Abstract
The current dissertation is focused on exploring the firm-specific determinants of profitability of the industrial companies of the United Kingdom-based on the information from the OSIRIS database for the years 2015, 2016, and 2017. Three models were fitted to estimate the determinants of profitability, each of them being further evaluated with the help of the simple OLS, random effects, and fixed effects regression analysis frameworks.
The three models varied by the profitability proxies chosen as the dependent variables, which were the return on assets (ROA), the return of equity (ROE), and the profit margin. The selected independent variables of each of the models were the liquidity ratio, the gearing ratio, the asset turnover, the inventory turnover, the collection period, the credit period, the interest coverage, the market capitalization, and the number of employees of the considered companies.
Of all the indicators some were significant in one or two models, and only the indicators of market capitalization and gearing ratio were significant across all the three selected models. The gearing ratio was found to be negatively correlated with profitability, in which market capitalization showed a positive significant association. The conclusions of the study can be used in practice for making more optimal decisions in company management and financial investment activities
Introduction
Profitability as a business indicator has often been viewed by enterprises globally as one of the most important and desirable goals aimed to guarantee successful strategic development. As noted by Mendoza (2015), profitability can be used not as an efficient measure of business performance, but also as one of the key drivers of long-term survival of companies. Similarly, according to (Zimmerman and Bell, 2015), it can have a solid association with the levels of sustainability.
Looking at the specific case of the UK, exploring the most recent trends in its profitability can be interesting for several reasons. As noted by Newth et al. (2018), the recent Brexit vote of 2016 has caused the UK companies to reconsider the structures of their business models, which might have impacted the structural frameworks of profitability. Secondly, the increasing pressure of competition is causing the companies globally to invest more actively into advanced technological solutions of their businesses, which can also have an impact on the specific factors which used to be or are becoming the most solid predictors of companies’ profitability in the world and the UK specifically.
Thirdly, the volatile political environment of the UK of a recent couple of years introduces additional levels of risks into the portfolio of the local companies, and therefore identifying the actual drivers of profitability which work in the current environment can help the UK firms to design more optimal long-term development strategies.
In the current dissertation, the key aim is to explore the actual drivers of profitability of the companies headquartered in the United Kingdom-based on the set of available data for the years 2015, 2016, and 2017. The key source of the information for the analysis is the OSIRIS database, available through the University Library access. The literature review chapter will be focused on the three key topics. First of all, it will briefly outline the state of research on the profitability of UK companies within the recent decade.
This will help to establish some of the expectations of the trends and also compare the future analysis output with the previous findings. Secondly, the literature review will outline the existing evidence on examining the determinants of firms’ profitability globally. This will facilitate the decision on the optimal choice of proxies for the measures of profitability, as well as the possible, preferable sets of the independent variables to be used as predictors of profitability in the subsequent chapters. Lastly, the literature review will also explore the possible best options for handling panel data in the context of profitability concerning the choice of the most suitable models.
The methodology chapter of the dissertation will give a better idea of the chosen research strategy, outlining the research philosophy and the specific choice of the dataset variables and equation models. The findings chapter will present the summary statistics and the outputs of the OLS, fixed effects, and random effects regressions that were selected to analyze this dissertation. The discussion chapter will provide the interpretation of the obtained results and compare them to the expected results and the previously existing evidence. Lastly, the concluding chapter will summarize the conducted analysis, as well as will outline its limitations and the possible beneficial directions for future research.
Literature Review
As was already noted above, the current literature review will be comprised of the three major parts. The first one will be focused specifically on identifying the existing research gap – the lack of literature on the determinants of profitability of the companies in the United Kingdom published in the Brexit and post-Brexit period. To fill in this gap properly, the two subsection sections will be useful. One of them will be focused on the most recent efforts of conceptualizing and predicting corporate profitability in the different regions of the world.
The third and last section of the literature review will explore the nuances of dealing with balanced panel data in the context of exploring various dimensions of financial ratios and specifically the indicators of profitability. The synergy of all three sections will help to define the optimal research strategy for the subsequent sections, which will be based on the most solid existing empirical and theoretical evidence and the identified research gap.
Exploring the Existing Recent Literature on the Profitability of UK Companies
Looking at the most recent literature on the determinants of profitability of UK companies published after 2014, it can be seen that it is rather scarce and fragmentary. Besides, more often than not it is focused on a single profitability dimension, relationship, or context, which impedes generalizations. Due to this, it can be rather difficult to extrapolate the findings of the authors onto a wider context of environments and associations and thus see a wider context of profitability determinants.
For instance, a recent study by Quader, Kamal, and Hassan (2016) focused on exploring the relationship between the quality of environmental performance and the levels of profitability in the United Kingdom. The paper was specifically based on the case study evidence collected through qualitative research methods for small and medium-sized enterprises (SMEs). The authors used in-depth interviews with industry experts to figure out the prevailing opinions on the impacts of environmentally focused decisions on the strategic success of the business.
As the findings suggested, these types of sustainable development efforts can often be used by SMEs for marketing purposes solely as a marketing activity and in attempts to increase the levels of clients’ loyalty. However, as an outcome, these efforts can also have positive environmental impacts on the local communities. While the commitment to sustainability can indeed be a workable predictor of profitability, it is hard to quantify its degree and impact, and therefore detailed quantitative information for the subject is not readily available. This is one of the most probable reasons why Quader, Kamal, and Hassan (2016) chose a descriptive qualitative case study research design to explore the relationship of interest.
Another recent study on exploring profitability in the UK (Jones et al., 2015) was focused on determining the future profit patterns of companies in the fishing industry. The key rationale of the authors was based on the recent global climate changes that altered the placement and distribution of the marine species and thus imposed additional difficulties on the processes of profit-centered fishery management.
The authors suggest the method of scenario development and testing to predict future profitability for UK fisheries. For the mentioned scenarios, the methods of species distribution modeling were linked with the instruments of cost-benefit analysis. As a result, the authors calculated the profitability of different scenarios for the 45-year horizon with the help of the NPV indicator. While the method employed by the authors can be the best choice for the fishery industry, it can be rather hard to extrapolate it onto the wider context of companies from other industries and especially for the case of a large dataset, as the scenarios should then be developed for each company or industry individually.
One more recent paper that explored the patterns of profitability in the UK environment is the research by Pacini et al. (2018). In this article, the authors specifically focused on the influence of external factors on the profitability of enterprises in the United Kingdom. The selected dependent variable to measure profitability was the factor of asset productivity, calculated as the ratio of the company’s net value added to total assets.
