ANALYSIS OF DRIVERS OF INATTENTION IN CONSUMER CREDIT BASED ON EVIDENCE FROM OVERDRAFTS

(1)

UNIVERSITY OF LJUBLJANA

SCHOOL OF ECONOMICS AND BUSINESS

MASTER’S THESIS

ANALYSIS OF DRIVERS OF INATTENTION IN CONSUMER CREDIT BASED ON EVIDENCE FROM OVERDRAFTS

Ljubljana, January 2021 ŽIVA PERFETA

(2)

AUTHO RSHI P ST ATEMENT

The undersigned Živa Perfeta, a student at the University of Ljubljana, Faculty of Economics, (hereafter: SEB LU), author of this written final work of studies with the title Analysis of drivers of inattention in consumer credit based on evidence from overdrafts prepared under supervision of prof. Matej Marinč, PhD

D E C L A R E

1. this written final work of studies to be based on the results of my own research;

2. the printed form of this written final work of studies to be identical to its electronic form;

3. the text of this written final work of studies to be language-edited and technically in adherence with the FELU’s Technical Guidelines for Written Works, which means that I cited and / or quoted works and opinions of other authors in this written final work of studies in accordance with the FELU’s Technical Guidelines for Written Works;

4. to be aware of the fact that plagiarism (in written or graphical form) is a criminal offence and can be prosecuted in accordance with the Criminal Code of the Republic of Slovenia;

5. to be aware of the consequences a proven plagiarism charge based on the this written final work could have for my status at the FELU in accordance with the relevant FELU Rules;

6. to have obtained all the necessary permits to use the data and works of other authors which are (in written or graphical form) referred to in this written final work of studies and to have clearly marked them;

7. to have acted in accordance with ethical principles during the preparation of this written final work of studies and to have, where necessary, obtained permission of the Ethics Committee;

8. my consent to use the electronic form of this written final work of studies for the detection of content similarity with other written works, using similarity detection software that is connected with the FELU Study Information System;

9. to transfer to the University of Ljubljana free of charge, non-exclusively, geographically and time-wise unlimited the right of saving this written final work of studies in the electronic form, the right of its reproduction, as well as the right of making this written final work of studies available to the public on the World Wide Web via the Repository of the University of Ljubljana;

10. my consent to publication of my personal data that are included in this written final work of studies and in this declaration, when this written final work of studies is published.

Ljubljana, ________________________ Author’s signature: _________________________

(Month in words / Day / Year, e. g. June 1^st, 2012

(3)

LIST OF FIGURES

Figure 1: Overdrafts in relation to personal current accounts and consumer loans ... 4

Figure 2: The value of total overdraft market in years 2010-2019, the Eurozone (in mio EUR) ... 7

Figure 3: The value of overdraft market by country in years 2010 – 2019 (in mio EUR) ... 8

Figure 4: Hofstede’s six categories ... 13

Figure 5: Heterogeneity across selected EMU countries ... 27

LIST OF TABLES

Table 1: Variable definitions ... 25

Table 2: Clusters of cultures... 28

Table 3: Summary statistics for basic sample ... 28

Table 4: Fixed effects regression table ... 31

Table 5: Random effects estimates ... 32

Table 6: Hausman test ... 32

Table 7: Fixed effects estimates with cultural dimensions ... 33

Table 8: Random effects estimates ... 34

Table 9: Hausman test ... 35

Table 10: Fixed effects estimates ... 36

(5)

LIST OF APPENDICES

Appendix 1: Povzetek (Summary in Slovene language) ... 1 Appendix 2: Complete database used in the analysis ... 4 Appendix 3: R-Studio commands used in estimation ... 8

LIST OF ABBREVIATIONS

ATM – (sl. Bankomat) Automated Teller Machine

ECB – (sl. Evropska centralna banka); European Central Bank

EMU – (sl. Ekonomska in monetarna unija) Economic and Monetary Union GDP – (sl. Bruto domači proizvod) Gross Domestic Product

IMF – (sl. Mednarodni monetarni sklad) International Monetary Fund IVC – (sl. Individualizem / Kolektivizem); Colecivism vs. Individualism

LVS – (sl. Dolgoročna usmerjenost / Kratkoročna usmerjenost); Short-term vs. Long term orientation

MVF – (sl. Ženskost / moškost); Femininity vs. Masculinity PDI – (sl. Razlika moči); Power distance

RVI – (sl. Nebrzdanost / zadržanost); Resistant vs. Indulgence UAI – (sl. Izogibanje negotovosti); Uncertanty avoidance index US – (sl. Združene države) United States

(6)

(7)

INTRODUCTION

This master thesis investigates overdrafts and the drivers that affect consumers when entering into overdrafts. My aim is to provide some insights for understanding consumer attitudes towards overdrafts, especially determinants which facilitate overdraft behaviour, namely with the evidence from cross-cultural differences as measured by the Hofstede's model of cultural dimensions (Hofstede, 2020). People are involved in their culture to such extent that they often do not notice how it affects their behavioural model and economic thinking – they become inattentive.

Little is known about the drivers of overdrafts in general or the relevance of cultural and behavioural characteristics when entering into overdrafts and limited consumer attention is often put forth as an explanation. In the context of inattention, overdrafts have been studied by Stango and Zinman (2014) via extensive surveys of micro level data. They analysed data from thousands of individuals’ checking accounts for the period of up to three years. With this study, Stango and Zinman (2014) proposed a quite general definition of limited attention, playing a role in overdrafts: “It means incomplete consideration of information that would inform choices, whether that information is about account terms or available balances.” Their definition resembles the view of inattention in overdrafts by Grubb (2012), who sees it as limited awareness of consumers about their own past account usage. In both views, the consumer with limited attention is uncertain about the consumption that might overdraw the account. This master thesis can be classified under the growing domain of behavioural inattention in economics and finance. Behavioural inattention is a large and heterogeneous field of study which enters many domains of human behaviour. See Gabaix (2019) for an overview of the behavioural inattention with some of the main applications of behavioural inattention in economics, and DellaVigna (2009) for an overview of the literature about inattention in finance.

An overdraft enables the account holders to continue with their consumption even when their bank account has no funds in it or has insufficient funds to cover the amount of the consumption. When that occurs, the bank allows the overdraft of the account, ignoring the insufficient funds on it to cover this amount, which leads to a negative balance in the account.

The (borrowed) overdrawn amount is not pre-defined and there is no specified deadline when the amount needs to be repaid. In an overview of the US overdraft market, Zernik (2018) explains that the US banks composed 32 billion dollars in overdraft fees in 2013, which is approximately 60-75% of all the banks’ revenues from consumers.

