Despite the increase in literature on financialisation, most fails to cross theoretical paradigms with empirical data at firm level. Furthermore, very few sectorial studies have been conducted to date. This paper aims to bridge this gap by estimating the importance of sectorial affiliation in explaining the financialisation of large companies in the United Kingdom and Spain. As the definition of financialisation is often imprecise and unclear, we begin with a discussion of the limitations of indicators used in financialisation literature. Secondly, we build a financialisation ratio based on a unique dataset obtained from balance sheets and income statement variables of the companies listed on the FTSE -100 and Continuous Market index of 2000-2015. Finally, and for the first time in literature, we have applied a multilevel methodology in order to study similarities between the estimated sectoral patterns for Spain and the United Kingdom. Preliminary results reveal a relevant influence of the corporate sector in financialisation ratio variability. The paper opens up a new theoretical and methodological research agenda for analysing financialisation.

A pesar del aumento de la literatura sobre financiarización, la mayoría de las contribuciones no ofrecen una vinculación satisfactoria de los paradigmas teóricos con los datos empíricos a nivel de empresa. Además, hasta la fecha se han realizado muy pocos estudios sectoriales. El objetivo de este trabajo es cerrar esta brecha estimando la importancia de la afiliación sectorial en la financiarización de las grandes empresas en el Reino Unido y España. Dado que la definición de financiarización suele ser imprecisa y poco clara, comenzamos con una discusión de las limitaciones de los indicadores utilizados en la literatura sobre financiarización. En segundo lugar, construimos un indicador de financiarización basado en un conjunto de datos único, obtenido de los balances y las variables de las cuentas de resultados de las empresas que cotizan en los índices FTSE -100 y del mercado continuo de 2000-2015. Finalmente, y por primera vez en la literatura, hemos aplicado una metodología multinivel para estudiar las similitudes entre los patrones sectoriales estimados para España y Reino Unido. Los resultados preliminares revelan una influencia relevante del sector empresarial en la variabilidad del indicador de financiarización. El artículo propone una nueva agenda de investigación teórica y metodológica para analizar la financiarización.

The concept of financialisation is widely used to describe certain structural changes in advanced economies. The process of financialisation is associated not only with an expansion of the financial system, but also with a change in the behaviour of non-financial actors, the source of their monetary earnings and the structure of their balance sheets (Davis,

Most of the contributions on financialisation studies fail to cross appropriately theoretical approaches with empirical data at firm level. The aim of this paper is to discuss some of the difficulties involved in measuring the financialisation of non-financial corporations (NFCs) and to propose a multilevel modelling approach for estimating the importance of sectorial affiliation to explain financialisation. These difficulties have to do basically with two features of different nature: the imprecision of the literature in defining financialisation from an operational way and the characteristics and nature of the available data. Our proposed methodology of measurement is illustrated by estimating sectoral financialisation ratios in two different countries, namely, Spain and the United Kingdom, for one of the possible channels of corporate financialisation, related to income origin.

Studies carried out to date have linked financialisation to a management model based on a shareholder value firm approach that emerged in the 1980s. However, the empirical measure of financialisation in existing scientific literature is often unclear and imprecise (Davis,

Some authors have pointed out the importance of considering the role of the corporate sector in financial investment and the differences between listed and non-listed firms (Davis,

The principal studies in this area have summarised corporate financialisation in a few “stylized facts” (Davis,

The fundamentals of the case selection process lay on our aim of analysing to what extent financialisation is more dependent of the structural characteristics of the sector of activity rather than the institutional settings that situate the UK as an accumulation model dominated by finance (Jessop, 2013) and Spain as a mix market economy (Molina and Rhodes,

The main contributions of our research are the following. Firstly, the paper provides crucial yet hitherto unavailable evidence for measuring financialisation in sixteen industries. For first time in the literature, we use multilevel modelling techniques to study whether part of the variability of the firm level financialisation ratios is given by sectoral variability. Secondly, we use those sectoral estimates to compare their patterns in two samples of listed firms from Spain and the United Kingdom. Thirdly, we propose a new theoretical and methodological research agenda for analysing financialisation in various corporate sectors, using firm-level data and multilevel modelling.