Some of the studied independent predictors used by the authors in the regression model were the inflation rate, the exchange rate, the interest rate, the Gross Domestic Product (GDP), the ratio of interest payments on domestic debt to the number of net borrowings, and the rates of current foreign debt to the existing international reserves of the central bank. The sample collected by the authors included the performance data on 100 top companies of the United Kingdom observed over the period between 2000 and 2014. As a result, the study found that some of the factors having a positive influence on firm profitability are inflation and the GDP.
While this type of study is indeed valid to explore the macroeconomic factors of firm profitability in the UK, it gives little guidance on what could be the beneficial actions of the corporate management on the micro-level which could have the potential to stimulate profitability. Also, given that the dataset used by the authors was collected in the pre-Brexit period, it does not necessarily reflect the existing patterns of profitability which are constructed as a response to the changing political and economic integration conditions.
One more recent study related to profitability patterns in the UK was published by Saeed (2014). The author focused exclusively on the factors determining the profitability of organizations in the banking sector and based the analysis on the performance information obtained from 73 commercial banks operating in the United Kingdom, for the period between 2006 and 2012. The analytical methods employed by the author included the correlational and regression analysis.
The two key selected measures of bank profitability were the return on assets (ROA) and return on equity (ROE). The author found that factors such as capital ratio, bank size, and liquidity have a positive impact on bank profitability measured through ROA and ROE, while GDP and inflation had a negative impact.
The analysis performed in the article is reasonably comprehensive, but it was only focused on the banking institutions rather than industrial companies, and besides as of today the used dataset is rather outdated. Also, the middle of the analyzed period contained the years 2008 and 2009 in which the global financial crisis was happening and which could have altered significantly the patterns and frameworks of profitability in the UK economy.
Summing up all of the above it is possible to conclude that there does not seem to exist a recent study on the industrial companies of the United Kingdom aiming to explore the internal micro-level determinants of firm profitability. Therefore, this gap in the empirical research can be filled by developing a workable analytical model of profitability determinants basing on the most recent data available from the OSIRIS database. The subsequent sections of this literature review are aimed at identifying the most optimal model for further analysis taking into account the aims of the study and the available set of data.
Exploring the Empirical Literature on the Determinants of Profitability
One of the most recent comprehensive studies of determinants of corporate profitability was published by Nanda and Panda (2018). The authors were focused on the predictors of profitability of a sample of Indian manufacturing firms observed over the period between 2000 and 2015. The study formulated two key sets of factors predicting firms’ profitability – firm-specific and macroeconomic.
The two proxies which the authors used for profitability were the net profit margins and the values of returns on assets of the considered firms. Some of the independent variables used for predicting profitability were the levels of liquidity, firm size, leverage, exchange rate, and industrial growth. The hypothesized directions of the relationships between the observed variables were also described. For instance, it is claimed that the positive sign of relationship is expected between the variables of firm size and profitability.
This is explained by the economies of scales and by the fact that in markets with a high concentration of industry competition larger firms are expected to make higher returns and compared to similar smaller enterprises. As for the relationship between leverage and profitability, the authors suggest the existence of an inverse relationship due to the impacts of the trade-off and pecking order theories. They also mention a recent paper by Al-Jafari and Samman (2015) which empirically showed the negative relationship between firm leverage and profitability. Then, Nanda and Panda (2018) suggest the duality in the relationship between the variables of profitability and liquidity.
In some cases, holding excess levels of liquidity can be associated with additional expenses and unearned profits, as the same funds could have been invested at some higher rates. This type of pattern can be expected to influence profitability negatively. On the other hand, profitability and liquidity can also be positively related, as insufficient liquidity can generate the needs for expensive short-term loans and lower than necessary levels of cash flow.
Besides, high levels of liquidity also serve as a positive signal to investors indicating that there are relatively lower short-term risks associated with investing in the stocks of a certain company. Consistently with the posed hypotheses, the authors of the paper found that the indicators of leverage were inversely related to profitability, while the liquidity indicator was positively associated with profitability.
One more paper exploring the determinants of profitability for the set of micro, small and medium enterprises in the Philippines (Mendoza, 2015) was based on examining over 200 cases of financial statements of 67 companies observed between 2011 and 2013 to determine a set of significant independent predictors for this country. Basing on the exploration of previous research, the author tested the variables of current ratio, quick ratio, collection days, inventory days, asset turnover, and the debt ratio for being significant predictors of enterprise profitability in the Philippines.
The authors tested three separate models for each of the three dependent variables selected to be the proxies of profitability: the return on assets (ROA), the return on equity (ROE), and the return on sales (ROS).
The authors found that ROS has a negative association with the logarithmic transformation of the asset turnover and the logarithmic transformation of the debt ratio. For the case of ROA, the log of asset turnover showed a positive association with profitability, while the log of debt ratio showed a negative significant association. For the case of ROE, only one of the variables had a significant coefficient, which was the positive association between the level of profitability and the logarithmic transformation of assets turnover.
Azhagaiah and Deepa (2012) focused on determining the drivers of the profitability of the companies operating in the Indian food industry. The paper mainly highlights the impact of companies’ size on the levels of profitability, classifying the selected companies into the three groups – small, medium, and large – depending on the scope of their sales turnover. The dependent variable of the analysis which served as proxy measuring profitability was the return on assets, while the set of the selected independent variables included the levels of capital intensity, proxies for liquidity and firm size, as well as the volatility of growth and the levels of growth in total assets.
The analysis showed that capital intensity was an especially significant positive predictor of profitability for the cohort of larger firms. This conclusion was specifically relevant for the case of the higher levels of investment in long-lived assets, as they predicted higher profitability for the large enterprises, but not for the small- and medium-sized. A paper by Mary et al. (2012) based their analysis of profitability determinants on researching the trends of the four beer brewery firms in Nigeria observed within the period between the years 2000 and 2011.
The authors used the indicator of the gross profit margin as a proxy for profitability and the variables of borrowing, depreciation, inventory levels, and operating expenses as the independent variables of the analysis. The authors used the simple OLS method to analyze the collected data, which might seem a suboptimal choice as compared to special models designed to handle panel datasets. The article concluded that the ratio of inventories to COGS and account receivables to sales are significantly associated with the dynamics of the gross profit margin.