According to the British Bankers Association, there were 8 billion pounds outstanding on overdrafts in large retail banks as of July 2013. Overdraft debt doubled between 1997 and 2007, although it was stable until 2010, but started falling afterwards (FCA, 2015). Reed von Hinten, Mendys-Kamphorst and Jannsen (2014) estimate that on the Dutch overdraft market between mid-2012 and mid-2013, approximately one third of current accounts were overdrawn at one point.

(8)

The master thesis is focused on the empirical analysis of aggregate values of overdrafts in the EMU countries by using Hofstede’s cultural dimensions in order to test whether and to what extent the measures of cultural dimensions influence the differences between countries in the levels of overdrafts. These dimensions are (i) power distance, (ii) individualism vs.

collectivism, (iii) masculinity vs. femininity, (iv) uncertainty avoidance, (v) long vs. short- term orientation and (vi) indulgence vs. restraint.

For this aim, the panel data regression analysis is implemented on the panel data for the EMU countries in a way that aggregate values of overdrafts are regressed on macro- economic variables and on cultural dimensions by Hofstede (2020). The data were obtained from the International Monetary Fund – IMF (GDP, unemployment, etc.), European Central Bank – ECB (overdrafts in the EMU countries), and from Hofstede (2020).

The obvious factors impacting overdrafts are macro-economic variables. Despite them, we assume there are some additional effects which significantly affect the decisions on making overdrafts. My first step of the analysis is to answer the question about whether macro- economic variables sufficiently explain the differences in overdrafts between the EMU countries. In particular, I will focus on three macro-economic variables: gross domestic product in current prices, inflation rate and unemployment rate. After checking the existence of country specific fixed effects in the model of overdrafts that are not related to macro- economic variables, the remaining variability will be checked whether it could be explained due to cultural differences. The following hypotheses will be tested:

• H1: On average, the power distance score will tend to have a negative impact on overdrafts.

• H2: On average, individualism will tend to have a positive impact on overdrafts.

• H3: On average, masculinity will have a negative impact on overdrafts.

• H4: On average, the uncertainty avoidance index will tend to have a negative impact on overdrafts.

• H5: On average, long-term orientation will tend to have a negative impact on overdrafts.

• H6: On average, indulgence will tend to have a positive impact on overdrafts.

The master thesis is divided into chapters, where the first chapter presents a general overview of consumer credit overdrafts and discusses certain backgrounds of different types and functions of overdraft lending with providers and the use of overdraft lending. At the end, the chapter brings an overview of the overdraft market with some concrete market data to better understand the wide use of overdraft lending. The second chapter turns to the topic of behavioural inattention, starting with a brief overview. It continues with evidence pointing to the important role of behavioural inattention in finance. The discussion proceeds with the drivers of overdrafts, where limited consumer attention is often put forth as a possible explanation. The chapter brings forth some insights to understand consumers’ attitudes toward overdrafts, especially cultural determinants facilitating overdraft behaviour. It also brings a brief introduction into the Hofstede's model of aggregate cultural dimensions. The

(9)

last three chapters cover empirical analysis of the master thesis. In particular, the third chapter gives the theoretical framework of panel data regression and shows a brief explanation of the model with a suitable econometric methodology. The fourth chapter explains the variables of the empirical model which were used to test the main hypotheses of the master thesis in the fifth chapter. Chapter six concludes with the results and main findings.

1 CONSUMER CREDIT OVERDRAFTS

This chapter is a general overview of consumer credit overdrafts and begins with a description of overdrafts definition. Once the latter is in place, the chapter discusses some background information on different types and functions of overdraft lending with providers and the use of overdraft lending. The chapter concludes with an overview of the overdraft market, which surveys some concrete market data about overdrafts, with the aim to better understand the wider use of overdraft lending.

1.1 Basic definition

Throughout this work, we assume the general definition of overdraft as pointed out by Reed von Hinten, Mendys-Kamphorst and Jannsen (2014), who describe overdraft as a negative balance on the current account. In case the current account’s balance falls below zero, the account will be overdrawn, and the account holder will have to pay interest on the amount overdrawn – the overdraft.

Similarly, Zernik (2018) defines overdraft as an open line of credit, linked to a checking account. An overdraft occurs when an individual’s account is debited, even when the account has no funds in it or has insufficient funds to cover a given charge, which leads to a negative balance in the account. The (borrowed) overdrawn amount and the deadline to for the amount to be repaid is not pre-defined. This resembles a revolving loan due to the unspecified length enabling the account holder to draw money at any time, without the need to reapply for the right to overdraw the account once the funds have been repaid. The consumer can overdraw his/her account until the credit limit is reached.

In general, any account charges can result in an overdraft: money withdrawal at the ATM, debit/credit card payment, etc. When an account is overdrawn, the bank can approve or decline the transaction according to the overdraft policy in the individual’s current account agreement. Normally, banks will cover transactions up to the overdraft limit, entering the account into an overdraft, and reject transactions over the limit.

1.2 Types of consumer credit overdrafts

Zernik (2018) classifies overdrafts into approved and unapproved. Approved overdrafts

(10)

include a credit limit and a rate, generally defined in relation to individual’s creditworthiness.

On the other hand, unapproved overdrafts are transactions which surpass the approved credit limit or transactions that have no approved overdraft facility. The banks choose to either approve or deny these transactions at their own choice.

According to Reed von Hinten, Mendys-Kamphorst and Jannsen (2014), an overdraft is authorised if there is a previous arrangement between the account holder and the bank which determines, for instance, the overdraft limit and the allowed amount of overdraft. On the other hand, an overdraft is unauthorised when there is a negative balance on the account without any prior arrangement regarding the overdraft use between the account holder and the bank or if the balance falls below the authorised overdraft limit.

1.3 Relevant market

In defining the relevant market, Reed von Hinten, Mendys-Kamphorst and Jannsen (2014) analyse whether overdraft ending should be considered as being a part of a group of services which together combine personal current accounts, or it should be considered as a part of consumer's loans. This elementary relationship between them is shown in Figure 1.

Figure 1: Overdrafts in relation to personal current accounts and consumer loans

Source: Reed von Hinten, Mendys-Kamphorst and Jannsen (2014).

When thinking about the relevant market, current accounts seem to be an obvious candidate as overdraft lending is not possible without having access to a current account, however, the authors (Reed von Hinten, Mendys-Kamphorst & Jannsen, 2014) find evidence that overdraft lending resembles characteristics of consumer loans. When discussing overdraft lending as a part of the market for consumer loans, authors discuss substitution between overdrafts and stand-alone consumer loans by arguing that when a consumer intentionally overdraws, it might also be possible to consider taking out a consumer loan rather than overdrawing the account.

