Competition in the banking sector, real or not? : a non-structural approach of the ecuadorian experience

(1)

COMPETITION IN THE

BANKING SECTOR, REAL OR

NOT?

A NON-STRUCTURAL APPROACH OF THE ECUADORIAN EXPERIENCE

Paulo Bermeo B.* July 2016

Keywords: banking, competition, market structure, Panzar and Rosse

* University of Amsterdam · Faculty of Economics and Business · Industrial Organization · Roetersstraat 11, 1018 WB Amsterdam · The Netherlands · www.fee.uva.nl · paulo.bermeo@gmail.com · Master of Sciences in Economics · Major in Industrial Organization · Master Thesis (15 ECTS) · Student ID 11084502 · I would like to thank the following people and organizations for helpful support, comments and discussions: Dr. Maarten Pieter Schinkel, the supervisor of this thesis, Pablo Zumarraga and the SBS for their data, and Paulo Bermeo M. and Sylvia Brazzero for their patience and suggestions · All errors are my sole responsibility.

Under supervision of Prof. Dr. M.P. Schinkel

(2)

Statement of Originality

This document is written by Student Paulo Bermeo who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

-1. Introduction

Nowadays, concentration, competition and markets’ competitive structure are, without question, important economic issues for policy makers. These matters are in constant need of improvement and regulation. Aiming for competitive non concentrated markets, allegedly, enhances consumers’ welfare. But what happens in industries with complex structures, like the banking sector or the financial services in general, where concentratio n may not be a sufficient proxy for a lack of performance.

The aim of this research is to assess correctly the market structure of the banking system, in order to gauge the level of competition present in the industry, and give some inferences on the type of competition behavior this kind of market faces. The empirica l analysis will be based on the specific case of the Ecuadorian Banking Sector. A non-structural approach, based on the Panzar & Rosse (1987) model and its recent improvements, is used for obtaining the desire conclusions. Recent improvements on the model solve the biased results from the original version by including a dynamic analys is. A suitable fit for the Ecuadorian reality is expected.

Studying competition and modeling its behavior has been a great deal for economists and econometricians involved with market analysis topics. Many models have been developed in this matter, from the traditional Industrial Organization to the New Empirical Industrial Organization (NEIO). In the later we found to different approaches to solve the problem of competition assessment: a structural approach and a non-structura l one. The non-structural approach has gain some importance in addressing the market behavior in for industries like the financial services, with complex costs functions and earnings structures.

Assessing the degree of competition can help policy makers to regulate and take measures in economically determinant markets, such as the financial sector. Given the importance that financial markets have in the global Economy it is important to keep an eye on the market behavior and the performance of the firms. Many studies have address this issues for different countries, trying to find the best empirical model to suit its reality. In the Ecuadorian case, there’s not been yet an individual analysis of the financial system, policy

(5)

makers have been taking decisions on this matter based on the world’s regulatory tendency, without a proper inside analysis. It’s been intended in the last couple of years to issue an exhaustive study an investigation of this industry, but not advances have yet been known, even though the Government has been regulating the industry.

The Ecuadorian Financial System is, in more than an 81%, constituted by banks. It makes sense that the analysis undertaken in the thesis focuses only on the Ecuadorian Banking Sector and not on the whole Financial System. Then, the inquiry of “How competitive is

the Ecuadorian Banking Sector?”, is raised as the main research question.

To solve this important issue, it is necessary to discuss: the structure of the banking sector in Ecuador; the type of competition that it faces, the evolution of competition over the years; the effect of political stability and recent regulations on the competition structure; and the closing issue, is to know if the results are supported by Industrial Organizat io n Theory, specifically, analyze if the Panzar-Rosse model can gauge the competit io n structure in an adequate way for this Industry.

In this research, a description of the Ecuadorian financial industry is performed, focusing mainly on the banking sector, which is going to be used for empirical proposes. To establish an adequate empirical model, it is first reviewed the theory involvi ng competition analysis, ending with recent applications of the Panzar-Rosse model and its empirical modifications. To conclude the study, the results on the Ecuadorian Banking system will be described as well as the possible recommendations.

2. The Financial Industry in Ecuador

The Ecuadorian financial system consists of institutions that capture the savings of their customers and provide these resources to those who are in need of them. The process of the redistribution of these resources is call financial intermediation. This industry is constituted by public and private financial institutions, regulated by:

 Superintendencia de Bancos y Seguros (The Banking Authority)

o Banks: receive the savings of their clients and place these funds as loans to the public.

(6)

o Mutualistas (Mortgage Companies): capture public resources for housing and construction loans.

o Financial Societies: grant loans and capture investments.

o Credit Card Companies: allow user for a credit line to buy goods and services on affiliated establishments.

 Superintendencia de Economía Popular y Solidaria (The Credit Union Authority) o Credit Unions: formed by natural or legal persons, grouped voluntarily to

perform financial intermediation activities and social responsibility with its members.

The regulatory agencies are responsible for supervising that the financial institut io ns comply with the law. They work together with the Central Bank of Ecuador and, act when there are liquidity problems. In Ecuador, the General Law for Financial Institut io ns establishes which operations are allowed and which ones are prohibited for this institutions. It details the obligations to be met with the regulatory agencies, ensuring that the financial system remains in an optimal condition, increasing the depositors’ trust and confidence.

There is no question that stability in financial markets is necessary to be sustained, a healthy financial system can boost other industries when necessary. Hence Competitio n authorities all over the world have turned their sights to this industry. In Ecuador, after the major financial crisis lived in 1999, which cause the foreclosure of many of these institutions, many monetary policies have been designed to secure stability in this industry.

Since the outstanding economic growth from 2011, the Ecuadorian financial industry was facing a remarkable growth and stable period, that just ended in 2014, as a result of the economic slowdown produce by the decrease in oil prices. The Ecuadorian Government, following the global tendency of regulating this industry, modified the Organic Monetary and Financial Code in this same year, claiming that regulation on financial activities will improve competition and stability.