One more related paper by Boadi et al. (2013) focused on exploring the profitability drivers of 16 insurance firms in Ghana over the period between the years 2005 and 2010. The authors used the variable of return on assets as the proxy for profitability, while liquidity, tangibility, leverage, size, and the percentage change in premiums were conceptualized as the set of the relevant independent variables.
The methodological approach used for the data analysis was the panel method and the simple OLS multiple regression. The indicator of tangibility was negatively related to the dynamics of the return on assets, while the indicators of leverage and liquidity were positively related to profitability. An article by Yazdanfar (2013) considered the influence of firm-related factors on productivity for the case of a developed country – Sweden.
The author collected an impressive dataset of 12,530 nonfinancial micro firms, observing them between the years 2006 and 2007. Similarly to the previous studies, the chosen proxy for profitability was the return on assets. The set of independent variables included the indicators of growth, firm size, firm age, lagged values of profitability, industry affiliation, and productivity. The methodology used by the authors is referred to as the Seemingly Unrelated Regression (SUR).
It helped to find the optimal combinations of variables that could serve as predictors of profitability. In the outcome, the authors found that the factors of firm age and industry affiliation are negatively influencing the observed profitability.
At the same time, lagged profitability values, growth, firm size, and productivity was found to be positively associated with the dynamics of the return on assets. Mohamed and Jones (2014) explored the frameworks of profitability determining indicators on the example of 60 information and communication companies in Egypt. The proxy of profitability was returned on assets, while the independent predictors were the firms’ cost drivers, asset drivers, and revenue drivers. Empirical evidence was found on the importance of each of the three factors in defining the dynamics of profitability.
The two recent papers by Dimitropoulos (2018) and Magoutas et al. (2018) are focused on the related topic – the determinants of profitability in the Greek hospitality and tourism industries, respectively, in the period after the crisis. Magoutas et al. (2018) observe an extensive sample of the Greek tourism industry companies over the period between 2005 and 2014. The authors are specifically interested in how the Greek financial crisis affected the relationships between the profitability and its identified determinants.
It is highlighted in the conclusions of the paper that in the post-crisis period the levels of capitalization of firms affect the profitability in the negative direction, which can be related to the negative events happening in the Greek economy. The higher profitability in the Greek tourism sector could have multiplier effects on the rest of the industries of the country due to its importance, but for that access to low-cost financing for the local enterprises would be necessary. To reach the above conclusion, the authors estimated the designed regression equation on three separate periods – total (2005 – 2014), pre-crisis (2005 – 2008), and post-crisis (2009 – 2014).
Similarly to the previous paper, Dimitropoulos (2018) used ROA as a predictor of profitability in the hospitality industry of Greece, against the variables of firm size, capital intensity, leverage, liquidity, sales revenues, and the effective tax rate on the left-hand side. The size of the company and the number of sales were found to be the positive predictors of ROA, while liquidity, leverage, and capital intensity were the negative set of predictors. To account for the panel nature of the data, the authors included a set of year dummy variables into the final regression equation.
Looking at the set of the described literature exploring the determinants of profitability in different settings around the world, it is possible to conclude that the most frequently used proxies for profitability are ROA, ROE, and the profit margins.
For this reason, they will also be used as the dependent variables measuring profitability in the current study. As for the decision on the set of the independent variables to be used in the further analysis, it was made based on the intersection of the two key factors – presence in the previous literature on the topic and in the OSIRIS database, which will be used as a source of the considered dataset in the subsequent chapters. For this reason, nine key independent variables were chosen as the profitability predictors in the current study, which are the liquidity, gearing, collection period, interest coverage, inventory turnover, assets turnover, credit period, market capitalization, and the number of employees.
Exploring the Literature on Handling Panel Data in the Context of Financial Ratios and Indicators of Profitability
In general, the notion of panel data means that the variables of the considered dataset are observed over several similar periods for several ids, countries, individuals, or other types of pre-determined groups. For the case of this dissertation, the observations are made for 284 different companies within the period of three years – 2015, 2016, and 2017. For this type of data, the specific influences associated with either years or companies can interfere with determining the desired effect coefficients. In this situation, there arises a need for specific econometric models to come to the conclusions of interest.
For instance, in an article by Nanda and Panda (2018) the authors used the data on 173 Indian firms observed over the period between 2000 and 2015. To handle the dataset, the generalized least square (GLS) approach with random effects design was selected. This helped to capture the fraction of variance associated with period-specific effects and obtain more representative results in the outcome.
Similarly, Petria et al. (2015) were exploring the determinants of profitability of banks basing on the evidence of the 27 banking systems of the EU. The authors used the return on average assets (ROAA) and return on average equity (ROAE) as the variables measuring profitability, and explored them for the period between 2004 and 2011. The authors chose to estimate two frameworks – fixed and random effects models to deal with the panel nature of the chosen dataset.
Following the estimation of the two models, a conclusion was reached that the fixed effects model is optimal in estimating the relationship of interest as compared to the random effects. This judgment was made based on the outcome of the Hausman test, as its null hypothesis was rejected according to the test statistics probability of 0.000. The final estimated models showed that the variables of inflation, credit risk, and efficiency are negatively associated with profitability, while the variable of growth has a positive impact on both ROAA and ROAE.
One more study using the panel data analysis techniques in the context of measuring profitability is the article by Abduh et al. (2017), focused on exploring the profitability of Islamic and conventional banks in Malaysia. The profitability predictor in the regression is the return on assets, and the independent variables are the indicators of liquidity, capital adequacy, and the real GDP. The authors estimate three separate regressions: the simple OLS and the fixed and random effects frameworks.
Under the pooled OLS estimation, the authors found that both real GDP (the external macroeconomic variable) and the liquidity ratios are the significant predictors of the return on assets in Malaysia between 2005 and 2009. Then, the authors estimated the random effects and fixed effects models and found that the fixed effects model should be preferred due to the inter-model test outcome. The random-effects models concluded that only liquidity ratios are significantly relevant to predicting profitability.
However, for the preferred fixed effects model there were two significant predictors of banks’ profitability: the type of banks and the real GDP (macroeconomic conditions). Onuonga (2014), Alalaya and Khattab (2015), Raashid et al. (2015), Hirsch (2018), Almaqtari et al. (2018), Charles et al. (2018), Dalci (2018), Montoro-Pons and Cuadrado-García (2018), Salike and Ao (2018) are some of the other recent papers using fixed and random effects estimation models for handing panel data in the context of exploring profitability.