Ttheir analysis suggests that overdraft lending partly belongs to the market for current accounts and partially to the market for consumer loans.

1.3.1 Overdraft lending as a component of personal current accounts

In the first case, one assumes that overdraft lending is not a product that is available

(11)

separately, therefore, consumers must initially hold a current account at a bank before they can have access to an overdraft facility. Current account is a product consisting of various services. Therefore, overdraft lending does not have its separate price and is determined jointly with the prices of all the services within the package (current account).

1.3.2 Overdraft lending as a part of market for consumer loans

When Reed von Hinten, Mendys-Kamphorst and Jannsen (2014) analyse overdraft lending as a part of the market for consumer loans, they discuss to what extent other consumer loans can be an appropriate substitute for overdraft lending. Possible suitable substitutes for overdrafts are term loans, revolving loans, credit cards and short-term loans (i.e., loans with a maturity of up to four weeks).

Reed von Hinten, Mendys-Kamphorst and Jannsen (2014) explain that in order to find appropriate substitutes, it is not necessary that all or even the majority of consumers consider them as such products. Moreover, as further claimed by the authors, consumers do not need to replace their entire overdraft with consumer loans, in fact, it is sufficient to have a few consumers willing to replace a part of their overdraft.

1.4 Varying uses of consumer credit overdrafts

Reed von Hinten, Mendys-Kamphorst and Jannsen (2014) classify overdraft usage into two main categories:

a) Overdrafts used as a source of liquidity b) Overdrafts used as a loan

Overdrafts used as a source of liquidity. The first category describes overdrafts which are used to balance instant liquidity needs when consumers have insufficient funds in their current account, irrespectively of the availability of funds elsewhere, i.e., as on the savings accounts. Such use of overdrafts usually occurs due to consumers' lack of accurate monitoring of their financial flows and consequently their funds fail to meet their payment requirements. These types of overdrafts are likely of small amounts and have a duration of a few days. When consumers use overdraft lending as a source of a very short-term liquidity, they most likely do not consider it as borrowing.

Of all the alternatives to overdrafts used as a source of liquidity, it seems that payments by credit cards could be the closest substitute. Reed von Hinten, Mendys-Kamphorst and Jannsen (2014) point out the following two main characteristics of credit cards as substitutes to overdrafts: (i) consumers do not have to pay interest on their credit, but they should repay within the repayment period and (ii) consumers can use their credit card directly when making a purchase. The substitution argument is supported by the data, that about 11% of credit card owners say that they can make payments and draw money when they have

(12)

insufficient funds on their current account, and around 28% of cash withdrawals on credit cards indeed happen because of insufficient funds in current accounts. On the other hand, they also identify annual credit card fee as being a major limitation of this substitution effect.

Overdrafts used as a loan. The second category includes overdrafts used as a loan to fund consumers’ consumption when they do not have sufficient funds. These types of overdraft are usually of larger amounts and for longer time periods. However, the closest alternative to overdrafts when the consumers are in need of short-term financing appear to be revolving loans.

According to Zernik (2018), overdrafts perform three functions: (1) an insurance mechanism against rejected payments due to inattention and unintentional overdrawing of the account;

(2) short-term, high-cost credit; and (3) a long-term revolving credit line. The definition of these three functions is not overly strict, since there is not always a clear dividing line between them, and they can interact with each other. However, the framework mentioned is useful to better understand which product can be an appropriate substitute for overdrafts, which behavioural consumer biases impact the use of overdrafts, the pricing of overdrafts and the levels of risk connected with them.

Overdrafts as an insurance mechanism. AN overdraft can happen unintentionally due to consumers' lack of attention to their account balances and/or unawareness of the timing and amounts of their transactions (credited and/or debited). In addition, many consumers do not even realise ex-post that they exceeded their approved balance and how many times they did it. In this sense, overdrafts are precautionary bank instruments which basically protect consumers from the costs of having a payment rejected as the rejection charge is usually higher than the overdraft charge.

Short-term, high-cost credit. Overdrafts can be used as a short-term credit to balance a short-term discrepancy in the timing of current expenses and future income or as a form of a last-resort high-cost credit to help fund unexpected expenses. This use of overdrafts is appropriate when the consumers have no cheaper borrowing option, and when the current transaction brings more value to the consumers in comparison to overdraft fees and repayment.

Long-term revolving credit line. Overdrafts can be used as an open-ended, revolving credit line.

1.5 Overview of the overdraft market: patterns of use

Overdraft lending is a widely used form of credit. According to the study by Reed von Hinten, Mendys-Kamphorst and Jannsen (2014), approximately 44% of current accounts at the four largest Dutch banks had an authorized overdraft facility in June 2013, while the total number of current accounts amounted to 18.6 million. Furthermore, between mid-2012 and

(13)

mid-2013, one third of current accounts were overdrawn at least once at these banks.

Similarly, Zernik (2018) argues that approximately 20% of the US consumers incur at least one overdraft charge per year, of which 13.5–28% incur more than ten charges annually. In 2006, according to the UK Office of Fair Trading, the total arranged overdrafts in the UK had an average daily balance of 8.4 billion pounds. In addition, at least 35% of consumers were occasionally or permanently overdrawn, maintaining an average daily debt of over 1.000 pounds. For the Israel market, the Israel central Bureau of Statistics in their 2013-14 research concluded that approximately 54% of consumers used their overdraft line of credit for at least a month, 41% borrowed most of the time and 28% had been overdrawn for more than a year.

Stango and Zinman (2014) claim that overdraft fees have become common in recent years as banks based their pricing model primarily on service fees (ATM fees, overdraft fees). In an overview of the US overdraft market, Zernik (2018) points out that in 2013, the US banks collected 32 billion dollars in overdraft fees, which amounts to approximately 60-75% of all banks’ revenues from consumers, while, according to Alan, Cemalcilar, Karlan and Zinman (2018), banks in the UK derive almost as much income from overdrafts as from re-investing checking account deposits. According to the British Bankers Association, there were 8 billion pounds outstanding on overdrafts in large retail banks as of July 2013, where overdraft debt doubled between 1997 and 2007, remaining stable until 2010, and then falling afterwards (FCA, 2015).

In the Eurozone, overdrafts have steadily declined in recent years, according to the European Central Bank data. From Figure 2, we can denote that the total value of the overdraft market in period between 2010 and2019 decreased from approximately 900 billion Euros in 2010 to approximately 700 billion Euros in 2019.