(7)

Parallel to this, the Ecuadorian Government, in 2013, created The Ecuadorian Competition Authority, Superintendencia de Control del Poder de Mercado (SCPM), and released the first official Organic Law in matters of Competition and Regulation of the Market Power. In its first official report of recommendations, the SCPM claimed that it was necessary to improve the transparency of information to optimize competition in the Financial Services Industry. The SCPM, in the same year, released a general analysis of concentration and inequality for all the industries in Ecuador; the Financial Services Industry was ranked as the 8th_{most concentrated industry out of 400. Considering that}

banks represent approximately 81% of the whole Ecuadorian Financial system, in 2014, the SCPM begun gathering information from banks and other financial institutions but not resolution has been emitted yet.

2.1. The Banking Sector

Banks in Ecuador are the most important financial institutions. They are divided in Private Banks and Public ones. The latter is concentrated in granting credit loans for social developing and micro-productive activities, representing the 7% of all the banking system in terms of total assets. Private banks are the other 93%, and they have a wider catalogue of financial services offered.

After the Ecuadorian financial crisis of 1999, there were, in January 2000, a total of 32 private banks engaging in financial activities. By March 2016 there are registered 21 institutions in total, in 2014 there were 25 private banks. The decrease in the quantity of banks can be attributed to market adjustments due to the current recession the Ecuadorian economy is suffering. This sector is characterized for its market concentration, the seven major players, by 2015, represented the 87.4% of the market, in terms of total assets. The concentration of the industry has been growing, particularly in the last couple of years. In 2013 the seven major players had a joint market share of 83.9%, and in 2014 it was 86.7%.

It is argued that this increase in market concentration levels are due to the reputation that this major banks have on the market, granting better complementary services and a better access to cash disposals, given their wide presence over Ecuadorian territory. Lafuente & Valle (1997) identified and established five factors that impact Ecuadorian banks’ performance and their market shares. This factors are size, administrative efficie nc y,

(8)

capitalization, liquidity, and assets quality. Is important to consider the year of this publication to make assumptions on the current market structure.

On a more recent study, Baquero D. (2013) performed some analysis on the banking structure, using data from 1994 until 2012. The study shows through the HHI index analysis that during the 90s the Ecuadorian Banking Industry was not concentrated, but after the 1999 crisis, the industry was catalogue as mildly concentrated. The vast number of banks influenced negatively the concentration levels before the crisis. After the crisis, the consolidation of some of the major banks increased the HHI index values. It was also pointed out that since the 90s this industry has been characterized by a monopolist ic competition, with no trace of collusive behavior. Also there’s no evidence of a bank having a dominant position in the industry. Which means that the concentration level seen in the Ecuadorian banking sector does not allow the existence of dominant positions, hence there is no evidence of any type of market power abuse. These results were also validated by the Ecuadorian Central Bank, as well as the International Monetary Fund (IMF). Further studies performing deep analysis of competition and market structure in last years have not been done. Most of the reports are based on concentration indexes and performance comparisons between banks.

In geographical terms, the two major cities in Ecuador, Quito and Guayaquil, concentrate more than the 60% of the banking activity in terms of clients. Figures provided by the Superintendencia de Bancos y Seguros (SBS) on 2015 showed that this relationship also holds for offices and ATMs. It is also worthy to mention that the banking index grew from 24.98% on 2005 to 46.21% on 2015.

Rodríguez (2006) mentions that there is no evidence of price fixation agreements between entities in terms of the services offered by the banks, given that the regulations ruling this market have efficiently contributed to free market structure in terms of price fixatio ns. Each entity has the right to determine its own prices according to the regulation. On the other hand, in terms of the interest rates, the Ecuadorian regulation stablishes a minimum and a maximum rate, which implies a tacit fixation of prices. This does not allow the existence of an adequate level of competition between entities in terms of financ ia l intermediation prices, the current market structure causes that banking operations, depending on the corresponding market segment, tend to settle on the roofs of the

(9)

maximum conventional rate with the exception for the mortgages loans and the corporate credits.

The rigidity of prices conditions the strategies that entities can pursue to compete in the market and impairs the efficiency of the system. Nowadays it has been noticeable that some Ecuadorian banks, in order to get more clients into the banking system and remain in the market, have been lowering interest rates at a point that is barely possible to cover costs, this can be evidence of some type of price war in the different financial markets in the industry. The Ecuadorian Central Bank (2016) published an analysis on the average interest rates showing that depending on the market segment, certain banks manage interest rates that differ downwards from the average rate of that segment. Banks lower the interest rates on the financial products they wish to sell more or increase their participation.

In terms of market barriers, the Ecuadorian banking sector does not limit the entry of new banks or foreign entities. Perhaps the biggest constraint presented by the market is the political risk, which implies the constant political changes (which in the last years has been counteracted by the prolongation of the office term of the current president) or decisions that may affect economic activity operators, such as price controls, new regulations, the introduction of subsidies, etc.

By the end of 2014, from the whole financial system, private banks represented the 75,9% of total assets, 77,2% of total gross loans; 79,3% of total liabilities, 80,9% of the public deposits; in other words, approximately the two third parts of the Ecuadorian Financia l System are held by private banks. Considering this, it makes perfect sense to aim this research on assessing competition based on a sample of the Ecuadorian private banks.

3. Literature Review: Competition Assessment

Competition has always been an important issue to economists. The existence of competition in a market economy represents a central foundation for economic and social development. Competitive markets are believed to incentivize innovation in production processes, reducing costs, and innovation in products, extending the diversity of options a consumer can get. The end game is to maximize social welfare.

(10)

Industrial Organization, a branch of economic studies, has dedicated its efforts in analyzing different market structures and their economic relevance, mainly addressing competition issues. Studying competition and revenue levels in different industries has become a major focus for economists and policy makers. It is important to use appropriate methodology to assess properly the competition and concentration levels of a specific industry. Obtaining accurate results can be useful to develop correct regulations and measures, that can be used to enhance social welfare.

The literature referring competition and how to gauge it can be divided in two approaches: structural and non-structural. The first one refers to assessing the degree of competit io n from the market structure, while the second approach, as part of the New Empirica l Industrial Organization, infers the level of competition by the observation of the firms’ behavior in the market. Given that many models have been developed to evaluate different industries, it is important to recognize the peculiarities that each market has, regarding its structure and behavior. The more accurate the measure method and data, the more precise the empirical results will be.