Basing on the evidence presented in this section, it is possible to conclude that it may be optimal to choose three models for estimating the data in the subsequent analysis. These models are the simple OLS regression framework, the fixed effects regression, and the random effects regression.
The two latter models can help in accounting for the heterogeneous time-related and company-related effects and in this way better identify the coefficients associated with the predictors of profitability of the chosen UK industrial companies and their significance. The latter framework is relevant to the research aim of uncovering the determinants of profitability is the UK, the research question if those can be the firm-specific indicators, and the simple OLS / random effects / fixed-effects models.
Methodology
The Limitations of Secondary Research
Having observed comprehensively the pillar of the literature exploring the determinants of profitability in the UK for recent years, it can be concluded that no comprehensive study on this topic was conducted. The articles that exist about the subject address the issue of determinants of profitability of the UK companies fragmentarily – they are focused on a specific environment, specific industry, or a limited set of indicators.
This represents a revealed gap in the secondary literature and therefore highlights a need for a study of the most recent evidence on the profitability drivers in the United Kingdom. It will be filled with the help of the data analysis plan and the methodology outlined further in this chapter.
The Research Philosophy
The research philosophy pursued in this study is positivism. This means that the trustworthiness of the study is sourced from observing the factual knowledge which can be reflected through measurement (Ghamrawi et al., 2018).
In studies based on the philosophy of positivism, the subjectivism on the side of the researcher is minimized, as both data interpretation and data collection are exercised objectively. Besides, in such types of studies, the findings are usually quantitative, such as the regression coefficients and their associated p-values. The key source of positivism in research is the process of observation which subsequently results in the statistical analysis. In the current study, consistently with the logic of the positivist research philosophy, the personality of the researcher is independent of the study, and there are no human interests directly related to the study outcomes.
The Research Strategy
The research strategy will be based on collecting the set of panel data from the OSIRIS database and processing it with the help of the Stata statistical package software. The data on the collected dependent and independent variables will be analyzed with the help of the OLS, random effects, and fixed effects regression frameworks and further tested for the model fit with the help of the Hausman test. The obtained coefficients will be interpreted and compared to the previously existing evidence. Lastly, the conclusions will be made on the obtained results, including the identified limitations of the study, practical application of the findings, and the possible future directions of the research.
The key research question of the study is as below:
- Can any of the selected firm-level indicators be the significant determinants of profitability of the UK industrial companies, basing on the dataset of the years from 2015 to 2017?
The associated null and alternative hypotheses are as below:
- H0: None of the selected firm-level indicators can predict profitability measured through ROE, ROA, or profit margins, and their associated coefficients are either equal to zero or insignificant;
- H1: At least one of the selected firm-level indicators can predict profitability measured through ROE, ROA, or profit margins, and the associated coefficient(s) is (are) non-zero and significant.
The Process of Data Collection, Inputting, and Analysis
The data was collected with the help of the OSIRIS database, which contains the information on the financials of the thousands of companies globally. The database allows exporting the data in the MS Excel format, which can, in turn, be pasted into the data editor of the Stata statistical package. Therefore, the process of inputting the data into the final statistical package of choice is rather simple and straightforward. The preliminary analysis of the collected dataset was conducted in MS Excel, to understand the general nature of the considered data and how many missing values it contained.
The Population and Sampling Approach
The population of interest to the current study is the set of industrial companies of the United Kingdom that explicitly reported their financial information within the years between 2015 and 2017. The initial sample obtained with the help of OSIRIS included data for at least some of the variables for a total of 1875 companies. However, on the further stages of data analysis, the sample was reduced very significantly basing on the selected sampling approach.
The inclusion criteria for the companies of the final sample was that there exists a full set of observations for the company’s finances for each of the selected indicators for each of the included years (2015, 2016, and 2017). This reduced the initial number of observations to a more limited sample of only 284 companies. Therefore, the total number of observations in the final sample is equal to 852, which represents three entries (for 2015, 2016, and 2017) for each of the included companies.
Description of the Selected Variables
To conduct further analysis, the intersection of the examined literature, and the factor of availability in the OSIRIS database resulted in the final choice of 3 dependent variables and nine independent variables. The interpretation of the variables as it is reflected in the source database will be described further in this section.
The dependent variables chosen to be the proxies for profitability are the return on assets (ROA), return on equity (ROE), and the profit margin. The indicator of ROA is measured as a percentage based on the ratio of net income to total assets multiplied by the factor of 100. Similarly, ROE is the percentage reflection of the ratio of the net income divided by the number of the reported shareholders’ funds and multiplied by the factor of 100. The indicator of the profit margin is the percentage that reflects the ratio of the profit or loss before tax divided by the value of operating revenue (turnover) and multiplied by the factor of 100.
The dependent variable of liquidity ratio is represented by the current asset stocks divided by the value of the current liabilities. Interest coverage (in percent) is the operating profit or loss divided by the interest expense. The collection period in days is the indicator of liquidity calculated as the debtors divided by the operating revenue and multiplied by 360. A credit period in days is the ratio of creditors to operating revenue multiplied by 360.
Gearing (in percentage) is the leverage ratio calculated as the value of non-current liabilities with the added loans, divided by the number of shareholder funds and multiplied by the factor of 100. Assets turnover is the asset utilization ratio calculated as the operating revenue divided by the number of shareholders’ funds with the added amount of the non-current liabilities. Inventory turnover is one more asset utilization ratio which is calculated as the ratio of operating revenue to the number of stocks (inventories).
Two more variables to be used on the right-hand side of the equation are the market capitalization in US dollars and the number of employees working in the company. The two latter variables are very large in absolute values, and for this reason, they are included in the regression equation in the logarithmic form, to ensure more convenient interpretation.
Description of the Selected Model
The selected regression model will be estimated with the help of the three main methods: the simple OLS, and the fixed effects and random effects models for handing panel datasets. The specific equation to be used in each case will be as below:
profitability = β0 + β1 * liquidity + β2 * gearing + β3* collection_p + β4 * interest_c + β5* inventory_t + β6* assets_t + β7* credit_p+ β8* lmark_cap + β7* ln_employees + ε, where
- β0 is a constant term attributable to the selected equation;
- profitability is the dependent variable and a proxy of profitability (which is either ROE, ROA, or the profit margin);
- liquidity is the liquidity ratio reflected as a percentage;
- gearing is the gearing ratio reflected as a percentage;
- collection_p is the collection period reflected in days;
- interest_c is the interest coverage ratio;
- inventory_t is the inventory turnover ratio;
- assets_t is the inventory turnover ratio;
- credit_p is the credit period reflected in days;
- lmark_cap is the market capitalization of a company reflected as a logarithmic transformation;
- ln_employees is the number of employees a company reflected as a logarithmic transformation;
- ε – is the stochastic error term.