Figure 2: The value of total overdraft market in years 2010-2019, the Eurozone (in mio EUR)

*Note: The Eurozone data do not include Cyprus, as the overdraft data for Cyprus are not reported by ECB.

Source: ECB (2020).

898.761 924.357 918.005

881.529

844.833 854.619

797.537

751.148

733.024

705.930

500.000 550.000 600.000 650.000 700.000 750.000 800.000 850.000 900.000 950.000

2010 2011 2012 2013 2014 2015 2016 2017 2018 2019

(14)

Figure 3 shows the structure of the Eurozone overdraft market. The largest share represents the Italian market with its value of 166 billion Euros in year 2019, followed by France (140 billion Euros). Greece, Italy and Spain show a strong decline in overdrafts, while France, Germany and Austria maintain the same levels of overdrafts or even show a small increase in overdrafts over the observed period.

Figure 3: The value of overdraft market by country in years 2010 – 2019 (in mio EUR)

Source: ECB (2020).

2 BEHAVIOURAL INATTENTION

This chapter describes behavioural inattention, starting with a brief overview. Hereinafter, I added evidence pointing to the important role of behavioural inattention in finance as it had been intensively studied in recent empirical literature. The discussion proceeds about the drivers of overdrafts, where limited consumer attention is often put forth as a possible explanation. Some insights were provided to understand consumers’ attitudes toward overdrafts, especially cultural determinants which facilitate overdraft behaviour. People are involved in their culture to such an extent that they often do not notice how their behavioural model and economic thinking are affected. At the end, the chapter briefly introduced the Hofstede's model of aggregate cultural dimensions.

2.1 Introduction to inattention

Mainline economics assumes that humans make choices to maximize utility by using complete information that is immediately available and process it appropriately. Humans are assumed to be time-consistent, unbiased and independent decision makers. This is an over- simplification of true nature, as humans seem to violate rational expectations. The economics

1 10 100 1.000 10.000 100.000 1.000.000

Log scale

2010 2011 2012 2013 2014 2015 2016 2017 2018 2019

(15)

literature raises serious questions about how much these deviations from the standard theory affect economic decisions in the field. Contemporary behavioural works, such as Tversky and Kahneman (1981), Loewenstein (2000), Simon (1997), Gabaix (2019) and many others demonstrate that humans are fallible, rationally bound, inconsistent, biased, emotional and fail to differentiate between what is an immediately obvious portion of information from a tacit information that is not immediately seen. The finding that humans pay attention primarily to most obvious information dates back to Bastiat (1850), a famous French philosopher who argued that humans often fail to make a distinction between what is seen and what is not immediately seen. Modifying the assumption of a rational choice is necessary to attain greater realism and ultimately to improve our understanding of economic thinking.

Attention involves the allocation of scarce cognitive resources, i.e. limited working memory, limited ability to perform complex computations or lack of readily-accessible information and knowledge (Gabaix, 2019). Inattention occurs when agents do not act on information available to them and thus incur unnecessary costs (Scholnick et al., 2008). As argued by Caplin, Dean & Leahy (2019), one key implication of limited attention is that a decision maker may consider only a subset of the available alternatives, ignoring all others. Gabaix (2019) provides a simple example with wine drinking: “[...] we think about just a few considerations (the price and the quality of the wine), but not about the myriad of [income- related] components that are too minor.”

Introducing behavioural inattention into economic models captures a large number of circumstances of behavioural phenomena (Gabaix 2019): inattention to prices and taxes;

base rate neglect; inattention to sample size; over- and under-reaction to news; local inattention to details of the future; global inattention to the future; etc.

In recent empirical literature, inattention has intensively been studied in various fields. For instance, Hirshleifer and Teoh (2003), as well as Cohen and Frazzini (2008) studied inattention to financial news; DellaVigna and Pollet (2009) studied stock return response to earnings announcements released on Fridays, by finding out that as investors are distracted from their work on Fridays, they are also more inattentive to the earnings information which results in more delayed stock return responses; Andersen, Campbell, Nielsen and Ramadorai (2014) studied inattention in household finance based on the Danish household mortgage market; inattention in health plan choices was studied by Abaluck and Gruber (2011), Heiss, Mcfadden, Winter, Wuppermann and Zhou (2016) and Ho, Hogan and Scott Morton (2017);

Scholnick et al. (2008) identify empirical evidence on a sample of 75,000 credit card holders revealing that poorer individuals are more inattentive to credit card repayments than richer individuals; Stango and Zinman (2009) studied implicit and avoidable costs of consumers with respect to their checking and credit card accounts and found that consumers are inattentive to their payment choices by allocating them inefficiently, etc.

(16)

2.2 Inattention in overdrafts

As it has already been pointed out in the market overview in the previous chapter, overdrafts have become an increasingly important market segment over the past decade. However, little is generally known about drivers of overdrafts, and limited customer attention is often put forth as an explanation for overdrafts.

In their research, Stango and Zinman (2014) explored the dynamics of limited attention in the market for overdrafts. They analysed data from thousands of individuals’ checking accounts for the period of up to three years. The main question was how overdrafts were affected by shocks to checking the account holders’ attention, while the surveys used operated as shocks to the salience of overdraft fees. They used different types of surveys regarding household finance services, with specific topics and questions differing from one survey to another. In particular, three types of survey were used: (i) “generic” surveys which did not contain overdraft-related questions, (ii) “overdraft-related” surveys with basic overdraft-related questions, and (iii) “overdraft-focused” survey mostly asking a series of overdraft related questions.

They found that individuals who took overdraft-related questions were less likely to overdraw their accounts in the month of the survey. What is more, individuals who took many overdraft surveys were more resilient from making overdrafts for up to two years as repeated shocks to attention cause sustained changes in behaviour. This suggests that unintentional overdrawing demonstrates a learning curve. The effects are significant for individuals with lower education, lower financial literacy and lower income.

In addition, survey evidence suggests that limited attention is a plausible explanation for overdrafts as, for instance, in the overdraft-focused survey 60% of respondents reported overdrafts, because they thought they had enough money in their account or that the money they deposited was not yet available; both reasons are consistent with limited attention to checking account balances. In the same survey, 24% of respondents did not know or remember whether the bank described different overdraft coverage options at the time of opening the account, which is also evidence of limited attention to account terms.

Stango and Zinman (2014) propose a quite general definition of limited attention, playing a role in overdrafts: “It means incomplete consideration of information that would inform choices, whether that information is about account terms or available balances.” Their definition resembles the view of inattention in overdrafts by Grubb (2012), who sees it as limited awareness of consumers about their own past account usage. In both views, the consumer with limited attention is uncertain about the consumption that might overdraw the account, which leads to occasional overdrafts.