The analysis performed in this research is based on the usage of a non-structural approach to measure the competition degree in the Ecuadorian Banking Sector. The empirical study is framed around the Panzar & Rosse (1987) model, and its recent developments and improvements, considering that it is well recognized in the academic community as a suitable fit for the banking sector. It is intended to test this assumption in the Ecuadorian banking reality. For the viability of the results of this thesis it is important to incorporate resent improvements in the methodology of the original model, to guaranty that the results won’t be biased in any possible way.

3.1. Competition in the Financial Sector

As many other markets, researches have been trying to evaluate correctly the levels of competition in the financial markets. But this specific industry holds many peculiarit ies compare to the traditional economic activities, it plays a crucial role on non-financ ia l sectors. This, solely, implies an extra interest on addressing competition issues in the whole financial industry. But in recent years, competition in this industry has attracted

(11)

much interest from researchers, given that after the 2008 crisis, the world’s gaze turned to the financial industry. It was now time to stablish strict regulations and controls for the different financial activities, with special focus on the banking sector.

A financial system is a set of institutions whose main objective is to channel the savings received from the society. This channeling of resources enables the development of economic activity (production and consumption) making funds arrive from people who have a liquidity surplus to people who need this liquidity. Credit financial intermediar ies are responsible for capturing the deposits from the public and, secondly, for lending these resources to those who are in need of them.

The assessment of competition in the banking sector has a broad development of methodologies, given the complexity of the industry, and a long tradition in literature. Analyzing the degree of concentration and competitiveness in this sector has important implications for the economy. As mentioned before, banks are in charge of receiving the private savings and placing them as loans and credits in other sectors of the economy. They are responsible of the allocation of capital, the growth of companies and their abilit y to undertake investment projects.

In traditional Industrial Organization studies, the main focus of research was based on finding a direct relationship between concentration and market power in the banking sector, following the Structure-Conduct-Performance paradigm, which stated that the probability of collusion increases with market concentration. This studies have been inconclusive and failed to find a direct relationship between these two conceptions. Nevertheless, it is clear that concentration is a relevant variable to consider in competit io n analysis, but is not the only one that explains the competitive behavior of this complex market. There are other variables that can influence the banking systems and their level of competition, like the political heritage of a country, the market contestabilit y, institutional and regulatory environment, economic cycle, among others (Zurita, 2014).

The reliability of studies involving traditional approaches in this sector have been questioned and raised some doubts. In response to the deficiencies, more complex models, based on newer approaches of Industrial Organization, have been developed and improved to measure the degree of competition directly from market information. There

(12)

is no consensus on which measure is the best for gauging competition levels, different indicators of banking market competition do not provide the same inference about competition (Leon, 2014). Therefore, results and conclusions are biased by the methodology or indicator used for the study. It is important, given the tradeoff between models, to use the one that best suits the data available for the banking sector analysis.

3.2. Competition Assessment: Structural vs. Non-Structural Approach

As mentioned before, the existent literature addresses the competition assessment problem with two types of approaches, a structural one and a non-structural. The structural methodology is based on the Structure-Conduct-Performance (SCP) paradigm, traditional Industrial Organizatio n, and the market efficiency hypothesis. The whole point of this approach is to establish if a market with high levels of concentration generates collusive behavior between the main participants of the industry, which in turn leads to higher profits; or if higher efficiency of the major market players generates higher profits and turnovers.

In the other hand, the non-structural approach follows the principles of the New Empirica l Industrial Organization (NEIO), determining the degree of competition directly from the market’s behavior. This new approach arises as a response to the theoretical and empirica l deficiencies detected in the studies based on the structural approach. Its main focus is to analyze the competition levels between market players and the usage of market power position by the leader firms. Market concentration is not part of the analysis in the models used under this approach.

3.2.1. Structural Approach

The Structure-Conduct-Performance (SCP) paradigm and the efficiency hypothesis are the core ideas, behind the structural approach, used to establish the market competit io n level and the incidence of concentration in competition. The SCP paradigm, initia ll y developed by Mason (1939) and Bain (1956), seeks to explain aspects of the conduct and performance of firms in terms of the structural characteristics of the markets in which they operate (Leon, 2014). The market structure depends on basic conditions of supply and demand, and it influence the conduct of the participant firms in the market.

(13)

It is assumed that a higher degree of concentration eases a collusive behavior between the major market participants and leads to a lower level of competition in the market. Under the SCP paradigm believes, it is considered that the market structure influences firms’ behavior, determining firms’ capacity to maximize revenue, the interaction between competitors, and the industry’s behavior. Market structure determines how companies perform, and this, in turn, limits the outcomes a firm can obtain. Resulting in the assumption that this is a unidirectional line of causation.

The SCP paradigm claims that firms’ prime instinct is to restrict output and increase prices through exploiting their market power position or through collusion with other companies. The resulting benefit from this type of behavior is given by the differe nce between the market price, unnaturally high, and the firms’ costs. From a welfare point of view, this higher than normal profits can only be the result of an optimization behavior at the expense of consumers, rather than a firm’s accomplishment.

The relationship between the size of the companies and the obtained outcomes is of great importance for the SCP paradigm, since biggest firms are the ones to control substantia l proportions of the market, giving them greater opportunities and incentives to get involved in collusive or monopolistic practices. The presence of collusive behavior hampers the market structure and reduces competition, altering the size of the participant firms. Companies’ motivation to expand and grow lies in increasing their degree of market power or, alternatively, preventing that other organizations reach a position of monopoly or increase their market power. Thereby, vertical integration is seen as a way to outspread monopoly power to other related industries. Likewise, advertising and product differentiation are considered as actions aimed to establish entry barriers and increase market power.

This approach considers the concentration level of an industry as a good proxy for market power and, therefore, for its undesirable effects in economic welfare. Usually, the methods used are: the k-firm concentration ratio, measuring the market share of the “k” principal firms in the market; and the Herfindahl-Hirschman Index (HHI), measuring the lack of competition in the industry, the higher the index the more concentrated and less competitive the market is. Finding a causal relationship would mean clear implicat io ns

(14)

for competition policy, but no conclusive evidence has been found between the degree of market concentration and the industry’s average rate of return.

This line of study has received many criticisms, especially in reference to the unidirectional causality between market structure, firms’ behavior and the outcomes firms obtain. As a result of the discrepancies generated from this type of studies, Demsetz (1973) and Peltzman (1977) developed the efficiency hypothesis.

