable to be discussed. The eye-catching minimum of 0 value can be explained, upon further inspection, by five pharmaceutical companies with no operating revenues. Noticeably, this has an impact on the labour productivity calculation. These five companies are also classified as SMEs. The impact of these five companies on the whole sample and SMEs subsample is believed to be minimal. Therefore, the decision has been made to keep the companies in the sample for sake of diversity.

The companies in the sample have, on average, 1931 employees after controlling for inconsistencies. The smallest company has only one employee, and the biggest one has almost 600 thousand employees. The average number of employees seems to be influenced again because of the large businesses in the sample. The squared term is included in the regression when variable serves as control, to regulate the inverted U-shaped relationship. On average, it seems that the companies in the UK have high productivity. However, it has to be remembered once more that the sample consists of differently sized companies from various sectors where high productivity values differ per sector and company size. There, minimum productivity values of 0 are the reflection of operating revenue used for calculation of the labour productivity ratio. The high productivity values mean that dependent variable should be well represented in the regression.

(18)

17

transformation, all negative values have been excluded and therefore do not influence the results of regression.

Patents representing the second independent variable are also observed within all 2010 company cases. The average number of patents is 5.62, but the average number also includes the patent observation with the value of 0. These observations represent the years when companies might not have been granted any patents. However, it is necessary to mention that all observed companies had been granted patents within at least one year of the three-year observation period.

The average age of the companies in the sample is 35.46, which means that the sample includes mostly older companies. According to Baumann and Kritikos (2016), SMEs are often older than 35 years. They also say that younger firms put more effort into innovative activities than older firms. However, older SMEs are significantly much more successful in innovation. This could impact the results of regression. In order to prevent an inverted U-shaped curve in regression, square term of firm age has to be used as a control.

The sample includes companies that have a very low ROA, represented by the minimum of -99.43 per cent and very high ROA of 97.26 per cent. However, ROA is on average 3.21 per cent, which indicates that more companies in the sample are profitable. Because of the relationship of ROA and productivity, the results are expected to have positive significance. Profit/Loss before taxes shows high average profit among companies in the sample. The minimum value shows that there are companies that are very unprofitable. Upon further inspection, unprofitable companies are more represented in the SMEs subsample, which can potentially cause negative significance in the results, especially when testing the SMEs subsample.

(19)

18

Variables N Mean Std. Dev. Min Max

Operating Revenue 1991 4.44e+08 5.27e+09 0 2.36e+11

No. of Employees 1965 1931 20496.25 1 596452

Labour Productivity 1948 334554.2 2908872 0 1.99e+08

R&D Expense 2010 1.08e+07 1.22e+08 -5.89e+07 4.67e+09

Patents 2010 5.62 24.25 0 971

Age 1989 35.46 26.78 0 173

ROA 1971 3.21 20.93 -99.43 97.26

P/L before Taxes 2010 3.72e+07 5.30e+08 -2.39e+09 2.96e+10

Table 1: Descriptive Statistics

Variables N SMEs LBs

Company Size 2010 1299 711

N Service Manufacturing

Industry 2010 522 1488

SMEs per Industry 1299 348 951

LBs per Industry 711 174 537 N Yes No Foreign Ownership 1893 894 999 Corporate Group 2010 1937 73 Table 2: Frequencies Correlation

After performing the correlations test the results show that many variables are significantly correlated. Most of the significant relationships are weak and should not influence the results of the regressions. However, the variable profit and loss before taxes correlates on moderate to strong level with other variables. It correlates moderately with labour productivity and return on assets and correlates strongly with number of employees. Because of the strength of these correlations, profit and loss before taxes has been excluded from the regressions as a control variable. All correlations can be found in Table 5 in Appendix A.

Reliability and Validity

(20)

19

study are relatively poorly consistent with Cronbach’s alpha of 0.5914. This means that subsets of tested items of this research will likely not produce similar results.

Results

Hausman Test and Standard Error

Since this research uses panel data over a period of three years, the Hausman test has been performed before carrying out the regressions. The results show that regression should be using fixed effects estimator for testing of all hypotheses. However, because fixed effects estimator is used, dummy control variables, foreign ownership and corporate group, could not have been included into regressions, as fixed effect estimator omits constant variables. All standard errors have been clustered on industry level, to control for heteroskedasticity in the sample and subsamples.