In the study of overdraft pricing structures in the US overdraft market, Zernik (2018) argues that consumers may underestimate the likelihood that they will overdraw their account, as they inaccurately evaluate their own future behaviour and needs. Consumers may

(17)

overestimate their ability to better manage their budget in the future or they may underestimate the possibility that they will later incur an overdraft. Namely, consumers’

preferences may change over time. Thus, while they avoid borrowing in the future, the temptation of spending grows stronger over time and at a certain point, they overdraw their account. Consumers are consequently short-sighted, and they disregard future costs. For example, current costs for having a bank account are valued more than the future costs of paying higher overdraft fees. In addition, future financial decisions entail uncertainty, causing consumers to over- or under-estimate the odds to overdraft their account. Basically, people tend to be over-optimistic regarding their future financial state.

2.3 Economic significance of culture

Culture has an important impact on many dimensions of economic behaviour. Soares, Farhangmehr and Shoham (2006) underline one of the earliest definitions of culture that was proposed by Leopold (1980), who understands culture as “the complex whole which includes knowledge, belief, art, morals, custom and any other capabilities and habits acquired by man as a member of society”. However, Baskerville (2019) pointed to Freilich (1989) who opposed the general agreement about what culture involves. Namely, the concept of culture itself is difficult to describe, as it is all-embracing but contradictory. Moreover, some anthropologists and sociologists even discard the concept of culture as a whole.

This confusion of culture, particularly in terms of the lack of conceptual and definitional clarity, impedes research about the influence of culture on consumer behaviour. To support that, Soares, Farhangmehr and Shoham (2006) refer to Buzzell (1968), who pointed out that culture is merely a convenient determinant (or just a buzzword) for many differences in the market structure and behaviour that can hardly be explained in terms of more obvious factors. Usunier (1999) went a step further by claiming that the concept of culture is often used as an explanation for residuals when other explanations have been proven unsuccessful.

As argued by Guiso, Sapienza and Zingales (2006), when studying economic impact of culture, it is first necessary to define culture as narrowly as necessary in order to be able to identify inference of culture for economic performance. As a result, we will understand culture as those commonly observed beliefs and values that different groups pass through generations almost unchanged. This definition focuses on those particular dimensions of culture whic drive economic behaviour.

Hence, we adopt a widespread approach towards measuring cultural dimensions as developed by Hofstede (2020), who also acknowledges the above-mentioned difficulties in the nature of culture by claiming that “[...] culture is more often a source of conflict than of synergy.” Thus, it is of no surprise that in business and private life, it is sometimes astonishing people from different cultural backgrounds can exhibit such diverse behaviour patterns by. People are involved in their culture to such an extent that they often do not notice how it affects their behavioural model and economic thinking. Cultural background may

(18)

play an important role in daily human decisions and may explain the differences in spending patterns among nations. Culture has long been studied as an integral part of economic life.

One could agree with the assertion by Guiso, Sapienza and Zingales (2006) about the reluctance of economists to rely on culture as a possible determinant of economic phenomena. However, as authors further elaborate, much of this reluctance emerges in our broad understanding of culture and it is difficult to make and assess hypotheses. With the emergence of new techniques and as more data are available, it is becoming easier to identify systematic differences in people’s preferences and beliefs and relate them to culture. Many views have evolved towards the role culture plays in economics performance. For a detailed introduction into the study of interplay between culture and economic well-being, see the pioneering study by Guiso, Sapienza and Zingales (2006). These authors study relationships between economics and culture, in particular, how ethnic background or religion acts through beliefs and/or preferences, affects social capital and economic well-being.

Discussing the impact of ethnic diversity, religious beliefs, language and legal norms on economic, political and social outcomes are beyond the scope of this master thesis. We understand cultural dimensions in aggregate terms according to Hofstede (2020) as the collective category that distinguishes the members of one group or category of people from others. Our approach is similar to Iloie (n.d.), who used the Hofstede’s cultural dimension model to study foreign direct investment flows.

2.4 Hofstede’s cultural dimensions

In recent years, the world has become more and more globalized. As people interact with each other, different cultures intertwine. In the business world, this implies that the managers should be capable to cooperate with a large range of people from different countries and cultures. However, people are often involved in their culture to the extent that they do not notice how it affects their behavioural model and economic thinking. Researchers suggest different models to understand the impact of culture on work and life to determine the dimensions in which cultures vary.

Many different studies find cultural dimensions most appropriate for conceptualising and operationalising culture. Grove (2005) argues that cultural dimensions provide concepts which increase our awareness of the values within human culture and enable measurement and cross-cultural comparisons. However, Hofstede’s framework is the most often used and best-known framework to study cultural differences in a number of disciplines (i.e.

psychology, sociology and business studies).

The beginnings of Hofstede’s Cultural Dimensions Theory originate in 1970s, when Hofstede analysed a large survey database on values and related sentiments. As summarized by Baskerville (2019), this data set included results of an international attitude survey in the multinational corporation IBM with 88,000 employees, working in subsidiaries from 66

(19)

countries. This survey process was repeated for several years, resulting in 117,000 responses.

The results of these data led Hofstede to develop six cultural dimensions for each surveyed country. He constructed indexes for each dimension and linked them to demographic, geographic, economic and political aspects of a society. This extensive research enabled Hofstede to create an empirical framework, which is useful for studying statistical inference of culture in different empirical models and for discerning the ways in which people make culture-related decisions. According to his study, culture consists of the following six dimensions:

• power distance (PDI),

• collectivism vs. individualism (IVC),

• femininity vs. masculinity (MVF),

• uncertainty avoidance index (UAI),

• short-term vs. long-term orientation (LVS) and

• restraint vs. indulgence (RVI).

Cultural dimensions are expressed on a scale which runs roughly from 0 to 100. Their basic relations are shown in Figure 4. I explain them more thoroughly further below.

Figure 4: Hofstede’s six categories

Source: CFI (2020).

2.4.1 Power Distance (PDI)

The power distance index examines how less powerful members of organizations and institutions tolerate the distribution of power in society, which can be distributed unequally or equivalently (Hofstede, 2020; CFI, 2020).

The basic question is how a society deals with inequalities among people. Societies with a

(20)

high score of power distance accept hierarchical order, while people in societies with lower score of power distance seek for equivalent distribution of power and demand justification for inequalities of power. China, Saudi Arabia andIndia are countries with high power distance index. On the other hand, in countries with low power distance index cultures encourage organizational structures that are flat, designated with decentralised decision- making and participative style of government. In these societies, there is strong emphasis on the distribution of power among people. Examples of these countries are western democracies (Hofstede, 2020; CFI, 2020).