The efficiency hypothesis criticizes the reasoning behind the SCP paradigm, offering an alternative explanation for the relationship between market structure and the outcomes obtained by the participant firms. The hypothesis considers that if a firm achieves a level of operational efficiency that is comparably higher than the one from its competitors, the firm’s maximizing behavior will lead to a higher market share at the expense of its competitors. Given that the cost structure is now more efficient, the firm will be able to reduce its prices below the minimum prices that their competitors can set. Therefore, the market structure is determined endogenously by the firms’ performance, in a way that market concentration is a result of the higher efficiency of the dominant firms.

Studies trying to empirically gauge market competition under these structural conceptions can be divided in two groups, depending on the variable use as a measurement of the firms’ outcomes. These variables can be: the price of certain products or services as a measure of the performance of the firms; and the profitability of the entities as a measure of its obtained outcomes.

In the case of competition in the banking sector, the price of a good or service as the proxy of performance usually involves the usage of a banking product like the average interest rate of loans or deposits, the average cost of certain banking services, or the interest rate of a specific product, such as business loans. Using this proxy demands an adequate market definition, based on a particular product, which is usually not possible in the banking sector; and ignores the fact that larger institutions, present in different market segments and geographical areas, can subsidize some activities within the corporate group. Consequently, the use of prices, as a measure of banks' behavior, can lead to incorrect inferences when applied to the outcomes obtained by the organizations in the market.

(15)

In the other hand, using profitability measures to estimate the performance of the banks provides a greater ease for calculations and access to necessary data, as well as the advantage of having summarized in only one piece of data the gains and losses results, from the different banking products an institution has. However, considering the efficiency hypothesis, there is the possibility that the exertion of market power, by the dominant firm in the market, leads to laxer standards of operational efficiency. This relaxation of standards would harm the profitability of said firm, once they have reached a dominant position. Hence, market power and profits need not always to be directly correlated with each other.

The structural measurements for market competition provide incomplete explanations for competition. They do not consider other significant factors with great impact on the behavior of firms and the performance they can achieve. Among these determinants of competition, the regulation of financial markets and the markets’ contestability stand out, as substantial issues to take into account when gauging competition levels.

3.2.2. Non-Structural Approach

After identifying some concerns in the structural approaches, three main non-structura l models, for assessing competition in the market, were developed under the New Empirical Industrial Organization (NEIO) theory of behavior. This meant a substantia l evolution in competition studies, for the banking sector and other industries, from the simpler methods that were used in the first contributions in this matter.

The new procedures for assessing competition in a market are due to the models developed by Iwata (1974), Bresnahan & Lau (1982) and Panzar & Rosse (1987). These models estimate the level of competition and emphasize the analysis of the competit ive behavior of firms, regardless of the structure of the market. All three models share as their main feature, that the measurement of the degree of competition is done by estimating the deviation of the price level with respect to its level in perfect competition.

The model Proposed by Iwata (1974) is focused on the development of an econometric approach to address the problem of price determination in oligopoly. Specifically, for the analysis of the Japanese flat glass industry during the period from 1956 to 1965. Iwata

(16)

proposed in his model an estimator of conjectural variation to test for the existence of certain type of collusion among oligopoly firms in the market. The model is based on the construction of this estimator, using a demand function, in terms of the industry’s products prices, and a cost function, in terms of prices of all the inputs used in the production cycle.

Regarding a banking sector application, the model proposed by Iwata has barely been used in empirical studies, there are only two identifiable examples. The first one is the work done by Shaffer & Di Salvo (1994), who applied the model in a duopoly banking market with data from Pennsylvania, United States. The results provide a concrete counterexample to the presumption that high structural concentration must preclude substantially competitive conduct (Castellanos, Del Ángel, & Garza-García, 2016).

The second application can be found in the analysis done by Shaffer & Spierdijk (2013) for a highly concentrated market with two banks in South Dakota, United States. They estimated the Lerner index, the conjectural variation parameter and the Panzar-Rosse H-statistic, concluding that it might be a competitive behavior, according to the H-H-statistic, in a noncompetitive market, as suggested by the Lerner index and the conjectural variation correlation test (Castellanos, Del Ángel, & Garza-García, 2016).

In order to measure the degree of market power of the average bank, Bresnahan & Lau (1982) developed a model of profit maximizing oligopoly banks. The model assumes that the banks produce only one product using multiple inputs. The cost functions are based on the price of these inputs. And the equilibrium price equation includes a markup, not used under perfect competition, partly used under monopolistic competition or oligopo ly and fully used under monopoly. Assuming that the required productive factors differ between banking products, there is not interdependence between the different products ’ cost functions, the model allows testing the possible use of market power for submarkets. Based on these equations a conjectural variation parameter is estimated to gauge the level of competition in the market.

The Bresnahan-Lau model has been applied empirically in some studies of banking competition, from which the following studies developed for the European Union, Brazil and the United States stand out. In the case of the European Union (EU), Bikker (2004)

(17)

analyzed the degree of competition on the deposit and loan markets of nine European countries. The hypothesis of perfect competition was rejected for the deposit markets of the whole EU, and for the loan markets of Germany, Portugal, Spain, Sweden and the UK.

The Brazilian banking sector analysis by Rocha, J. et al. (2009) focused on regiona l competition, using data from eight Brazilian states and a dynamic panel. On average, the level of competition in the Brazilian banking sector was found to be high, even though perfect competition can be rejected. This result prevails at the state level.

In the last case, Chang, S. et al. (2012) used a static and a dynamic version of the Bresnahan-Lau model to study market power in the United States commercial banking industry since the early 1990s, when the US government began deregulating this sector. The static model showed a high degree of competition in the industry. The dynamic model, suggested a level close to perfect competition in the short-run, but a certain degree of market power in the long-run due to the slow adjustment speed of the demand and supply.

Panzar & Rosse (1987) developed a model to assess competition in the banking sector, based on comparative statistical properties of the reduced form of the income equation of a bank. This model uses data at the institutional level and allows a precise estimation of the competition degree in the banking system. The model assumes that banks produce only one good, financial services, hence all inputs are used in the production of this one good. Consequently, this model does not allow to distinguish between different banking products or geographical regions as other models. Despite this, the Panzar-Rosse model has proved to be a useful tool, being broadly used in many competition assessment studies, for the banking sector as well as other industries. The estimations of the competitiveness of the industry in this model are measured by the H-statistic, parameter based on the price of the production inputs. A deeper discussion of this model is bared in the next section.