Innovation Input and Labour Productivity

The first researched relationship is between innovation input and labour productivity, where innovation input is represented by R&D expense. The results of regression, using fixed effects estimator, show that this relationship is indeed positive and significant (p=.020) for the sample with the coefficient of 0.0153. If R&D expense would be increased by 10 per cent, the labour productivity would increase by only 0.15 per cent. R-squared indicates that 12.8 per cent of variation in productivity can be explained. Hypothesis 1 is therefore supported.

Firm size is also significantly related to the productivity; however, the relationship is negative. For the companies that invest into innovation, an increase in the number of employees by 1 per cent would decrease productivity by 0.68 per cent in this sample. Firm age is strongly significant, and also shows negative relationship with productivity. This relationship signalises that productivity decreases as the firm gets older, but non-linearly, meaning that old companies again become more productive. ROA is positively and significantly related to the productivity. Productivity increases by 0.45 per cent if ROA increases by 10 per cent, meaning that the profitability of the company increases the productivity.

(21)

20

increase in R&D spending. R-squared shows that 16.1 per cent of the variations in the dependent variable is explained. Hypothesis 1a is also supported.

The control variable firm size is significant and negative, but on a non-linear basis, suggesting that productivity of small SMEs and large SMEs is higher than productivity of companies in the middle of the subsample. Control variable age is also non-linearly negatively and significantly influencing the relationship between innovation input and productivity. This means that both variables are decreasing the elasticity of this relationship. On the other hand, the control variable ROA has strong significant effect on this relationship. Productivity increases by 0.61 per cent if ROA increases by 10 per cent. Compared to the whole sample, innovation input and productivity relationship in SMEs is more sensitive to changes in firm size and ROA.

In contrast, the relationship between innovation input and productivity is not significant (p=.134) for large businesses, resulting in Hypothesis 1b being rejected. The summary of regression results for Hypothesis 1, 1a, and 1b can be found in Table 3.

Labour Productivity (log)

Variables All Companies SMEs LBs

R&D Expense (log) 0.0153** 0.0122* 0.0179

(0.00635) (0.00636) (0.0118)

No. of Employees (log) -0.687* -1.168** -0.533

(0.354) (0.477) (0.953)

No. of Employees (log)2 0.0284 0.0839* 0.0129

(0.0330) (0.0499) (0.0690) Age (log) -0.781*** -0.556** -1.443*** (0.199) (0.236) (0.418) Age (log)2 _0.304*** _0.257*** _0.431*** (0.0377) (0.0489) (0.0636) ROA (log) 0.0472*** 0.0643*** 0.0306*** (0.00907) (0.0110) (0.00851) Constant 13.62*** 14.25*** 14.26*** (0.959) (1.176) (3.077) Observations 3,805 2,365 1,440 Number of BvD ID number 1,608 1,047 610 R-squared 0.128 0.161 0.111

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

(22)

21 Innovation Output and Labour Productivity

The second researched relationship is between innovation output and labour productivity, where innovation output is represented by number of patents. Firstly, two-stage least-square (2SLS) regression has been performed using R&D expanse as an instrumental variable to patents. However, because of the significance between innovation input represented by R&D expense and labour productivity, this instrument variable is not suitable. Therefore, OLS has been performed instead.

The results of regression, using fixed effects estimator, are insignificant (p=.498) for all companies in the sample. Thus, Hypothesis 2 is rejected. Upon further inspection of both subsamples, the relationship of innovation output and labour productivity is insignificant for SMEs (p.=0.704) as well as for large businesses (p.=0.520). As a result, Hypothesis 2a and Hypothesis 2b are rejected. The summary of regression results for Hypothesis 2, 2a, and 2b can be found in Table 4.

Labour Productivity (log)

Variables All Companies SMEs LBs

Patents (log) -0.00257 -0.00204 -0.00363

(0.00376) (0.00534) (0.00561)

No. of Employees (log) -0.730*** -1.357*** -0.761

(0.214) (0.247) (0.755)

No. of Employees (log)2 0.0164 0.0995*** 0.0116

(0.0201) (0.0332) (0.0493) Age (log) -0.389 -0.122 -1.029** (0.372) (0.696) (0.494) Age (log)2 0.231*** 0.172 0.332*** (0.0719) (0.129) (0.0793) ROA (log) 0.0484*** 0.0643*** 0.0296*** (0.0106) (0.0147) (0.0103) Constant 13.98*** 14.48*** 16.01*** (0.752) (1.120) (3.155) Observations 2,808 1,789 1,019 Number of BvD ID number 1,557 1,007 585 R-squared 0.181 0.172 0.216

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

(23)

22

Discussion and Conclusion

The results of the previously conducted studies vary per country, industry, company size and even innovation type. Therefore, the results of this research are mixed, but not surprising. To some extent, they even confirm the findings from the literature review.

