Can People Trust What They Don‘t Understand? Role of Language and Trust for Financial Inclusion

The topic of FI has been given a lot of attention by researchers and policymakers in the last decades. The efforts to provide as many people as possible with the means to conduct financial transactions, to transfer, save, or borrow funds have strong practical foundations.

According to the definition proposed by the World Bank (2018), FI can be described as access to useful and affordable financial products for both individuals and businesses. The definition also mentions the suitability of such services for various consumer needs and the sustainable way in which they should be delivered. For the purpose of quantitative research, we need to operate on some numeric measures of the level of FI. A number of studies have used the fraction of population over 15 years old having a banking account, having saved or borrowed in the past given months (Allen et al., 2016; Demir, Pesqué-Cela, Altunbas & Murinde, 2020; Grohmann et al., 2018; Xu, 2020), or the number of bank branches/ATMs per 100,000 adults (Xu, 2020) as a measure of FI. To ensure the robustness of the outcomes, usually results for a few measures are compared. A similar approach has been followed in this paper.

The studies on the determinants of FI usually focus either on the supply or on the demand side of the market. From the supply perspective, physical access to facilities such as bank agencies, ATMs, or cash points is key. The proximity to financial intermediaries, among others, has also been reported as a reason for owning an account (Allen et al., 2016; Faber, 2019). However, the number of bank branches, etc., per capita or per square kilometre does not take into account an uneven population density within a given territory. People living in remote areas remain physically excluded from convenient and effective access to financial services. Even in densely inhabited areas, spatial distribution of different financial service points has an impact on customers’ choices, as shown by Faber (2019). Faber (2019) documented the prevalence of areas in the United States with little or no formal financial facilities and linked them to the racial and ethnic distribution of neighbourhoods. It happens that one reason behind the lower formal FI among certain social groups is the insufficiency of bank entities in their area. Faber (2019) raises an argument that it may be banks’ attitude and preference towards a certain type of clients that makes their services not universally accessible. Physical distance may also create a metaphorical distance between certain groups of customers and financial providers, a sense of “not belonging”, which could have an impact on how they perceive financial industry in general (Rengert & Rhine, 2016).

Speaking about FI, we should also note that there is a huge disproportion between developed and developing countries in terms of availability and use of even basic products. Currently, much attention is put to further improve mobile money services, as there has been evidence of their successful adoption and potential in developing countries (Demirgüç-Kunt & Klapper, 2013; Malaquias & Hwang, 2016; Sanderson, Mutandwa & Le Roux, 2018) and their usefulness for people living in remote areas and for women (World Bank, 2017). Shifting to online delivery of financial services could surely improve the availability of financial services; however, in the case of online services, different barriers could emerge, for example, a technological barrier, like having access to the Internet (Hayashi & Minhas, 2018) or a computer literacy barrier. Also, concerns about data privacy or the threat of cybercrime are likely to influence the level of trust people have towards electronic banking products (Png & Tan, 2020). It seems impossible to distinguish this kind of trust from the level of trust in a bank as an institution, given the available data. That is why we did not include online financial services separately in our study. Instead, we focus on financial services as a whole.

The literature on the demand for financial services reveals that characteristics such as gender, age, income, education, and place of residency are important determinants (Africa, Zins & Weill, 2016; cross-country, Allen et al., 2016). Younger, less well-off individuals, those living in rural areas, with lower levels of education, are most likely to be excluded from the financial system. Low-income households with less-educated, minority, and unemployed members are most likely to be unbanked (Hayashi & Minhas, 2018). Across poorest households, sociodemographic factors such as age, race, homeownership, or Internet access matter more than for the average income household. Also, there is a disproportion in account ownership between women and men. Worldwide, 56% of the unbanked are women; this disproportion is observed across most economies. Especially vulnerable are women coming from the least developed countries and often those unemployed (World Bank, 2017). Women are also less likely to open an account themselves if another person in the household already has one (Demirgüç-Kunt & Klapper, 2013).