We show that 13-14% of the dispersion of the (logarithms of the) financialisation ratios of the firms in our samples of Spanish and UK corporations is attributable to differences between sectors. Additionally, we find preliminary evidence of a significant moderate positive correlation between the sectoral financialisation structures of both countries.

The paper is organised as follows. Section two discusses the definition of financialisation, introducing two main approaches: the regulation theory and shareholder value contributions. Section three presents a methodological discussion of financialisation at firm level, detailing the strengths and weakness of existing indicators. Sections four and five introduce the main hypothesis and the characteristics of dataset drawn up using firm level data of listed businesses in the UK and Spain. Section six details the multilevel methodology developed to measure the influence of sectorial affiliation to explain the financialisation ratio. Sections seven and eight present the main results and conclusions regarding the sectoral convergence of firm financialisation.

Financialisation can be broadly understood as the spread of the financial system, new legal and technological structures, as well as new models of relationships and interaction between economic actors and the financial sphere.

In this section we introduce two main approaches to the concept of financialisation in order to contextualise the specific objective of this paper. The first approach has been developed by the regulation theory and discussed by comparative political economy approaches. The second is centred on the shareholder value approach to firms based on contractual or agency theory and discussed in financialisation studies.

We aim to explore some elements of both approaches that have received scant attention to date. On the one hand, we attempt to shed light on financialisation within the varieties of institutional settings and more specifically within the varieties of sectors of activity. Unfortunately, there is no specific literature available on sectoral characteristics and the process of financialisation, thereby indicating a gap that must be filled in the future. A further objective is to identify the main indicators of the shareholder approach. Both insights provide the theoretical basis for a new research agenda on financialisation based on identifying regularities and differences at industry level in different spaces of financialisation.

According to this approach, the increasing role of financial motives, financial markets, financial actors and financial institutions in the operation of domestic and international economies is a consequence of economic globalisation and de-regulation (Epstein,

The regulation theory approach has provided a useful conceptual structure and indicators for analysing the changing nature of the economy. Within this framework, studies on financialisation have addressed the relationship between falling profitability in the productive system and the search for new sources of profitability of non-financial firms in financial markets. More specifically, Krippner (

Evidence provided by Krippner (

Comparative political economy approaches (CPE) reject the idea of convergence towards financialised capitalism insofar as finance-led capitalism is in itself an idiosyncratic form of capitalism (Kalinowski,

In addition, regulation theory and political economy approaches do not provide an explanation as to why firms financialise and why certain corporate structures and their position within production chains allows them more latitude to financialise (Soener,

The functionalist assumption of an evolution towards a single model of financial capitalism is based on a generalisation that cannot be applied to all industries and the behaviour of all non-financial corporations. Comparative political economy studies consider national states as a unit of analysis, simplifying the variety of industries and productive processes or the relationship between firm productive roles and global production. As with other reductionisms, financial capitalism is more a conceptual tool in exemplifying a process of change encompassing a wide variety of structures and strategies of business growth.

Given these limitations of the comparative political economy and regulation theory approaches, the present paper tries to identify possible similarities and differences in sectoral patterns of financialisation under two different institutional settings, Spain and the UK, that allegedly represent different models of financialisation.

The second approach to the concept of financialisation has a narrower scope. It focuses on the emergence of a management paradigm called “shareholder value” (Lazonick and O’Sullivan,

Recent reinterpretations of the shareholder value orientation stem from contractual corporate theories, such as agency theory

Later, in the 1990s, the Agency Theory was associated with processes of corporate restructuring and downsizing, justified because they guaranteed high returns for capital owners and managers. This approach brings together a large and diverse number of contributions focused on the study of the relationships between non-financial companies, the financial sector and the rules that govern the employment relationship and give rise to a new model of wage relations (Ruesga,

In sum, literature on financialisation has linked changes in corporate financial behaviour to shareholder value orientation (Lazonick and O’Sullivan,

Despite the growing interest in the process of financialisation in socio-economic literature, there is a lack of consensus regarding how to measure financialisation at firm and industry level from a quantitative and qualitative approach. One of the main reasons for that is the lack of consensus on how the process of financialisation has changed the behaviour of NFCs (Barradas,

Literature highlights three different channels of NFC financialisation (Orhangazi,

As mentioned above, studies conducted to date on the changing behaviour of NFCs are based on considerations regarding the evolution of those stylized facts that depict a broad process of structural change (Davis,

For the purpose of this paper, we have created a financialisation ratio in order to find synthetic evidence of financial sources of income. Following some of the most influential contributions (Stockhammer,

Notes: The table does not include an exhaustive review of literature. Instead it highlights some of the main contributions cited in the literature of non-financial corporation financialisation that use empirical data.