The first hypothesis of this research studied the relationship between innovation input and productivity. The most common outcome in the existing literature is that innovative expenditure increases productivity, whether it is with stronger or weaker significance. These findings compare to the findings of this research, as R&D expense increases labour productivity in the whole sample. Firm size is negatively significant with labour productivity in this relationship. This result can be driven by the negative non-linear significance of the LBs subsample, which is discussed later. However, it has to be acknowledged that another cause for this difference can be the way in which companies carry out the R&D activities. Duch-Brown et al. (2018) suggest that SMEs may allocate more resources to R&D, while large businesses may carry out more R&D activities at once.

There are differences in the outcomes between literature and this research comes when SMEs and large businesses are studied separately. The relationship of innovation input and productivity in SMEs is significant in this research. This is in line with the studies of Hall et al. (2009), Griffith et al. (2006), and Aboal and Garda (2015). These studies argue that the relationship between innovation input and productivity in SMEs is significant. Firm size is non-linearly and negatively related to this relationship, as study of Duch-Brown et al. (2018) suggests. However, it is necessary to point out that this can be caused by SMEs engaging in informal R&D practices, as suggested by Baumann and Kritikos (2016). Informal R&D is simply not reported and therefore could not have been taken into account when performing the analysis. The negative non-linear significance of age in this relationship is in line with the results proposed by Baumann and Kritikos (2016). They propose that younger SMEs and SMEs older than 35 years old are more productive than middle aged SMEs. Furthermore, these findings also differ significantly from the proposed theory by Bond and Guceri (2016) that firm size and firm age do not play an important factor in innovation input and productivity relationship is therefore rejected.

(24)

23

expense. There are two possible explanations of these differences. First, as Aboal and Garda (2015) suggest, innovation strategies executed by SMEs and large businesses differ and therefore they bring different results. Second, Hall and Sena (2017) provide evidence that R&D expense is only a fraction of total innovation spending. Even more when the sample consists of various industries. To summarize, the relationship of innovation input and productivity is significant for all companies in the sample and SMEs as a separate subsample, but it is not significant for large businesses subsample. In addition, it has to be considered that results of the whole sample can be driven by the significance of the result of SMEs subsample.

The second hypothesis of this research focuses on the relationship between innovation output and productivity. The results of this research show that the relationship is insignificant for all companies, even when studied separately. This is not in line with findings of existing literature, even though they vary per study and innovation type. One of the explanations can be the measure selection of innovation output, patents. Some studies consider patents to be only partial measure of the innovation output (Hall, 2011; Crespi and Pianta, 2008; Pakes and Griliches, 1984). Being only a partial measure of the innovation output can mean that the measure is insufficient and other types of protections of intellectual property could have been used in addition to the patents. Mohnen (2019) also contributes by saying that the frequency of patent use depends on the sector. Moreover, Hall and Sena (2017) confirm in their study that protection of intellectual property increases productivity. The intellectual property protection is especially important for SMEs which are accessing external input (Hall and Sena, 2017). Surprisingly, Balasubramanian and Sivadasan (2011) found weaker evidence of significant relationship between patents and productivity, which was not shown in the results of this research. Besides the intellectual property protections, another measure of innovation output could have been innovation type. Despite the differences in the outcomes, the innovation type seems to be a more suitable measurement of innovation output as shown in previous studies.

(25)

24 Implications of the Research

Regardless of the insignificance of the results, there are some theoretical and managerial implications to consider. From a theoretical point of view, this research can be used as an addition to already existing literature on the topic of innovation and productivity. From a managerial point of view, it can be advisable to invest in R&D in case the company wants to increase its productivity. However, depending on the size of the company, it is important to study the differences between forms of innovation inputs, such as formal and informal R&D, and their influence on productivity in the firm. Undeniably, the choice of innovation input form has to be in line with the strategy of the company. Only after that can the correct decision to invest in innovation be made.

Limitations and Further Research

Some limitations of this research are required to be noted. The first limitation is the standardised reporting done by SMEs. SMEs often choose not to report as they are not required to report their financial information, including R&D expense. This could have skewed the actual results of the research.

The second limitation of this research is the measure of innovation output. Many studies use the innovation type as the measure, however, the ORBIS database does not offer such an option. Because this research used patents as the form of the measure, all innovation types could have been combined and therefore further influence the results of the research.