The Hausman test will help to determine which of the two latter models, fixed or random effects, is preferable for the considered case. If the H0 of the test that the difference in coefficients is not systematic is rejected, fixed effects should be preferred, otherwise – random effects.
Conclusion
Summing up the above, it can be said that the current study employs the positivist research philosophy and is based on the extensive use of quantitative analytical methods. The dataset of interest will be obtained with the help of the OSIRIS database and processed with the help of the Stata statistical software. The three parallel methods to estimate the data will be the simple OLS, the random effects, and the fixed effects models that will help to account for the time- and company-related heterogeneity. The choice of the best model will be determined with the help of the Hausman test. The next chapter of the dissertation will outline the findings obtained by utilizing the described methodology.
Findings
The current chapter of the dissertation will give a detailed outline of the data analysis process employed in answering the hypotheses of interest. The first section will report the summary statistics of the collected dataset, which will help to have a better understanding of the average values of the collected variables, as well as their ranges and patterns of variation. The second section will discuss specifically the regression outputs and the magnitudes and levels of significance of the coefficients obtained within the estimation process.
It will be comprised of three subsections, each of them testing the research hypothesis basing on a different dependent variable on the left-hand side. These three dependent variables will be the return on equity, return on assets, and the net profit margin. For each of the three hypotheses, three types of econometric models will be estimated: the simple Ordinary Least Squares (OLS) and the fixed and random effect models for panel data.
Following the estimation of two of the later models in the Stata statistical software package, it will be necessary to compare their outputs within the Hausman test framework, to check for the consistency of the two estimates. In case if the null hypothesis fails to be rejected, random effects models should be utilized as it is associated with higher levels of efficiency. Once the null hypothesis is rejected in favor of the alternative hypothesis, the test suggests using fixed effects. Having obtained the regression and test outputs from Stata, it will be possible to make the conclusions on the research questions of interest.
Summary Statistics
As was noted previously, the data of the considered dataset was obtained from the OSIRIS database which contained the information on the variables of interest for the years 2015, 2016, and 2017. Once the companies with a lot of “no answer” and “not specified” values were eliminated, only 284 firms with full information were found for the UK sample. Multiplying the number of 284 companies by the three years of collected data gives a total number of 852 observations within the considered dataset.
Table 1 below presents the summary statistics for the obtained variables. Given that the variables of the number of employees and the market capitalization of the company are very high in absolute numbers, their logarithmic transformations will be used in further analysis. This provides a researcher with an easier interpretation of the beta coefficients, as logs of variables are interpreted in terms of percentage rather than per unit changes.
However, to give a better understanding of the nature of the data, both logarithmic transformations and the original forms of the variables of the number of employees and market capitalization will be shown within the summary statistics representations.
As can be seen from Table 1, the mean value of the return on equity is the highest of the three selected indicators of profitability and is equal to 6.69%. Similarly, the return on equity also has the highest levels of variation, which is reflected in the standard deviation of 33.79%, which is more than five times higher than the mean value. The range of the ROE indicator is from -378.12% to 213.86%, which represents a very high level of dispersion.
Return on assets is the least dispersed variable of the three included proxies of profitability, with a standard deviation of 10.24% and the range from -81.38 %to 40.24%, and the mean of ROE is 3.01%. The profit margin variable has a mean of 4.62% and a standard deviation of 16.97%. The average for the liquidity ratio of the companies included in the sample is 1.10, with the range between 0.03 and 23.89 and a standard deviation of 1.12, which is roughly equal to the calculated mean. The average for the gearing ratio in percentage is 100.93%, with a standard deviation of 127.11% and the range between 0 and 932.92%.
The average collection period observed for the companies of the sample is equal to 47.13, with a wide range between 0 and 202 days. The interest coverage ratio has a mean of 15.17%, with a standard deviation of 52.72 and the range between -86.59% and 780.51%. The asset turnover ratio averages 1.80 with the range from 0.10 to 16.45 and the standard deviation of 2.01. The credit period measured in days has a mean of 38.67, ranging from 3 to 320 days with a standard deviation of 26.66.
The average number of employees of the selected companies is 18,352, with the range between 13 and 611,366 and the standard deviation of 60,176. This shows that the majority of the companies included in the dataset are large. Lastly, the average market capitalization in the US dollar thousands of the included companies is $5,686,473, with a standard deviation of $15,900,000 and a range between $1,389 and $140,000,000.
Exploring the Three Sets of Hypotheses
In this section, regression outputs will be presented and discussed. The three subsections will cover the set of relevant predictors for the profitability estimates of the return on assets, return on equity, and the net profit margin. The undertaken calculation will show which sets of predictors are more appropriate for these specific measures of profitability and if the significance and effect size of the selected predictors varies across the hypotheses sets.
Exploring the Determinants of the Return on Equity
As can be seen from Table 2 below, three different models were estimated to explore the significant predictors of UK industry companies’ profitability as measured by return on equity. The estimated OLS regression showed that the associations between gearing and credit period and ROE are negative and significant, and the associations between the interest coverage, inventory turnover, assets turnover, and the log of market capitalization are positive and significant. The value of R-squared is equal to 0.16, which means that close to 16% of the variation in the chosen dependent variable of ROE is explained by the selected set of independent variables.
However, given that the dataset of consideration is panel data, it is more relevant to interpret it with the specifically designed regression techniques. Therefore, the fixed and random effects regressions were also estimated, and their outputs are presented in Table 2 to the right of the OLS. Following the estimation of the two latter models, the Hausman test was conducted which delivered the Chi-squared test statistics of 30.17 and the associated p-value of 0.0004.
This means that the difference in coefficients is systematic and therefore the fixed effects regression should be preferred. Looking at the preferred fixed effects model, it is possible to see that the overall R-squared is 0.05, which means that 5% of the variation in the dependent variable is explained by the set of chosen independent variables, accounting for the heterogeneous effects associated with specific companies.