Most studies focus the discussion of this ratio on whether the increase in financial profits is done at the expense of physical investment (Davis,

In addition, analysing the financial earnings has a limited scope because these instruments are not always a faithful reflection of the financial behaviour of NFC, fundamentally for three reasons: 1) accounting practices allow for different ways of calculating profits and amortizations (Miller,

Finally, the volatility and uncertainty of financial earnings is a key issue for consideration. To date it has not been addressed by literature and it is extremely difficult to find a robust interpretation of the financialisation ratio.

In order to fill the gaps in existing literature, we seek to test empirically the hypothesis that there are certain patterns of behaviour which are common to particular industries that depend more heavily on financial markets. The comparative approach of this paper leads to a dual hypothesis.

The first refers to the influence of sectorial affiliation to explain financialisation ratios. We hypothesize that the likelihood of financialising is conditioned not only by firm level decisions as a consequence of the declining of profitability rates, as stated by most literature (Crotty,

The second hypothesis refers to the existence of similar patterns of sectoral financialisation in the UK and Spain. This hypothesis would reinforce the previous one, provided that the existence of similar degree of sectoral influence in different national spaces confirms the existence of structural sectoral characteristics to face the worldwide challenges of big companies competing in a global market.

We compiled integrated time series at a corporate level from 2000 to 2015, from Thomson Reuters Eikon Desktop. This virtual desktop contains key economic data about the firms listed on the FTSE index in the case of the UK, and the Continuous Market for Spain. The only data cleansing was the exclusion of cases with a non-positive financialisation ratio. That was enough to exclude the firms in the financial and insurance sectors.

Firm level data of our financialisation ratio is volatile. For instance, specific large-scale corporate transactions may increase the ratio from 0.5 to 3.0 from one year to the next. The variation coefficient fails to provide and exhaustive vision. Firms with only one year of data will have a missing coefficient, so the sectoral mean of this indicator may be given by the data of one or a very small number of firms. Additionally, firm level coefficients across time can be calculated for 2, 3, 4 or 5 years, and not necessarily sequentially. However, the indicator is enough to illustrate the heterogeneous volatility of the data. In this context, we have included those sample cases with an extreme financialisation ratio, namely higher than 10 or 30. On the one hand, the volatile nature of the data prevents us from establishing a criterion regarding a “sensible” range of data to clean the data bank from outliers. On the other hand, the data do not allow us to distinguish between certain operations or accounts which are essential for measuring financialisation from others that should be ignored. We have therefore kept all financialisation ratio observations as they were calculated. This is also appropriate in order to show the strength of multilevel modelling when dealing with unbalanced and fuzzy data.

Note: The means in column (4) are first calculated by each firm, across years, and then those firm averages are later averaged for the sector. The means in column (5) are averaged for all available cases in each sector. The analogous is true for the medians in columns (6) and (7). The coefficient of variation is first calculated for each firm across time, as in (4) and (6), and then averaged for the sector. The row TOTAL includes the sum of columns (1) and (2), the ratio of those totals for column (3), the mean for columns (4), (5) and (8), and the medians for columns (6) and (7).

Note: see the note of Table 2.

The four central tendency measures in the descriptive table statistics show the consequences of this data heterogeneity (see the footnote of

In the previous section we saw that our data on financialisation have a wide range of sample sizes for each sector and variation across both firms and time. Our first research question addresses the possibility of some degree of sectoral correlation in those data: what proportion of the total dispersion of the data is due to sectoral heterogeneity? Moreover, the theory outlined in section 2 predicts the possibility of similar sectoral patterns of financialisation in different countries, which requires estimating and comparing sectoral financialisation ratios.