The third limitation is inappropriate instrumental variable, R&D expense, for 2SLS regression analysis. Because of this unsuitable instrumental variable, 2SLS regression analysis could not have been performed. This analysis could have provided better results to the study. Due to the fact that monetary values have not been deflated at the beginning of the analysis process, it can pose as the last limitation to the results to this research.

(26)

25

R&D practices in companies with different innovation types. Otherwise, the future research could use innovation types as an innovation output variable indicator even without using CIS. This could be done by using data collected by the future researcher or patent data collected by ORBIS and further divided into innovation types.

(27)

26

References

Aboal, D., & Garda, P. (2015). Technological and non-technological innovation and productivity in service vis-à-vis manufacturing sectors. Economics of Innovation and New Technology, 25(5), 1-20.

Aboal, D., & Tacsir, E. (2018). Innovation and productivity in service and manufacturing: The role of ICT. Industrial and Corporate Change, 27(2), 221-241.

Ahmad, S., Oliver, S., & Peters, C. (2018). Using firm-level data to compare productivities across countries and sectors: Possibilities and challenges. Economics Working Papers Series No. 2018-07-A.

Álvarez, R., Zahler, A., & Bravo-Ortega, C. (2013). Innovation and productivity in service: Evidence from Chile. Emerging Markets Finance and Trade, 51(3), 593-611.

Badiani, R. (2019). Outlook on UK-EU Brexit negotiations and possible economic risks. International Economics and Economic Policy, 16, 5-16.

Bahar, D. (2016). Using firm-level data to study growth and dispersion in total factor

productivity. The Brookings Institution Harvard Center for International Development. Balasubramanian, N., & Sivadasan, J. (2011). What Happens When Firms Patent? New

Evidence from U.S. Economic Census Data. Review of Economics and Statistics, 93(1), 126–46.

Baum, C. F., Lööf, H., Nabavi, P., & Andreas, S. (2017). A new approach to estimation of the R&D–innovation–productivity relationship. Economics of Innovation and New

Technology, 26(1-2), 121–133.

Baumann, J., & Kritikos, A. S. (2016) : The Link between R&D, Innovation and Productivity: Are Micro Firms Different? IZA Discussion Papers, No. 9734. Bond, S. R., & Guceri, I. (2016). R&D and productivity: Evidence from large UK

establishments with substantial R&D activities. Economics of Innovation and New Technology, 26(1-2), 108-120.

(28)

27

Castellacci, F. (2011). How does competition affect the relationship between innovation and productivity? Estimation of a CDM model for Norway. Economics of Innovation and New Technology, 20(7), 637–658.

Chudkovsky, D., Lópey, A., & Pupato, G. (2006). Innovation and productivity in developing countries: A study of Argentine manufacturing firms’behaviour (1992-2001). Research Policy, 35, 266-288.

Crépon, B., Duguet, E., & Mairesse, J. (1998). Research, innovation, and productivity: An econometric analysis at the firm level. National Bureau of Economic Research Working Paper No. w6696.

Crespi, F., & Pianta, M. (2008). Diversity in innovation and productivity in Europe. Journal of Evolutionary Economics, 18(3-4), 529-545.

Criscuolo, C. (2009). Innovation and Productivity: Estimating the Core Model Across 18 Countries. Innovation in Firms: A Microeconomic Perspective, 111–138. Paris: OECD Publishing.

Crowley, F., & McCann, P. (2015). Innovation and productivity in Irish firms. Spatial Economic Analysis, 10(2), 181-204.

Duch-Brown, N. de Panizza, A, & Rohnan, I. K. (2018). Innovation and productivity in a science-and-technology intensive sector: Information industries in Spain. Science and Public Policy, 45(2), 175-190.

Erken, H., Donselaar, P., & Thurik, R. (2008). Total factor productivity and the role of entrepreneurship. Jena Economic Research Papers No. 2008,019.

Gal, P. N. (2013). Measuring total factor productivity at the firm level using OECD ORBIS. OECD Economics Department Working Papers No. 1049.

Goedhuys, M. (2007). The impact of innovation activities on productivity and firm growth: evidence from Brazil. UNU-MERIT Working Paper Series, 02.

Gov.uk. (2020). Retrieved from https://www.gov.uk.

(29)

28

Hall, B. H., Lotti, F., & Mairesse, J. (2009). Innovation and productivity in SMEs: Empirical evidence from Italy. Small Business Economics, 33, 13-33.

Hall, B. H. (2011). Innovation and productivity. National Bureau of Economic Research Working Paper No. w17178.