The R-squared within groups is 9% (higher than 5% between groups), which means that company-related effects can have a high influence on the estimation outcomes. Within the preferred fixed effects model only four variables were found to be significant. The first of them is the gearing ratio, with the negative coefficient of 0.05 significant at 1% (as the p-value is below 0.01). This means that all other factors being equal, as the gearing ratio goes up by 1%, the profitability measured by ROA is expected to go down by 0.05%.
Similarly, as the assets turnover ratio goes down by 1, the profitability is expected to go down by 7.36%, the coefficient being significant at 1%. Log of market capitalization is positively related to profitability, and as market capitalization goes up by 1%, the profitability is expected to go up by 0.12% (12.01 / 100), the result being significant at 1%. The last variable significant in the model is the log of employee number, with a negative 8.17 coefficient significant at 10%. This latter coefficient means that as the number of employees goes up by 1%, the level of profitability measured by ROE goes down by 0.08% (8.17 / 100).
Exploring the Determinants of the Return on Assets
The second set of estimated models is using a return on equity as a proxy for profitability. The simple OLS model(Table 3) shows that the variables of gearing, collection period, credit period, and the log of employees are significant with the negative sign of the coefficient, while the values of coverage and market capitalization are significant with a positive size. The associated value of R-squared is 0.25, which means that the selected independent variables explain close to 25% of the overall variation in ROA. Further estimation of the fixed effects and random effects models are similarly reflected in Table 3.
Similarly to the previous case, the Hausman test of fixed versus random effects gives the preference to the first model, with the associated Chi-squared test statistics of 47.16 and the associated p-value of 0.0000. Looking at the preferred model, one can see that the overall R-squared is 0.17. Also, the R-squared within is 0.15, which is higher than the 0.14 R-squared between, meaning that the higher impact on the coefficients can be caused by the within-group heterogeneity.
The coefficient of liquidity is positive 1.00, significant at 1%. This means that all else being equal, as the liquidity ratio increases by 1%, the associated profitability level is expected to also go up by 1%. The gearing ratio coefficient is the negative 0.01 significant at 5%. This coefficient means that the gearing ratio goes up by 1% the profitability is expected to go down by 0.01%. The positive 0.04 of interest coverage significant at 1% means that as the interest coverage ratio goes up by 1%, the profitability is expected to go up by 0.04%.
As the asset turnover goes up by 1, the profitability is expected to go down by 1.58%, the result being significant at 5%. Similarly to the case of ROE, the log of market capitalization is positively related to profitability, and as the market capitalization goes up by 1%, the associated profitability is expected to go up by 0.05% (4.85 / 100). Lastly, the variable of the log of employees was found to be negatively related to profitability. As the number of employees goes up by 1%, the associated profitability is expected to go down by 0.03% (3.25 / 100), the result being significant at 1%.
Exploring the Determinants of the Profit Margin
The third selected measure of profitability is the profit margin. Under the OLS framework, the variables of the log of employees, gearing, and credit period were found to be statistically significant with a negative coefficient, which the variables of the log of market capitalization and interest coverage being significant with the positive coefficient. The R-squared value for the simple OLS model was found to be 0.18. Given the panel nature of the data, the fixed and random effects models were further estimated, with the fixed effects model preferred similarly to the two previous cases (Hausman Chi-squared statistics is equal to 22.94, and the p-value is 0.0063).
The preferred fixed effects model showed the negative association between profitability and gearing significantly at 5%. As the gearing goes up by 1%, the associated profitability is expected to go down by 0.02, all else factors being equal. As the variable of interest coverage goes up by 1%, the profitability goes up by 0.03%, the result being significant at 1%. The credit period was found to be significant at 1%, and as the credit period goes up by one day, the profitability is expected to go down by 0.06%.
Similarly to the cases of ROA and ROE, the market capitalization is positively associated with profitability (1% increase drives profitability up by 0.07%), and the number of employees is negatively associated with profitability (1% increase drives profitability down by 0.03%). Having explored the outputs of the estimated regression equations and the levels of significance of the estimated coefficients, the next chapter will be focused on the interpretation of the obtained results within the broader context of the existing empirical evidence.
Discussion
The previous chapter presented the estimates of the regression equations calculated for the separate cases of the predictor variables of profitability – the return on assets, return on equity, and the profit margin. Figure 1 was constructed to briefly summarize the evidence collected within the findings chapter, make it easier for understanding and present within a single context.
From Figure 1 it can be seen that only two of the estimates were found to be producing consistent influence across the three estimated regression equation modification. One of them is the gearing ratio, which was found to produce a negative effect on profitability with each of the profitability proxies being used as the dependent variables. Similarly, the log of market capitalization was found to have a positive and significant impact on profitability in each of the three cases.
The indicators of inventory turnover and collection period were found to be insignificant in each of the regression equation modifications. Liquidity ratio was found to positively impact profitability in the model with ROA, but not in the two other models.
Similarly, the credit period was found to hurt profitability only for the model with a profit margin as the proxy for profitability. Both of the two variables of the log of employees and asset turnover hurt profitability in the models with ROA and ROE, but not with the model which uses profit margin as the profitability proxy. Lastly, interest coverage was found to have a positive impact on profitability for the case of the models with ROA and profit margin.
The key research question of this dissertation was as below:
- Can any of the selected firm-level indicators be the significant determinants of profitability of the UK industrial companies, basing on the dataset of the years from 2015 to 2017?
The associated null and alternative hypotheses were:
- H0: None of the selected firm-level indicators can predict profitability measured through ROE, ROA, or profit margins, and their associated coefficients are either equal to zero or insignificant;
- H1: At least one of the selected firm-level indicators can predict profitability measured through ROE, ROA, or profit margins, and the associated coefficient(s) is (are) non-zero and significant.
Judging by the outputs demonstrated in the previous chapter and Figure 1, it is possible to reject the null hypothesis above and claim that some of the selected indicators were found to be significant predictors of profitability. In particular, the gearing ratio and the log of market capitalization were found to be significant predictors in each of the three models. The answer to the research question posed at the beginning of the study is that indeed there are firm-specific indicators which can be interpreted as predictors of profitability for the case of the industrial companies of the United Kingdom.
The assumption of the positive relationship between the firm size and profitability is consistent with the previously discussed literature, such as Saeed (2014), Dimitroupoulos (2018), and Nanda and Panda (2018). The positive relationship between size and profitability can be partially explained by the effect of scale. For instance, the companies in some industries might only be able to achieve high levels of profits upon making extensive investments into some types of expensive capital assets.