Given the great requirement of data that the Iwata and Bresnahan-Lau models have, they have had a low application in empirical studies. While, the Panzar-Rosse model, with low data requirement and easy access to it, has been widely accepted in the academic

(18)

community. Particularly for assessing competition in the banking sector. The Panzar-Rosse model has been improved by the academic community over the years, making it an even more powerful and accurate tool for competition analysis.

3.3. The Panzar-Rosse Model

Panzar & Rosse (1987) shaped a model for assessing the degree of competition in a specific market and, to estimate the competition structure of it. The different types of competition, according to the model specifications, can be: perfect competit io n, monopolistic competition, collusive behavior or monopoly. The model is designed to obtain an indicator, known as the H-statistic, that, under certain conditions, can be interpreted as an increasing and continuous measure of the degree of competition in a particular market in a given moment of time.

The methodology implemented in the Panzar-Rosse model is based on the fundame nta l assumption that firms’ response to changes in the inputs prices will be different depending on the competitive environment of the market in which they operate. In other words, in this model, the degree of competition exhibit in the market is measured by the effect that changes in the inputs prices have over the total earnings perceived by the firms in an equilibrium condition.

Starting from a generic bank 𝑗, the model assumes that the double condition of market equilibrium is sustained, both at the industry level as well as at the individual level, of each firm in the market.

Thus, equilibrium, in the market, is obtained when profit is zero, in other words, when revenues are equal to the costs and there are no extraordinary benefits:

𝑅_𝑗(𝑦_𝑗, 𝑍_𝑗𝑅) = 𝐶_𝑗(𝑦_𝑗, 𝑊_𝑗, 𝑍_𝑗𝐶)

where 𝑅_𝑗(∙) and 𝐶_𝑗(∙) are the revenue and cost functions of the bank 𝑗; 𝑦_𝑗 is the production of the bank 𝑗; 𝑊_𝑗 is the price vector of the 𝑘 production factors of the bank 𝑗, 𝑊_𝑗= (𝑊_1𝑗, ⋯ , 𝑊_𝑘𝑗 ); 𝑍_𝑗𝑅 is a vector of exogenous variables that impact the revenues of the bank 𝑗; 𝑍_𝑗𝐶 is a vector of exogenous variables that influence the costs of the bank 𝑗.

(19)

At the entity level, the equilibrium condition requires that marginal revenue of the bank 𝑖 be equal to its marginal costs:

𝑅′_𝑗(𝑦_𝑗, 𝑍_𝑗𝑅) = 𝐶′_𝑗(𝑦_𝑗, 𝑊_𝑗, 𝑍_𝑗𝐶)

From this initial conditions, the Panzar-Rosse model establishes the market competit io n level estimator, the H-statistic, as the elasticity of the bank 𝑗 total revenue with respect to changes in the prices of the 𝑘 production factors:

𝐻 = ∑ 𝜕𝑅𝑗 𝜕𝑊_𝑘𝑗∙ 𝑊_𝑘𝑗 𝑅_𝑗 𝐾 𝑘=1

According to the model, the H-statistic summarizes in a single figure the overall level of competition in the relevant market. It measures the strategic behavior that firms exhibit, as from their ability to transfer the variations in the input prices to the final production prices and to the output offered in the market. The H-statistic can take values going from minus infinity till one:

 𝐻_{𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐} ≤ 0 if the market faces a monopoly or collusive behavior  𝐻_{𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐} = 1 if the market faces perfect competition

 0 < 𝐻_{𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐} < 1 if the market faces monopolistic competition

The effects that the changes in the input prices have over the market equilibrium prices and output depends on the slopes that the supply and demand curves have. It depends on the competition context of the market. In the first scenario, when a market faces a monopoly or an oligopoly with collusive behavior, an increase in the price of the production factors will increase the marginal cost, reducing the market’s equilibr ium production level, therefore decreasing the total revenue. Given this, the H-statistic can´t be positive, if costs increase revenue decreases.

In the second setup, a market under perfect competition, a proportional increase in the input prices will cause an equiproportional increase in the marginal cost and in the average cost, due to the cost function first order homogeneity in the input prices. In order

(20)

for the firms to keep their market operations ongoing, they would have to rise prices until the increase in costs is compensated. Firms residual demand is higher, due to the foreclosure of the inefficient firms, such that the increase in costs is translated directly into higher prices. Therefore, the total revenue remains constant, there is no monetary benefits. Consequently, under perfect competition, the H-statistic will be positive and equal to one. This result depends entirely on the market being in long-run equilibr ium, where the market has already been adjusted after the entrance and exit of firms reaching a competitive equilibrium.

The last scenario is the case of a market with monopolis tic competition, where firms are in long-run equilibrium. In this situation, even though they act as benefit maximizi ng individuals, the possibility of free exit and entrance of other firms in to the market guarantees that the benefits at the aggregate level are equal to zero. Therefore, even if firms act as monopolists they may not be able to exploit a market power condition. The H-statistic will be equal to or less than one. As the perceived elasticity of demand approaches a value closer to one, this case will seem increasingly the case of perfect competition, due to a higher entrance of new firms into the market to satisfy the demand . On the other hand, as the demand gets more inelastic, this case will resemble the monopoly case. Hence, for this scenario the H-statistic will have a value between zero and one.

Some characteristics of the Panzar-Rosse model have raised certain criticism. On the one side, we have a static model that measures the degree of competition in a given period of time, making it necessary, to estimate the results of the model, the use of observations from markets in a long-run equilibrium situation. The practical difficulty in estimat ing the long-run equilibrium revenues and costs for all the firms in the market, represents a significant criticism of this model. In addition, the fact that in the banking system, the entrance and exit of firms occur, leads to the conclusion that in practice there is not a long-run equilibrium situation.

On the other side, the Panzar-Rosse model is to be applied exclusively to firms that produce only one output or product. As mentioned before, in this model banks are considered to be producer of a single good, financial intermediation services. And it is obtained through the usage of specific inputs (labor, physical capital and financ ia l

(21)

capital). Nonetheless, to establish the degree of competition in a market, it would be necessary to consider the totality of the activities undertaken by the firms in the industr y. Given the lack of data to do so, the model has to be apply considering the whole production as one single product, rather than discerning between submarkets.