While standard regression models ‘average’ the data, multilevel (mixed or hierarchical) modelling takes into consideration the sample size of each group and several types of data variability. This makes it a suitable technique for modelling complexity and heterogeneity. Multilevel modelling is the appropriate technique when considering nested or clustered data, breaking away from the traditional assumption of independent observation owing to data dependency (correlation) at different levels of data aggregation.

We are going to consider three-level data of a financialisation ratio _{tfs}_{t··}_{f}_{f}_{tf·}_{s}_{s}

Here we are only interested in estimating a sensitive financialisation ratio by sector. This is done with an empty random intercept model, in which the term ‘empty’ (or ‘null’) model refers to the absence of explanatory variables.

_{tfs}_{fs}_{tfs}_{tfs}_{tfs}

where _{fs}_{tfs}_{tfs}_{f}

The level two model allows us to estimate a financialisation ratio _{fs}_{f}

_{fs}_{s}_{fs}_{fs}_{fs}

where _{s}_{fs}_{fs}

Finally, the level three model identifies a financialisation ratio for each of the _{s}

_{s}_{0}_{s}_{s}_{s}

where _{0} is a global intercept for all the observations (cases of firms in sectors) and the _{s}_{s}_{s}

Putting them all together provides us with a full model:

_{tfs}_{0}_{s}_{fs}_{tfs}

where we are interested in the _{s}_{0 }+ _{s}_{f}_{0 }+ _{s }_{fs }_{tfs}

_{tfs}_{s}_{fs}

The Intraclass Correlation Coefficient (_{fs}

_{fs}_{s}_{fs}_{s}_{fs}

We focused on the proportion of the dispersion of the financialisation ratio given by the variability between sectors, which is given by the following sectoral

_{s}_{s}_{fs}

As mentioned above, each of the group ‘random effects’ _{fs}_{s}_{fs}_{s}_{0} (the global average of _{tfs}

Estimates of group effects are weighted averages that combine information from the group itself with information from the mean for all groups. Estimating with random effects is a conservative approach, giving less weight to less reliable data. The random effects are precision-weighted residuals called ‘posterior residuals’, ‘empirical Bayes estimates’ or ‘shrunken residuals’. When the sectors are very similar (low between-sector variance), sectoral residuals _{s}_{0}: _{s}_{0} + _{s}_{s}_{s}_{0}. In other words, if the data of a sector are not reliable, the estimate for that sector borrows information from the global sample. That is an advantage over the alternative estimation with sectoral fixed effects or dummy variables, in which the sectoral estimates are arithmetic means, independent of the data reliability.

Our goal is to estimate global and sectoral financialisation ratios in a way more robust to data problems than the arithmetic means showed in Tables _{s}

We estimate these models by restricted maximum likelihood using the “lme4” package (Bates et al.,

Our dependent variable is the natural logarithm of the financialisation ratio, which has an approximately normal distribution for our samples of cases in Spain and the United Kingdom. _{0} and the (weighted) global mean of the logarithm of the financialisation ratio.

The Intra-class Correlation Coefficient at sectoral level (_{s}

Note: Correlations between the figures for both countries are Pearson correlations for the logarithm of the financialisation ratio and for its exponential. Spearman correlation was used for the sectoral rankings. The p values of those correlations compare the null hypothesis of zero correlation with the alternative of correlation greater than zero. A p value smaller than 0.05 implies the acceptation of positive correlation.

From the models shown in _{s}_{0} + _{s}

The arithmetic mean of financialisation for Spain was 2.6 (^{–0.69} = 0.50). This shows that the arithmetic mean of financialisation was severely affected by extreme values, while the multilevel estimate implies that listed Spanish corporations tend to have a financial income equivalent to the 50% of their net sales. That is similar to the multilevel estimate for the United Kingdom, which is 55% (^{–0.60} = 0.55), even though the arithmetic mean of cases in

The multilevel sectoral estimates can be used to study possible similarities between the estimated sectoral patterns for Spain and the United Kingdom. For this purpose, we considered both the sectoral ratios and the sectoral rankings of those ratios, in which number one corresponds to the sector with lower financialisation and number 16 to the sector with the highest ratio. We found preliminary evidence of a significant moderate positive correlation (see the note in