Besides, as was previously mentioned in the literature review, some of the industries that are characterized by high levels of market concentration and competition are also associated with producing higher profits for the companies of the larger sizes. The fact that the gearing ratio was found to be negatively associated with profitability is also consistent with the previous literature such as Al-Jafari and Samman (2015) and Nanda and Panda (2018).
The latter authors are talking about the negative relationship between the dimensions of leverage and profitability, and the gearing ratio used in the current studies is one of the instruments of measuring the companies’ leverage. Leverage can be negatively associated with profitability due to the higher long-term risks associated with the operations of the companies that take on additional risk.
At the same time, for some companies, higher leverage can mean more efficient use of these funds for more rapid future development, and thus to not be a negative indicator. However, from the output of the regressions for the case of the industrial companies of the UK, it can be seen that leverage indeed comes out to be a negative factor of the financial performance of firms, being negatively associated with each of the three selected measures of profitability.
Having obtained the above-mentioned conclusions, it is possible to discuss some of the practical applications of the identified relationships. For the company’s managers in the UK in the current state of the economy, the results imply that taking additional leverage can be a suboptimal decision that has a possibility of negatively affecting profits.
For equity investors, the results may mean that in determining the stocks with higher profitability potential more emphasis should be made on the variables of market capitalization and leverage, as well as their potential future forecasts. Similarly, the creditors may also be better off giving preference to companies with higher market capitalization and lower levers of gearing, as this can increase the probability of receiving the future pay-outs of interest and principles that are directly related to the profitability dynamics of the considered enterprises.
At the same time, the conclusions summarized in Figure 1 can serve as the guidance for more specific cases of decision-making of equity investors, managers, and creditors once they are interested in the dynamics of a specific profitability indicator such as the return on equity, the return on assets or the profit margin.
Conclusion
As noted by Mendoza (2015), profitability determinants analysis can be specifically useful for such groups of corporate stakeholders as creditors, equity investors, and the management of the considered companies. Looking at the group of equity investors, profitability is important for them as this indicator can be one of the most important determinants of future price movements. In most general terms, income forecasting can be referred to as one of the key skills of a successful equity investor.
For the case of creditors, the ability to have insights into the companies’ profitability potential can be a predictor of the future incoming cash flows (such as the credit interest and principal) and their volatility. For the corporate management, profitability can be viewed as the measure of the quality of financial and investment decisions, as well as how effectively the revenues and profits are generated from the scope of available assets, sales, and shareholders’ equity.
In the current research paper, the focus was made on exploring the determinants of profitability of the industrial companies of the United Kingdom basing on the financial data reported across the years 2015, 2016, and 2017. The obtained data was obtained through the OSIRIS database and further processed with the help of the Stata software. Three different modifications of the regression equations were specified with the help of the varying depending variables measuring profitability – ROA, ROE, and the profit margin.
Also, three different regression frameworks were used in estimating the models of interest – the simple OLS, the random-effects model, and the fixed-effects model. The two latter frameworks are specifically used for handling the panel data samples, like the one used in the present study. The Hausman post-estimation test helped to select the preferred model between the random and fixed effects. For each of the three model modifications the fixed effects model was preferred, rejecting the null hypothesis of the Hausman test. As a result, it was found that the only two variables the results for which were consistent across the three models where the gearing ratio and the logarithm of market capitalization.
For the case of the gearing ratio, a statistically significant negative relationship between gearing and profitability was found for each of the profitability indicators. In contrast, higher levels of market capitalization were found to be associated with higher levels of profitability as measured by ROE, ROA, and profit margins. Some of the other variables were also negative and significant or positive and significant for some of the profitability proxies but not for the others.
Basing on the obtained evidence the practitioners should place a higher emphasis on examining the indicators of market capitalization and gearing when interested in the profitability of the UK companies. This can happen in the situation when the managers of a company are seeking more optimal ways of ensuring the successful future strategic development. Alternatively, the creditors and equity investors should consider the identified significant indicators when making important investment decisions in the UK market. The researchers should make further efforts into exploring the patterns of profitability of companies in the UK, as after the Brexit decision is officially finalized, the structural frameworks of companies’ profitability can change in the directions yet unknown.
Even though the current study was based on a high-quality dataset with detailed information on the financials of the selected UK industrial companies, it still has several limitations. One of such limitations could have been imposed by the selected inclusion criteria – the existence of the information on each for the selected variables for each of the observed years. This specific inclusion/exclusion criteria could have affected the observed relationships in unknown ways and directions.
This could happen because the factors due to which the companies chose to either to disclose or not disclose their full financial information can be in some way correlated with the studied determinants of profitability. One more limitation of the current paper is the fact that it did not take into account that some of the independent variables could be influencing profitability in a non-immediate manner, but with a certain lag.
For instance, the increase in the liquidity of a company could be positively influencing profitability not in the current year, but with a lag of two or three years, as some more operational cycles pass and ensure the multiplication effect. Similarly, profitability variables could also be subjected to auto-correlation which was not taken into account in the current research.
Looking at the possible future directions of research on the topic it is possible to suggest the analysis including the lags of independent and dependent variables and exploring if these modifications being any changes to the sizes and degrees of the significance of the estimated coefficients. Besides, some of the observed financial indicators could also have been subjected to short-term changes invariance that was later on returned to the equilibrium.
This represents the econometric issue of the variables having some of the co-integrated vectors of variance. These types of issues can be resolved by using the statistical framework of the Vector Error Correction Models, as suggested, for instance, in the papers by Amountzias et al. (2017) and Yusi (2018).
One more possible direction for future research could be exploring the influence of the industry dummies on profitability for UK companies. They could explain some part of the macroeconomic influences relevant for groups of companies rather than individual companies in the market. Also, some of the vectors of future research could be based on collecting more data on the same environment and verifying if the results of the estimation are consistent.
This type of validation can be ensured, for instance, by obtaining the quarterly data on the same indicators and the same companies and estimating the similar fitted models. Alternatively, this can be done basing on the data for the upcoming years, taking into account the heterogeneous influences in the economic, social, and political environment of the United Kingdom.
Reference List
Abduh, M., Omar, M.A. and Mesic, E. (2017) “Profitability determinants of Islamic and conventional banks in Malaysia: a panel regression approach”. Terengganu International Finance and Economics Journal (TIFEJ), vol. 3, no.1, pp. 1-7.