Another aspect of FI is the type of services people decide to use, by which we mean more than just different types of products. In this study, we focus only on formal financial services and mobile money services and do not consider the market for alternative financing. The reason for this is quite simple–the data in the Global Findex Database, based on which we construct our data set, does not clearly define informal institutions. Payday lenders, pawn shops, family loans, etc., all constitute the shadow financing sector. Their common characteristics are that they operate outside the regulated banking sector and lack formal intermediation (Allen, Qian & Xie, 2019). Another issue is that the market for alternative finance is not at all homogeneous. In the most general manner, Allen et al. (2019) distinguish between constructive and derogatory informal financing. They argue that constructive alternative services are vital for small and medium enterprises (SMEs), which often do not qualify for formal financing. Also, a high concentration of banks’ assets creates an ineffective system for small, riskier enterprises, as they are likely to be denied credit (Karpowicz, 2016; Tsai, 2004). Using informal sources is a way for SMEs to obtain necessary funds. On the other side, predatory lending entities charge their customers extremely high fees, offer inflexible repayment schedules, and due to that, are used for managing temporary liquidity problems (Allen et al., 2019). Gathergood, Guttman-Kenney, and Hunt (2019) report that customers in the United Kingdom who take payday loans are often struggling to repay their overall debt. A reason why individuals may still choose to satisfy their financial needs outside the banking system may be, on the one hand, familiarity, trust, or language compatibility with local financial entrepreneurs, and on the other hand, a lack of the documentation that is required by banks (Faber, 2019; Rengert & Rhine, 2016). The nature of informal financing is different from that of the regulated sector; hence, we have decided to focus only on formal services. Also, other studies adopt a definition of FI as being the use of just formal services (Allen et al., 2016).

The financial system needs both the right infrastructure and well-informed customers (Grohmann et al., 2018). Since people are more responsible for managing their personal finances and securing retirement funds than ever before (Lusardi, 2015; Lusardi & Mitchell, 2014), it is crucial to ensure that they are capable of making reasonable decisions. The literature indicates that to act reasonably, one should be familiar with the concept of interest compounding, inflation, risk, etc., at least on a basic level (Lusardi, 2015). Individuals having such knowledge can be deemed financially literate. This naturally calls for proper early education on how to deal with money (Lusardi, 2015; Lusardi et al., 2010), apart from just assuming consumers are rational and prepared to successfully plan their consumption and savings over a lifetime (Lusardi & Mitchell, 2014). Young adults and women were found to be generally less financially literate and less likely to effectively plan for their retirement (Lusardi & Mitchell, 2008).

In this study, we use the data from Klapper, Lusardi, and Oudheusden (2015), who describe FL rather generally as an understanding of basic financial concepts. They construct their measures based on four FL questions (discussed in detail in the Data section). A more detailed definition was proposed by Lusardi and Mitchell (2014). It names FL as ‘peoples’ ability to process economic information and make informed decisions about financial planning, wealth accumulation, pensions, and debt’. We adopt this definition in our study and aim at expanding the rationale behind the crucial role of processing not only economic or financial information, but information in general. We argue that failing to do so impedes fully informed decisions. The willingness to engage in a relationship with a financial institution also depends on how much an individual knows about the industry (Allen et al., 2016). Later in this paper, we empirically investigate the importance of comprehensible information, which determines how much an individual knows.

It has been shown that a higher level of FL is associated with greater FI, both at the country level (Grohmann et al., 2018) and the individual level (Aslan et al., 2017). The effect is most profound in countries with lower gross domestic product (GDP) per capita. Also, more financially inclusive societies perform better on FL scores, meaning that the relationship between FI and FL can be reciprocal (Klapper et al., 2015). Historical encounters with financial products or phenomena such as inflation make consumers aware of what those phenomena mean for their personal money and savings. To avoid causality issues in our analysis, we follow the understanding of Lusardi and Mitchell (2017), who established a basic causal relationship; it is FL that enables retirement planning, and consequently FI. There is no doubt that factors such as historic background, religion, or language have important roles in the success of economic policies and should be included in studies on policy effectiveness. The importance of cultural factors for economic behaviour has been documented in the literature (Alesina, Devleeschauwer, Easterly, Kurlat & Wacziarg, 2003; Desmet, Weber & Ortuño-Ortín, 2009). Indeed, some behaviours are so deeply rooted in culture that they are profound even if a person changes their living environment. To give an example, Davoli and Rodriguez-Planas (2020) investigated the role of country of ancestry for FL of the population of immigrants in the USA. They found a significant relationship between financial knowledge, risk taking, and patience in the country of ancestry and individual's FL.