In terms of the sectoral ranking, technological, telecommunication and real estate industries in both countries have the highest financialisation ratio. These similarities are attributable to the structural characteristics of the technological and telecommunication sectors, which are based heavily on brand imagining and on a particular innovation and risk structure (Mazzucato,

A sectoral pattern in the data, with an of around 10-15% of the total variance of the (log) financialisation ratio, is a robust finding in our multilevel models for some estimations after cleaning outliers or making small sectoral aggregation changes. However, specific results about sectoral similarity should be taken with caution. Our sample size is limited and multilevel techniques behave better when the number of groups is high.

Indeed, the confidence intervals of our estimated sectoral deviations from the global intercept (estimates of _{s}

Moreover, even if the statistical results were more conclusive, we could not make any assertions regarding causality. Similar sectoral patterns of financialisation between different countries might be the consequence of a process of sectoral convergence towards a common global market and business structure. However, they might also be due to common technical characteristics of the firms in those sectors: technological processes; the business cycle; regulation trends; trade patterns; cash flow features; sensitivity to global, regional or local conditions; particular strategies of leading companies in each sector and so on.

Despite the growing amount of literature on financialisation, most fails to address theoretical paradigms with empirical data at firm level. Additionally, sectorial studies are very scarce. In this paper, we have argued that financialisation is a structural process of change of NFC behaviour, that affects a range of industries to a greater or lesser extent, depending on their structural characteristics. These include the productive roles played by firms in global production chains or the importance of intangible assets such as brands or patents relative to physical assets. Following Soener (

This paper describes a multilevel analysis of a sample of 843 companies to examine the influence of the sector in the variability of a firm-level financialisation ratio. We have selected 326 listed companies from the United Kingdom and Spain between 2000 and 2015. The data reveal the difficulties associated with measuring financialisation at firm level such as the problem of data volatility, the existence of extreme cases and abundant missing data.

After calculating a financialisation index for each firm in each period, we applied a multilevel model to estimate precision-weighted sectoral financialisation ratios for Spain and the United Kingdom. Those estimated sectoral financialisation ratios are more robust than firm-level arithmetic means to the data problems noted above.

The results confirm our first hypothesis: around 13-14% of the dispersion of firms’ (log) financialisation ratios is due to sectoral heterogeneity; namely to differences in financialisation between sectors. This implies that the structure and nature of certain industries may shape the financialisation process. In addition, and in line with our second hypothesis, the sectors with the highest and lowest financialisation ratios are similar in both countries.

Our analysis has a series of shortcomings, attributable to a number of factors: the availability of data, their unambiguous interpretation in terms of the financialisation theories, the influence of outliers and the impact of using a particular classification of sectors. However, the results are encouraging. From the point of view of a methodological research agenda, multilevel modelling can be a suitable technique to study financialisation at firm level, both in order to compare the trends in a country over time and to compare countries for several possible sectoral aggregations. Empirical financialisation literature may benefit greatly from applying these techniques to larger firm samples, focusing on detailed sectors for a broad sample of countries over extended sample periods. Empirical studies should also model changes in financialisation, rather than merely the level, and study firm-and-sectoral-level covariates. This approach will provide the necessary evidence to rebuild financialisation theories close to the data. Only a vast accumulation of empirical evidence will allow us to disentangle the varying and complex causal relationships and trends operating simultaneously at different levels of aggregation.

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 746622, and the project “Social consequences of financialisation: monetary accumulation, lack of employment, social inequality” funded by the Ministry of the Economy and Competitiveness- Government of Spain, AEI/FEDER, UE, reference number: CSO2016-78122-R.

This measure has been used by Alvarez (

Following Ertuk et al. (

A complete description of the model assumptions can be found in Snijders and Bosker (

Technically, the variance partition coefficients (_{s }_{s}_{f} = σ_{fs}_{s}_{fs}_{fs}

In the same token, in a model without explanatory variables, the global intercept _{0} estimated by ordinary least squares is the arithmetic mean of _{tfs}_{0} gives more weight to the groups with the most reliable data.

See Stegmueller (