Ultimately, the model requires the cost functions of all banks to be homogeneous, as well as for the market’s price elasticity of demand to be greater than one. In practice this may not be true. Other requirement is that the market has to be in long-run equilibrium, as shown in Nathan & Neave (1989) it is important to test the model observations to see if this condition holds. To verify the condition of long-run equilibrium is necessary to estimate the equilibrium E-statistic and perform a t-test to find if it is equal to zero. The E-statistic follows from a revenue specification using the Return on Equity (ROE) or the Return on Assets (ROA) as the dependent variable, it is equal to:

𝐸 = ∑𝜕𝑅𝑂𝐸𝑗 𝜕𝑊𝑘𝑗 ∙ 𝑊𝑘𝑗 𝑅𝑂𝐸𝑗 𝐾 𝑘 =1

In general terms, this model has been considered as a valuable tool for gauging the level of competition in a market. Firms’ revenues and costs are usually known in a banking system, making the data required for the calculations available under regular basis. This is why the academic community has been working, over the years, on improving and solving some of the concerns raised around the Panzar-Rosse model. One of the main weaknesses of the model is that it doesn’t tell much about the sources of imperfec t competition, hence is difficult to know what might be done to change matters.

3.4. Applications and Improvements to the Model

The Panzar-Rosse methodology has been used in many studies as the optimal measure of an industry’s competition level, especially for the banking sector and the financ ia l industry in general. One of the first applications in an empirical analysis is the one done by Nathan & Neave (1989) using an annual cross-sectional sample for the Canadian financial system between 1982-1984. An H-statistic is estimated in each year for a separate group of banks, credit companies and mortgage companies. It was found that in 1982 data is coherent with a perfect competition behavior, while between 1983-1984 the

(22)

hypothesis for monopoly and perfect competition are rejected, meaning that there is evidence of monopolistic competition.

In Vesala (1995) a competition study for the Finnish banking sector is performed, using annual cross-sectional data. For the analysis, structural and non-structural methodologies are used, finding a high resemblance between the results obtained with the use of the different methodologies. The Panzar-Rosse revenue test shows that in the periods between 1985-1988 and 1991-1992 there is evidence of a behavior resembling monopolistic competition. While, in the period between 1989-1990 the results show evidence of a perfect competition behavior.

Later on, more complete studies were developed, trying to explain the degree of competition in the banking sector, using the Panzar-Rosse model, for a broader diversit y of countries. In Claessens & Laeven (2004) a study for 50 developed and emergent economies, between the period of 1994-2001, was undertaken. The H-statistic shows that monopolistic competition is the best way to describe the behavior of these countries banking sector.

In Bikker, Spierdijk, & Finnie (2006a) the Panzar-Rosse model is used for the estimat io n of the H-statistic in 101 emergent and developed countries from the five different continents, in a period between 1986-2004. Using data for over 25,000 banks, the average value obtained for the H-statistic of the 101 countries analyzed was approximately 0.50; but with a clear variation between countries. The 101 economies were classified in three groups according to the degree of competition that the H-statistic reported for each one of them: low, medium and high levels. It can be said that in most of the cases the competitive environment in the banking sector is not determined by perfect competition.

In Delis et al. (2008), using econometric technics for panel data analysis, the H-statistic is estimated, as well as the conjectural variation parameter from the Bresnahan- La u methodology. The study uses panel data for banks in Greece, Latvia and Spain, in a period between 1993-2004. For this analysis a dynamic estimation of the H-statistic is proposed, and compare with the static estimation. The importance of using a dynamic estimator lies on the statistical significance of controlling for short-term variations in the data. The formulation solves the problem of inference using non-stationary data. It was found that

(23)

the static estimation understates the market power degree compare to its dynamic counterpart.

Bikker et al. (2012) in an empiric analysis performed to a panel data of banks in 67 countries, show that a price equation specified through a scaled revenue function or other specifications that control for firms’ size are not adequate for measuring the competitive behavior of an industry. This means that in a specification where the dependent variable is defined as a ratio of a size variable, like total assets in many studies, or if a size variable is used as a regressor, the yielded results will be biased. If we include a control variable for firm scale, like the logarithm of total assets, the estimated regression keeps the quantity produced statistically constant. Which means that the coefficients that add up to be the H-statistic, will represent the response of total revenue to changes in the input prices, for a fixed production scale. In other words, the change in prices when the production remains constant. In this case, the H-statistic will be positive for all monopolies or oligopolies whose total revenue equation is controlled for firm scale.

An appropriate H-statistic must be derived from a non-escalated revenue equation, which means, in many cases, that additional information, regarding costs, market equilibr ium and the market’s elasticity of demand, would be required. Then, testing the H-statistic would imply a one-tailed test, where a positive H-statistic is inconsistent with imperfec t competition, but a negative value may be consistent with perfect competition, if the market is in structural imbalance.

A problem arises when firms face imperfect competition, which, according to the Panzar -Rosse model refers to the case where de H-statistic is negative. As mentioned in Bikker et al. (2012), if a monopolist faces a perfectly inelastic demand curve, there won’t be any production adjustments if there are changes in the input prices, thus total revenue will move in the same direction as the output price, which, in equilibrium, has the same direction as the marginal cost. Hence, total revenue will move in the same direction as the input prices, therefore, a positive H-statistic will be obtained, as long as the demand is inelastic. The profit maximization condition, stating that ‘marginal revenue equals marginal cost (MR=MC), rules out this situation, implying, as well, an elastic demand at equilibrium output levels.

(24)

Considering the results obtained in Bikker et al. (2012), in Sun (2011) the Panzar-Rosse methodology is used for a panel of banks from the European Union, the United States and the United Kingdom. For the competition analysis in this paper, the estimations of the H-statistic are undertaken using specifications with and without a control for firms’ size in the regression equations. As a result, it was found that the estimations obtained do not differ, with each other, in a significant way.

In Goddard & Wilson (2009), the arguments surrounding the escalated revenue funct io ns and the controls for firm size, gain a more significant empirical robustness. The analys is demonstrates, through the use of various simulations, that a poor specification of the revenue equation, either by scaling the dependent variable or by the inclusion of total assets as a control variable, represents a form of misspecification bias, affecting the estimated H-statistic. In addition, it is proved that a static panel estimation for fixed effects in a context of partial adjustment and not instantaneous, leads to an H-statistic biased towards zero. The correct method would be to obtain the estimations using dynamic panels.