2.4.2 Collectivism vs. Individualism (IVC)

The collectivism versus individualism dimension measures how important are individual versus group interests in a society. The dimension quantifies the degree to which individuals are integrated into groups and how dependent from and obligated they are to them (Hofstede, 2020; CFI, 2020).

People in societies with grater individualism are more focused on tending to personal goals.

Individualism usually indicates a preference for a distant social relationship in which individuals are expected to look after themselves and their nearest family. The USA is known as one of the most individualistic countries in the world (Hofstede, 2020; CFI, 2020).

On the other hand, collectivism denotes a preference for closely integrated social relationships and greater importance on the goals and well-being of the group. In more collectivist societies individuals are integrated into close group from their birth onwards (extended families) (Hofstede, 2020; CFI, 2020).

2.4.3 Femininity vs. Masculinity (MVF)

The femininity versus masculinity dimension, also referred to as “tough vs. tender cultures,”

examines what values people find more important in a society (Hofstede, 2020; CFI, 2020).

This dimension considers the preference for achievement, attitude towards sexuality, equality, etc.

Masculinity represents a preference for achievement, heroism, assertiveness and material rewards for success and wealth-building. Usually, societies with higher masculinity score are more competitive (Hofstede, 2020; CFI, 2020).

On the other hand, femininity denotes a preference for modesty, cooperation, people in those societies are more concerned with the quality of life and caring for the weak. Usually, societies with higher femininity are more consensus-oriented (Hofstede, 2020; CFI, 2020).

(21)

2.4.4 Uncertainty Avoidance Index (UAI)

The uncertainty avoidance index expresses the degree to which the members of a society feel uncomfortable with uncertainty and ambiguity in unusual situations. In other words, this dimension considers how people deal with circumstances and events that are unknown, surprising and/or unexpected (Hofstede, 2020; CFI, 2020).

High uncertainty avoidance index indicates low tolerance for uncertainty, ambiguity and risk-taking. Societies with a high score of uncertainty avoidance index use strict rules and regulation with the intent to minimize the effect of uncertain situations. People in these countries are more conservative (Hofstede, 2020; CFI, 2020).

In contrast, low uncertainty avoidance index indicates high tolerance for uncertainty, ambiguity and risk-taking: the unknown is more openly accepted. Societies with a low score of uncertainty avoidance index are more relaxed and more tolerant of situations different from those they are used to. In those societies, practice counts more than principles, ambiguity is accepted and they have minimal rules to constrain uncertain situations (Hofstede, 2020; CFI, 2020).

2.4.5 Short-term vs. Long-term Orientation (LVS)

The short-term orientation vs. long-term orientation dimension is about how people in the society view time horizon (Hofstede, 2020; CFI, 2020).

Societies which are more long-term oriented are focused on the future. People in those societies are long-term driven, which entails giving up on short-term success or gratification in order to achieve long-term success. Values associated with long-term orientation are persistence, perseverance, frugality and long-term growth (Hofstede, 2020; CFI, 2020).

Short-term orientation indicates focus on the near future. People in short-term oriented societies emphasise short-term success or gratification. They rather focus on present than the future. Values associated with short-term orientation are quick results, respect for time- honoured traditions and norms, and fulfilling social obligations (Hofstede, 2020; CFI, 2020).

2.4.6 Restraint vs. Indulgence (RVI)

The dimension restraint vs. indulgence is relatively new. This dimension indicates how well people control their desires and impulses. Indulgent societies denote relatively weak control, they allow relatively free gratification of basic and natural human drives related to enjoying life and having fun. On the other hand, restrained societies denote relatively strong control, those societies suppress the gratification of needs and regulate it through social norms (Hofstede, 2020; CFI, 2020).

(22)

3 EMPIRICAL ANALYSIS: METHODOLOGY

The empirical model of overdrafts will be tested by static linear panel data regression analysis. The methodology is appropriate for panel data, featuring observations on the same entities (country, family, individual, region, plant, company) in different time periods for one entity (Greene, 2012). This section presents a brief overview of the methodology of the panel data regression analysis. The focus is on presenting two basic approaches to the panel data regression: (i) fixed effects estimation and (ii) random effects estimation. The latter part of the section brings an introduction into the Hausman Estimator, which is a widely applied standard test for choosing the more convenient of the two regressions.

3.1 Methodology

Panel data (also known as longitudinal or cross-sectional time-series data) is a dataset containing observations of different cross-sections across time. The fundamental advantage of panel data is that they contain both changes in time and across data sections, which allows great flexibility in modelling differences in behaviour across individuals (Panel Analysis, 2020). Panel data contains observations collected at a regular frequency, chronologically and across a collection of units. These units could represent countries, companies, individuals, demographic groups, etc.

According to Greene (2012), panel data regression analysis is an empirical method, commonly used in social sciences to analyse cross-sectional and longitudinal panel data. The data are usually collected over time and over individuals, upon which regression is run over these two dimensions. Panel data sets are more oriented toward cross section analyses; they are wide but typically short. Heterogeneity across units is an integral part and indeed, often the central focus of the analysis. In a typical panel, there is a large number of cross-sectional units and only a few periods. On the other hand, time series panels have a small number of units observed over a longer period of time.

Examples of panel data include:

• individual level time series data about income taxes, wealth, wages, employment, education, etc.;

• stock-exchange time series data on prices of individual stocks over time¹;

• company level time series data about revenues, employment, fixed assets, profits, inventories, etc.;²

• country level time series data about GDP, unemployment, inflation, money demand, public debt, etc.³

1 Such as the yahoo! finance® data: https://finance.yahoo.com/.

2 Such as the AJPES database: https://www.ajpes.si/.

3 Such as the EUROSTAT database: https://ec.europa.eu/eurostat/data/database or the European Central Bank

(23)

3.2 Panel data regression: basic model

At this point study the linear panel data regression model in a basic framework. Obtaining data over time for the same unit of a cross-sectional observation is useful for many different reasons, but the most important is having the possibility to control the unobserved heterogeneity across units of observation.

As explained by Greene (2012), a standard situation is to observe each 𝑖 ∈ {1,2, … , 𝑘} unit for 𝑡 ∈ {1,2, … , 𝑇_𝑖} time periods. In the panel data setting, the basic model

𝑦_𝑡 = 𝒙_𝒕^′𝜷 + 𝑢_𝑡 (1)

may appear to be overly restrictive, having the same regression coefficient in each time period and the same constant term for all cross-sections. The classical OLS regression model does not allow neither for the presence of time-specific effects nor for the presence of time- invariant differences between cross-sections, such as for instance, differences in cultural backgrounds among units of observation that are in the main focus of the analytical part of this master thesis.