Alalaya, M. and Al-Khattab, S.A. (2015) “A Case Study in Business Market: Banks Profitability: Evidence from Jordanian Commercial Banks (2002 – 2015)”. International Journal of Business Management and Economic Research (IJBMER), vol. 6, no. 4, pp. 204-213.
Al-Jafari, M.K. and Al Samman, H. (2015) “Determinants of Profitability: Evidence from Industrial Companies Listed on Muscat Securities Market.” Review of European Studies, vol. 7, no. 11, pp. 303-311.
Almaqtari, F.A., Al‐Homaidi, E.A., Tabash, M.I. and Farhan, N.H. (2018) “The determinants of profitability of Indian commercial banks: A panel data approach.” International Journal of Finance & Economics, pp. 1-18.
Amountzias, C., Dagdeviren, H. and Patokos, T. (2017) “Pricing decisions and market power in the UK electricity market: A VECM approach” Energy Policy, vol. 108, pp. 467-473.
Azhagaiah, R. and Deepa, R. (2012) “Determinants of Profitability of Food Industry in India: A Size-Wise Analysis.” Management, vol. 7, no. 2, pp. 111-128.
Boadi, E.K., Antwi, S. and Lartey, V.C. (2013) “Determinants of profitability of insurance firms in Ghana.” International Journal of Business and Social Research, vol. 3, no. 3, pp. 43-50.
Charles, D., Ahmed, M.N. and Joshua, O. (2018). “Effect of Firm Characteristics on Profitability of Listed Consumer Goods Companies in Nigeria.” Journal of Accounting, Finance, and Auditing Studies, vol. 4, no. 2, pp. 14-31.
Dalci, I. (2018) “Impact of financial leverage on the profitability of listed manufacturing firms in China.” Pacific Accounting Review, vol. 30, no.4, pp. 410-432.
Dimitropoulos, P.E. (2018) “Profitability Determinants of the Greek Hospitality Industry: The Crisis Effect” In Innovative Approaches to Tourism and Leisure, Springer, Cham, pp. 405-416.
Ghamrawi, N., Ghamrawi, N.A. and Shal, T. (2017) “Lebanese Public Schools: 20th or 21st Century Schools? An Investigation into Teachers’ Instructional Practices”. Open Journal of Leadership, vol. 6, no. 1, p. 1.
Hirsch, S. (2018) “Successful in the Long Run: A Meta‐Regression Analysis of Persistent Firm Profits.” Journal of Economic Surveys, vol. 32, no. 1, pp. 23-49.
Jones, M.C., Dye, S.R., Pinnegar, J.K., Warren, R. and Cheung, W.W. (2015) “Using scenarios to project the changing profitability of fisheries under climate change.” Fish and fisheries, vol. 16, no. 4, pp. 603-622.
Magoutas, A., Papadoudis, G. and Sfakianakis, G. (2018) “Determinants of profitability in the Greek tourism sector – assessing the effect of the crisis.” International Journal of Tourism Policy, vol. 8, no. 1, pp. 65-72.
Mary, O.I., Okelue, U.D. and Uchenna, A.S. (2012) “An examination of the factors that determine the profitability of the Nigerian beer brewery firms.” Asian Economic and Financial Review, vol. 2, no. 7, p. 741.
Mendoza, R.R. (2015) “Predictors of Profitability of Micro, Small, and Medium Enterprises in the Philippines.” International Journal of Business Research, vol. 14, no. 3, pp. 35-46.
Mohamed, A.A. and Jones, T.A. (2014) “Relationship between strategic management accounting techniques and profitability-a proposed model.” Measuring Business Excellence, vol. 18, no. 3, pp. 1-22.
Montoro-Pons, J.D. and Cuadrado-García, M. (2018) “” Let’s make lots of money”: the determinants of performance in the recorded music sector.” Journal of Cultural Economics, vol. 42, no. 2, pp. 287-307.
Nanda, S. and Panda, A.K. (2018) “The determinants of corporate profitability: an investigation of Indian manufacturing firms.” International Journal of Emerging Markets, vol. 13, no. 1, pp. 66-86.
Newth, F., Hansen, D., Kalinowski, B., Kokias, P., MacEachin, R., Marshall, R., Napolione, K., Smith-Wilks, B. and Williams, C. (2018) “Brexit Report-Impact on Business Models of Scottish Companies.” Business Commons, pp. 1-38.
Onuonga, S.M. (2014) “The Analysis of Profitability of Kenya’s Top Six Commercial Banks: Internal Factor Analysis.” American International Journal of Social Science, vol. 3, no. 5, pp. 94-103.
Pacini, K., Mayer, P., Attar, S. and Azam, J. (2018) “Macroeconomic Factors And Firm Performance In The United Kingdom.” Journal of Smart Economic Growth, vol. 2, no. 3, pp. 1-11.
Petria, N., Capraru, B. and Ihnatov, I. (2015) “Determinants of banks’ profitability: evidence from EU 27 banking systems”. Procedia Economics and Finance, vol. 20, pp. 518-524.
Quader, M.S., Kamal, M.M. and Hassan, A.E. (2016) “Sustainability of positive relationship between environmental performance and profitability of SMEs: A case study in the UK.” Journal of Enterprising Communities: People and Places in the Global Economy, vol. 10, no. 2, pp. 138-163.
Raashid, M., Rasool, S.A. and Raja, M.U. (2015) “Investigation of Profitability of Banking Sector: Empirical Evidence from Pakistan.” Journal of Finance, vol. 3, no. 1, pp. 139-155.
Saeed, M.S. (2014). “Bank-related, industry-related and macroeconomic factors affecting bank profitability: a case of the United Kingdom.” Research Journal of Finance and Accounting, vol. 5, no. 2, pp. 42-50.
Salike, N. and Ao, B. (2018) “Determinants of bank’s profitability: the role of poor asset quality in Asia.” China Finance Review International, vol. 8, no. 2, pp. 216-231.
Yazdanfar, D. (2013) “Profitability determinants among micro firms: evidence from Swedish data.” International Journal of Managerial Finance, vol. 9, no. 2, pp. 151-160.
Yusi, Z. (2018) “Research on Bank Profit Level from The Perspective of Interest Rate Liberalization – Empirical Analysis Based on the VECM Model.” Journal of Jilin Financial Research, vol. 1, p.003.
Zimmerman, S. and Bell, J. (2014) The sustainability mindset: Using the matrix map to make strategic decisions, John Wiley & Sons, San Francisco.