3.5. Reinterpretation of the H -Statistic

Following the work done in Ravizza (2012) and the statements from Bikker et al. (2012), the Panzar-Rosse test is presented as a one-tailed test, prone to additional consideratio ns : a positive H-statistic is inconsistent with any form of imperfect competition, but a negative value may arise from various conditions. Such like, short-run competition or long-run competition with constant average cost. The short-run competition occurs when the market has not yet faced the outflow and inflow of new companies, as a result of demand or cost shocks.

Considering the estimation of the unscaled revenue function, the estimated values of the H-statistic can be reinterpreted as follows:

 𝐻_{𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐} = 1: the market faces a long-term competitive equilibrium, a contestable natural monopoly or, profit maximizing firms with restrictions on their equilibrium point.

 0 < 𝐻_{𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐} < 1: the market is not compatible with imperfect competition, whether it is or not in equilibrium. If the null hypothesis of the H-statistic being

(25)

negative is rejected, without a scale control, the existence of a monopoly, a cartel or any other collusive behavior is disregarded.

 𝐻_{𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐} ≤ 0: this case requires additional information to make proper conclusions, given that it won’t distinguish between perfect and imperfect competition. Rejecting the hypothesis of an H-statistic equal to one does not imply disregarding perfect competition. Information regarding the long-run structural equilibrium of the market allows for a much deeper analysis of the case, obtaining clear conclusions.

For this last scenario it is of enormous relevance performing a long-run equilibrium test on the regression, using the ROE as a dependent variable. The resulting estimation is the E-statistic, used as a joint test for the presence of a competitive behavior and a long- run structural equilibrium. Not rejecting the null of the E-statistic being negative implies that the market is, either facing a monopoly or oligopoly, or it is in a short-run equilibrium.

Testing the H-statistic together with the E-statistic has the following interpretations:  𝐻_{𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐} = 1 and 𝐸𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐 = 0: the market faces perfect competition with

long-run equilibrium.

 𝐻_{𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐} < 0 and 𝐸𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐 < 0: the market faces a monopoly or collusive

behavior, or it is in a short-run competitive equilibrium.

 𝐻_{𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐} > 0 and 𝐸𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐 < 0: the market faces monopolistic competition or

is in transition to a long-run competitive equilibrium.

It is important to notice that, after the reinterpretatio ns, the competition implicat io ns under the monopoly, oligopoly or monopolistic competition models are still valid. The statistical precision of this results have been empirically and theoretically proven in Goddard & Wilson (2009) and in Bikker et al. (2012). To do so, multiple simulations and demonstrations were performed in these analysis.

3.5.1. Oligopoly Interpretation with the H-Statistic

Acording to Panzar & Rosse (1987) and Vesala (1995) an 𝐻_{𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐} > 0 is possible in a conventional static oligopoly. Shaffer & Spierdijk (2011) prove that this argument is not the only true statement, an oligopoly can occur under certain scenarios. Specifically, they

(26)

prove that the H-Statistic can take either sign (positive or negative) for a Stackelberg duopoly facing linear cost and demand functions. And an 𝐻_{𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐} > 1 is possible for a low-cost firm in a homogeneous Cournot duopoly, with asymmetric costs and a linear demand function.

In the first case, Shaffer & Spierdijk (2011) claim that given that the marginal cost is homogeneous of first degree in all input prices, the H-Statistic will be equal to the elasticity, formed by the derivative of the Total Revenue with respect to the Margina l Cost (𝜕𝑇𝑅 𝜕𝑀𝐶⁄ ). They consider a price function of the type 𝑃 = 𝑎 − 𝑏𝑄, with 𝑄 = 𝑞₁+ 𝑞₂, and a marginal cost function equal to 𝑐 (𝑀𝐶 = 𝑐). Solving for 𝜕𝑇𝑅 𝜕𝑀𝐶⁄ , the result obtained for the H-Statistic can take either sign for Stackelberg duopolists, depending on the relative magnitudes of 𝑎 and 𝑐. Both firms will have the same sign for the H-statistic, but the leader’s value of the statistic will be twice as far from zero as the one of the follower.

In the second case, Shaffer & Spierdijk (2011) set a price function of the type 𝑃 = 𝑎 − 𝑏𝑥 − 𝑏𝑦, where 𝑥 is the output quantity of one firm and 𝑦 is the output of the other one. The cost functions are: 𝑇𝐶_𝑥 = 𝑐𝑥 and 𝑇𝐶_𝑦 = 𝛼𝑐𝑦. It is assumed that 𝑎 > 𝑐 to ensure non-negative profits in equilibrium, and 𝛼 > 1 so the first firm has lower marginal costs. Solving by standard first-order conditions for profit maximization, it is obtained: 𝐻_𝑥 = (𝑐/𝑇𝑅_𝑥) and 𝜕𝑇𝑅_𝑥 /𝜕𝑐 = [𝑎𝑐(𝛼 − 1) + 2𝑐2_(𝛼2_{− 𝛼 − 1)]/[𝑎}2_{+ 𝑎𝑐(𝛼 − 1) +}

𝑐2(𝛼2 − 𝛼 − 1)]. Given that 𝑎 > 𝑐 and 𝛼 > 1, 𝑎2 > 𝑐2(1 + 𝛼 − 𝛼2) and so 𝐻_𝑥 > 1.

With similar calculations it can be shown that 𝐻_𝑥 > 0 for all 𝛼 > {2𝑐 − 𝑎 + [(2𝑐 + 𝑎)2+ 16𝑐2]1⁄2_{} 4𝑐}⁄ _{. Shaffer & Spierdijk (2011) point out that this is}

another example of how it can be obtained 𝐻 > 0 with imperfect competition.

3.6. Long-Term Equilibrium in a dynamic analysis

The long-term equilibrium assumption determines the results from the H-statistic under a perfect competition scenario, but not in the monopoly case. As pointed out in Ravizza (2012), this can wrong conclusions obtained from the interpretations of the degree of competition in the industry, nonetheless, econometric progress in dynamic estimat io ns allows to dissipate the problems generated by the static disequilibria.