One possible extension of the standard OLS model, when dealing with panel data, is to add time-specific effects 𝒅_𝒕 and/or time-invariant cross-section (group) effects 𝒈_𝒊 in the regression model. The extended model is of the form (Polanec, 2018):

𝑦_𝑖,𝑡 = 𝒙_𝒊,𝒕^′ 𝜷 + 𝒅_𝒕^′𝜽 + 𝒈_𝒊^′𝜹 + 𝑢_𝑖,𝑡^∗ . (2) In case we have the data for 1989, 1990, 1991 and 1992 on a cross section of individuals and would like to estimate the effect of real-wage and money demand on how much individuals overspend according to the limit on their bank accounts. In addition, we would also want to see if cultural background has any meaningful impact on their overspending. A possible empirical model for such purposes is:

𝑙𝑛(𝑜𝑣𝑒𝑟𝑠𝑝𝑒𝑛𝑑𝑖𝑛𝑔)_𝑖,𝑡 = 𝛽₀+ 𝛽₁𝑟𝑒𝑎𝑙𝑤𝑎𝑔𝑒_𝑖,𝑡+ 𝛽₂𝑚𝑜𝑛𝑒𝑦𝑑𝑒𝑚𝑎𝑛𝑑_𝑖,𝑡+ 𝛿₁𝑐𝑢𝑙𝑡𝑢𝑟𝑒_𝑖+ 𝑢_𝑖,𝑡 . (3) In this master thesis, the time-invariant effects of culture on the value of overdrafts are estimated by using panel data regression and overdrafts are represented as an overall value of spending over the bank account limits. Note that in the panel data setting, among other things, it could also be possible to use time dummy variables and observe whether the overdrafts for a particular cultural background have changed over time or not.

In fact, Wooldridge (2010) suggests that the basic rule, with a large number of cross-sections (K) and a small number of time periods (T) is generally considered a good idea, to allow separate intercepts for each time period. This allows aggregate time effects to have the same

database: https://sdw.ecb.europa.eu/.

(24)

influence on 𝑦_𝑖,𝑡 for all i.

However, in the case of the present master thesis there will be more units than periods, but neither of them will be particularly large. Therefore, the empirical investigation is focused on cross-sectional heterogeneity in cultural dimensions across countries, ignoring time- specific effects altogether.

As claimed by Greene (2012), anything that can be done in a classical regression model can also be done in a panel data setting. The analysis of panel data is the subject of intense research in econometrics. This is not a coincidence, as the panel data approach provides a wide variety of ways in which to develop estimation techniques. The approach enables the study of a rich set of issues that could not be studied in isolation in neither cross-sectional nor longitudinal settings.

The general empirical framework for panel data without time-specific effects is a regression model of the form as suggested by Greene (2012):

𝑦_𝑖,𝑡 = 𝒙_𝒊,𝒕^′ 𝜷 + 𝒛_𝒊^′𝜶 + 𝜀_𝑖,𝑡

= 𝒙_𝒊,𝒕^′ 𝜷 + 𝑐_𝑖+ 𝜀_𝑖,𝑡 (4)

where

• 𝑦_𝑖,𝑡 is the dependent (endogenous) variable;

• 𝒛_𝒊^′ 𝜶 = 𝑐_𝑖 is a joint time invariant individual effect, measuring the effect of all the factors that are specific to individual i, but constant over time;

• 𝒛_𝒊 contains a constant term and a set of individual or group-specific variables, which may be observed, such as race, sex and location or unobserved, such as family specific characteristics, individual heterogeneity in skill or preferences, etc. - all of which are taken to be constant over time;

• 𝒙_𝒊,𝒕^′ is a row vector of observations on K explanatory factors for each i at time t, not including the constant term;

• 𝜷 is a column vector of K parameters;

• 𝜀_𝑖,𝑡 is an independent and identically distributed (i.i.d.) error term.

The main objective of the regression is to find consistent and efficient estimates of regression coefficients of all variables used in the model. Whether and to what extend this is possible depends on the assumption of the unobserved effects, where the presence of unobserved fixed effects needs to be tested. This is usually done in a standard way by comparing fixed effects panel data model to random effects panel data model. The latter topic is briefly covered in the remaining subsections, although it is not the aim to discuss theoretical underpinnings of the fixed and random effects model. Therefore, the discussion will be limited to those parts of the methodology which are most relevant for the empirical estimation of economic and cultural determinants of overdrafts as studied in this master

(25)

thesis. Therefore, let us briefly discuss some major features of fixed effect and random effect panel data models and then conclude this part with a brief introduction of Hausman test, which is a standard procedure for comparing the mentioned models.

3.3 Fixed effects model

As explained by Polanec (2018), in panel data where longitudinal observations exist for the same subject, fixed effects represent subject-specific means. In panel data analysis, the term fixed effects estimator is used to refer to an estimator for the coefficients in the regression model including those fixed effects (one time-invariant intercept for each subject).

The fixed effects approach allows for the identification of group specific constant terms in the estimation equation for unobserved effects. In the above general case, these group- specific fixed effects are captured by 𝑐_𝑖. Note that the main aim of this master thesis is to empirically test the statistical inference between the Hofstede’s cultural dimensions and the aggregate levels of overdrafts, where cultural dimensions will be treated as time-invariant group specific observed effects. Thus, we can imagine 𝑐_𝑖 including eventual effects of cultural characteristics on the dependent variable that are specific to country i and constant over the time.

3.3.1 Least Square Dummy Variable (LSDV) specification

In this master thesis, I am interested in the existence of (unobserved) fixed effects in the empirical model of overdrafts and their eventual attribution to cultural dimensions as developed by Hofstede. To estimate the estimation equation (1) with time invariant or fixed effects is to include a dummy variable for each individual in the sample (Polanec, 2018).