Hall, B. H., Lotti, F., & Mairesse, J. (2013). Evidence on the impact of R&D and ICT investments on innovation and productivity in Italian firms. Economics of Innovation and New Technology, 22(3), 300-328.

Hall, B. H., & Sena, V. (2017). Appropriability mechanisms, innovation, and productivity: Evidence from the UK. Economics of Innovation and New Technology, 26(1-2), 42-62. Huergo, E., Jaumandreu, J. (2004). Firms' age, process innovation and productivity growth.

International Journal of Industrial Organization, 22(4), 541–559.

Kalemli-Özcan, Ş., Sorensen, B., Villegas-Sanchez, C., Volosovych, V., & Yesiltas, S. (2015). How to Construct Nationally Representative Firm Level Data from the Orbis Global Database: New Facts and Aggregate Implications. National Bureau of Economic Research Working Paper No. w21558.

Lopes, L. F., & Godinho, M. M. (2005). Services Innovation and Economic Performance: Analysis at the Firm Level. DRUID Working Paper No. 05–08.

Mairesse, J., and Robin, S. (2010). Innovation and Productivity: A Firm-Level Analysis for French Manufacturing and Services Using CIS3 and CIS4 data (1998-2000 and 2002-2004). Paris: CREST-ENSAE.

Mohnen, P. (2019). R&D, innovation and productivity. In The Palgrave Handbook of Economic Performance Analysis, 97-122.

Mohnen, P., & Hall, B. H. (2013). Innovation and productivity: An update. Eurasian Business Review, 3(1), 47–65.

OECD (2005). Oslo Manual: Guidelines for collecting and interpreting innovation data (3rd ed.). Paris: OECD. https://doi.org/10.1787/9789264013100-en.

(30)

29

OECD (2017). Enhancing the contributions of SMEs in a global and digitalised economy. Paris: OECD. https://www.oecd.org/industry/C-MIN-2017-8-EN.pdf.

Pakes, A., & Griliches, Z. (1984). Patents and R&D at the Firm Level: A First Look. R&D, Patents, and Productivity, 55–71. Chicago: University of Chicago Press.

Santarelli, E., & Lotti, F. (2008). Innovative output, productivity and profitability: A test comparing USPTO and EPO data. Industry and Innovation, 15(4), 393-409.

Segarra-Blasco, A. (2010). Innovation and Productivity in Manufacturing and Service Firms in Catalonia: A Regional Approach. Economics of Innovation and New Technology, 19(3), 233–258.

Schilling, M. A. (2017). Strategic management of technological innovation. New York: McGraw-Hill Education.

Stelios, K., & Aristotelis, S. (2008). Knowledge-Intensive Services, Innovation and Economic Performance: A Firm-Level Analysis. Athens: Centre for Planning and Economic Research (Greece).

Syverson, C. (2011). What determines productivity? Journal of Economic Literature, 49(2), 326-365.

Taber, K. S. (2018). The use of Cronbach’s alpha when developing and reporting research instruments in science education. Research in Science Education, 48, 1273-1296. Tevdovski, D., Tosevska-Trpcevska, K., Makrevska Disoska, E. (2017). What is the role of

innovation in productivity growth in Central and Eastern European countries? Economics of Transition, 25(3), 527–551.

Teplykh, G. V. (2014). Innovations and productivity: The shift during the 2008 crisis. Industry and Innovation, 25(1), 53-83.

(31)

30

Appendix

Appendix A – Correlation Analysis Table

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

(1) Labour Productivity (log) 1

(2) Patents (log) 0.02* 1

(3) R&D Expense (log) 0.1*** 0 1

(4) No. of Employees (log) 0.11*** -0.01 0.08*** 1

(5) Age (log) 0.16*** -0.07*** -0.05*** 0.36*** 1

(6) ROA (log) 0.08*** -0.02 0.05*** -0.14*** -0.03* 1

(7) P/L before Taxes (log) 0.34*** 0.01 0.22*** 0.71*** 0.18*** 0.32*** 1

(8) Corporate Group 0.17*** -0.02 0.01 0.13*** 0.15*** -0.01 0.1*** 1

(9) Foreign Ownership 0.16*** 0.01 0.14*** -0.05*** 0.01 0 0.01 -0.09*** 1

(10) NACE2 (Industry) -0.15*** 0.03** 0.03* -0.16*** -0.37*** 0 -0.02 -0.07*** -0.04*** 1

*** p<0.01, ** p<0.05, * p<0.1

Innovation and Productivity in SMEs and Large Businesses in the United Kingdom