(27)

Goddard & Wilson (2009) identify the effects for the H-statistic, of the bias generated by the misspecification of the revenue equation. The misspecification bias arises when the market adjustment towards equilibrium, as a response to shocks in the input prices, is partial and not instantaneous. To take into account the partial adjustment in the model, it is required to include a lag of the dependent variable in the revenue estimation, turning the model into a dynamic one. This suggests that the static version of the model, commonly used in many studies, without the addition of a lag variable, is specified incorrectly and produce biased estimations. Using a static estimator to predict the H-statistic from a revenue equation, with or without a lags, will produce an estimation biased towards zero.

Using an appropriate dynamic panel estimator in a dynamic revenue equation, correctly specified, allows for an accurate estimation, solving the bias problem. The H-statistic will be virtually unbiased. Ravizza (2012) mentions that the estimation obtained from dynamic panels allows for direct analysis of the rate at which the market adjusts towards equilibrium. For this porpoise, the estimated coefficient of the lagged dependent variable is used. With this method it is no longer necessary to test for the long-run equilibr ium hypothesis, which means that is also not necessary to estimate the revenue equation with the ROE as dependent variable. The instantaneous adjustment is considered as an especial case of this condition.

Bearing this in mind, if the model is well specified it won’t be necessary the assumptio n of market equilibrium to obtain an accurate estimation of the H-statistic. An empiric model, correctly specified, will lead to an unbiased H-statistic estimation. The inconsistency problems will be solved under the conditions of the instantaneo us adjustment or the partial adjustment.

4. Model Specification and Empirical Strategy

One of the purpose of this study is to establish an adequate methodology, able to measure the competition degree in the banking industry. For this matter, the Panzar-Rosse model is used as a starting point, followed by the advances and improvements found in recent studies, like in Goddard & Wilson (2009) and in Bikker et al. (2012). With the strategies

(28)

pursued in this analysis, an optimal application of the Panzar-Rosse model is expected for the Ecuadorian banking system.

Following the methodology and the strategies used in Panzar & Rosse (1987), the banks are considered as producers of a solely good, being this ‘financial intermediat io n services.’ This is assumed to be a homogeneous product for the final consumers, with not much price sensitivity. The firms differ with each other on the service and the personal experience they grant the consumer of their product, but the basic essence of the ‘financ ia l intermediation services’ is pretty much the same for all the banks, especially in terms of price.

4.1. Model Specification: The H-Statistic Estimation

The H-statistic is used as an indicative parameter for the type of competition found in an industry, in certain time periods. For a specific firm, the equilibrium revenue is given by the equilibrium production quantity times the equilibrium price. These variables are not easily observed on individual basis, unlike the total revenue that can be extracted directly from the firm’s financial statements. Both, price and quantity, depend on costs, demand and the competition context in which the firm is operating, consequently, the cost and demand shifters should be included as control factors in the revenue function. While the competitive structure in which the market operates will affect the values that the test parameter can take.

From the revenue specification equation and following the nomenclature used in Panzar & Rosse (1987), and the empirical strategy pursued by Claessen & Laeven (2004), for a bank 𝑗 in a period 𝑡, the total revenue is defined as:

𝑅_𝑗𝑡 = 𝑓(𝑊_𝑗𝑡, 𝑍_𝑗𝑡, 𝑦_𝑗𝑡, 𝑒_𝑡)

where the total revenue 𝑅_𝑗𝑡 is explained as function of different exogenous variables, which are the 𝑘 input prices 𝑊_𝑗𝑡 and the 𝑠 control factors that affect total revenue 𝐶𝐹_𝑗𝑡= (𝑍_𝑗𝑡, 𝑦_𝑗𝑡), either as demand shifters or as cost shifters.

(29)

In order to add in the model, the reinterpretations from Bikker et al. (2012) and the empirical safeguards from Goddard & Wilson (2009), dynamic panel estimations will be performed, with and without scale controls, to compare the obtained H-statistics. The equation to regress, then, follows a dynamic specification, including a lag of the dependent variable and lags of the proxy variables used as input prices. The dynamic revenue equation, in its reduce form, following the specification in Besar (2011) and Ravizza (2012), should state as follows:

𝑙𝑛(𝑅_𝑗,𝑡) = 𝛼₀+ 𝛼₁𝑙𝑛(𝑅_{𝑗,𝑡−1}) + ∑ 𝛽_𝑖𝑙𝑛(𝑊_𝑗,𝑡𝑖 ) 𝑘 𝑖=1 + ∑ 𝛽_𝑙𝑖𝑙𝑛(𝑊_{𝑗,𝑡 −1}𝑖 ) 𝑘 𝑖=1 + ∑ 𝛾_𝑠𝑙𝑛(𝐶𝐹_𝑗,𝑡𝑆) 𝑠 + 𝛿𝐷 + 𝜂𝑗+ 𝜀𝑗,𝑡

where 𝛼₀ is the constant, 𝛼₁ is the lagged dependent variable coefficient, the 𝛽_𝑖 and 𝛽_𝑙𝑖 are the vectors of coefficients for the 𝑘 input prices, cotemporary and lagged respective ly, and the 𝛾_𝑠 is the vector of the coefficient for the control factors used in the model. Also, 𝐷 is a vector of monthly dummy, 𝜂𝑗 is the unobserved time-invariant individual effect for

the bank 𝑗, and 𝜀_𝑗,𝑡 is the error term. It is assumed that 𝐸(𝜀_𝑗,𝑡|𝑊_𝑗,𝑡𝑖 _{, 𝐶𝐹}

𝑗,𝑡𝑆, 𝜂𝑗) = 0.

According to Ravizza (2012) the logarithm specification in the regression reduces the simultaneity bias. Besar (2011) mentions that the lagged value of the revenue variable is included on the right hand side to capture persistence in total revenue and also potentiall y mean-reverting dynamics in total revenue.

The dynamic H-statistic is explained as the sum of the elasticity of the total revenue wit h respect to each of the input prices. The short-run H-statistic is based on the coefficie nts of the contemporary inputs, and is defined as follows:

𝐻 =∑ 𝛽𝑖

𝑘 𝑖 =1

1 − 𝛼1

On the other hand, the long-run E-statistic, is estimated from the same equation but specifying it with the ROA as the dependent variable.

Competition in the banking sector, real or not? : a non-structural approach of the ecuadorian experience