The estimated model would be of the form

𝑦_𝑖,𝑡 = 𝑑₁𝛼₁+ 𝑑₁𝛼₁+ ⋯ + 𝑑_𝑚𝛼_𝑚+ 𝒙_𝒊,𝒕^′ 𝜷 + 𝜀_𝑖,𝑡 , (5) where

𝑑_𝑗(𝑖) = {1 if 𝑗 = 𝑖

0 if 𝑗 ≠ 𝑖, (6)

when 𝑗 = 1,2, … , 𝑚 while the remaining variables have the same meaning as in the general equation (1) above. Note that 𝑑_𝑗(𝑖) are dummy variables, each representing fixed effects of a single panel of observations, in our example, a country. They shall be seen as m separate arrays filled with the number of these that equals the number of time observations of a particular country. If the panel 𝑖 = 1 has 𝑇₁ observations, then 𝑑₁ also keeps the first 𝑇₁ values equal to 1 and all the remaining values equal to 0; if the panel 𝑖 = 2 has 𝑇₂ observations, then 𝑑₂ keeps the first 𝑇₁ values equal to 0, the second 𝑇₂ values equal to 1 and all the remaining values equal to 0; if the panel 𝑖 = 3 has 𝑇₃ observations, than 𝑑₃ keeps

(26)

the first 𝑇₁ values equal to 0, the second 𝑇₂ values equal to 0, the third 𝑇₃ values equal to 1 and all the remaining values equal to 0; and so forth.

As we shall see in the estimation part of the master thesis, the comparison between fixed effect and random effect model will unveil if the exclusion of time invariant explanatory variables from the model is also plausible in case of the underlying invariant factors that might cause differences across countries in overall values of overdrafts. Further elaboration will show these effects might be attributed to observable invariant cultural differences and/or to unobservable time invariant differences between countries.

3.3.2 Within groups and between groups specification

The fixed effects model may alternatively be split in its elementary constituent components.

Namely, any fixed effects model of the elementary form

𝑦_𝑖,𝑡 = 𝜗_𝑖 + 𝒙_𝒊,𝒕^′ 𝜷 + 𝜀_𝑖,𝑡 (7) can be split into two parts (Polanec, 2018):

1. within (groups) component:

𝑦_𝑖,𝑡 − 𝑦̅_{𝑖 ∙}= (𝒙_𝒊,𝒕^′ − 𝒙̅_{𝒊 ∙}^′ ) 𝜷 + (𝜀_𝑖,𝑡 − 𝜀̅_{𝑖 ∙}) and (8) 2. between (groups) component:

𝑦̅_{𝑖 ∙} = 𝜗_𝑖 + 𝒙̅_{𝒊 ∙}^′ 𝜷 + 𝜀̅_{𝑖 ∙} (9) Note that the sum of both components yields back the original fixed effects model.

In the within groups component, one estimates the regression coefficients of the modified basic fixed effects model. The modification is done in the way that the fixed effects are first removed from the model altogether. This is done by averaging in the following way:

First, for each observation unit i average over time t:

𝑦̅_𝑖 = 1

𝑇∑ 𝑦_𝑖,𝑡

𝑇

𝑡=1

, 𝑥̅_𝑖 = 1

𝑇∑ 𝑥_𝑖,𝑡

𝑇

𝑡=1

, 𝜗̅_𝑖 = 1 𝑇∑ 𝜗_𝑖

𝑇

𝑡=1

= 𝜗_𝑖 .

(10)

After the averaging is done, the initial fixed effects model is transformed by subtracting averages from the initial model to obtain

𝑦_𝑖,𝑡− 𝑦̅_{𝑖 ∙} = (𝒙_𝒊,𝒕^′ − 𝒙̅_{𝒊 ∙}^′ ) 𝜷 + (𝜗_𝑖− 𝜗_𝑖) + (𝜀_𝑖,𝑡− 𝜀̅_{𝑖 ∙}) . (11) This shows that the fixed effect term 𝑐_𝑖 is thereby effectively excluded from the regression and there is no unobserved fixed effect present in the model. This model is called the within model. Regression coefficients of the within specification are obtained by applying ordinary

(27)

least squares (OLS) and they are the same as regression coefficients of explanatory variables of the LSDV specification.

However, from the within estimation we cannot identify the existence of any time invariant effects in the model, such as the effect of culture. Therefore, the LSDV procedure is suggested in case there are fixed effects in the model and we want them to be estimated.

However, the observation specific effects might just be random and uncorrelated with regression coefficients and therefore, they might be seen as just another type of random noise, but attributed to the characteristics of each unit observed. This variability appears across units (i.e., in our case between states) and it does not change in time. If there are indeed random effects present in the data that are observation specific, then we cannot use the fixed effect specification as the results would not be valid.

In order to check if this is the case, the so called between specification needs to be estimated, which detects the remaining variation occurred due to differences between observations. Let us subtract the within variation from the total variation:

𝑦_𝑖,𝑡 − (𝑦_𝑖,𝑡− 𝑦̅_{𝑖 ∙}) = 𝜗_𝑖+ 𝒙_𝒊,𝒕^′ 𝜷 + 𝜀_𝑖,𝑡 − [(𝒙_𝒊,𝒕^′ − 𝒙̅_{𝒊 ∙}^′ ) 𝜷 + (𝜀_𝑖,𝑡− 𝜀̅_{𝑖 ∙})] . (12) Observe the resulting estimation equation:

𝑦̅_{𝑖 ∙} = 𝜗_𝑖 + 𝒙̅_{𝒊 ∙}^′ 𝜷 + 𝜀̅_{𝑖 ∙} (13) That is called the between estimation equation. According to Polanec (2018), the regression of the between specification turns out to be poor in the presence of observation specific fixed effects in the model and appropriate in case there are observation specific random effects.

As we will see from the remainder of this chapter, the between specification is an important element of the implementation of random effects estimation. The description of the main characteristics of the random effects specification and estimation is the subject of the next subsection.

3.4 Random effects model

The random effects model is an alternative to the fixed effects model. The estimation equation is the same as in case of the fixed effects model (Polanec, 2018):

𝑦_𝑖,𝑡 = 𝑐 + 𝒙_𝒊,𝒕^′ 𝜷 + 𝜗_𝑖 + 𝜀_𝑖,𝑡

= 𝑐 + 𝒙_𝒊,𝒕^′ 𝜷 + 𝜔_𝑖,𝑡 . (14) However, there is a slight difference to the fixed effects model. In the random effect model (Polanec, 2018):

ANALYSIS OF DRIVERS OF INATTENTION IN CONSUMER CREDIT BASED ON EVIDENCE FROM OVERDRAFTS

ANALYSIS OF DRIVERS OF INATTENTION IN CONSUMER CREDIT BASED ON EVIDENCE FROM OVERDRAFTS

TABLE OF CONTENTS

LIST OF FIGURES

LIST OF TABLES

LIST OF APPENDICES

LIST OF ABBREVIATIONS

INTRODUCTION

1 CONSUMER CREDIT OVERDRAFTS

2 BEHAVIOURAL INATTENTION

3 EMPIRICAL ANALYSIS: METHODOLOGY