Following an extensive study on FL, Lusardi (2015) noted that the effectiveness of current inclusion policies could be undermined by the very language used by industry and policymakers in their communication. It seems hardly questionable that technical and financial jargon is not well suited for those who lack the understanding of the concepts being described. In our study, we want to address the impact that communication has on FI–first, by distinguishing between the comprehension of financial concepts (measured by FL) and understanding of the real-life contracts, and second, by showing that the lack of this understanding is mediated via trust granted to the financial system in general, and represented by trust in banks in this study.

Among reasons for not having an account, 16% of respondents of the Global Financial Survey 2017 declared ‘a lack of trust’. The aim of this paper is to understand if and how the quality and clarity of information are related to the lack of trust. We argue that the relationship between a financial institution and its client requires a certain level of confidence towards one another. Plato-Shinar (2014) even names this relationship a relationship of trust and Guiso, Sapienza, and Zingales (2004) talk about the trust-intensiveness of financial contracts. Turning to individual determinants of trust in banks, we refer to a qualitative study by Rengert and Rhine (2016). Across several focus groups, they find that the second most common reason for not engaging in a relationship with a bank is the lack of trust and unfamiliarity with services. Financial entities are considered irrelevant for some consumers or are too conceptually distant to reach out to. Once again, the importance of creating an environment of mutual understanding emerges. Within our study, we argue that well-informed customers would be the ones willing to participate in the financial system, and the ones participating in it do so in the most optimal way. However, the effect created by the communication gap is intangible–there is no measure that captures how well the information is delivered and how well it is understood. The effectiveness of communication is unobservable, but the impact it has on the level of trust in banks, and consequently, on the level of FI, seems to be of great importance. The uncertainty stemming from limited understanding can be considered as a cost of a contract with a financial institution. Trust and confidence in a financial institution, on the other hand, can act as facilitators of such a contract. Together with country-level and individual characteristics, we try to incorporate the means of facilitation, i.e. trust in the concept of FI. Our study aims at adding the ‘understanding’ component to the analysis of the demand-side determinants of FI.

To address the concern of endogeneity caused by the omission of a variable reflecting the understanding, the lack of which underlies the prevalence of the communication gap, we follow the instrumental variable approach. As discussed, the component reflecting the understanding and comprehension of financial communication (in various means, starting from documentation through media advertisements to personal encounter) seems likely to be correlated with the level of trust exercised towards the banking system. Indeed, the starting point of our study is the observation that incomprehensible financial language impedes trust towards banks, making communication inefficient. That is why we start by looking at the influence of cultural differences on the level of trust to establish theoretical foundations for considering linguistic diversity as an instrument. In an experimental study, Glaeser, Laibson, Scheinkman, and Soutter (2000) have shown that the level of trust is lower between participants of different nationalities. Similar results have been reported by Tenzer, Pudelko, and Harzing (2014), who found language barriers to be significant trust resistors in multinational teams. Barcellos, Smith, Yoong, and Carvalho (2012) found that Hispanic immigrants to the USA experience lower trust towards financial institutions than natives, probably partly due to incorporating beliefs about their countries of origin onto their new reality. Bjørnskov (2007) finds that income inequality, religion, and communist history impact generalised trust on national level. Fungáčová, Hasan, and Weill (2019) observe that different social and cultural factors specifically impact the formation of trust in banks, which leads us to a belief that linguistic diversification may influence trust in banks as well. To our best knowledge, whether such influence is meaningful has not been studied before. The transmission channel would be communication facilitating mutual understanding of a contract between the two parties–in the case of finance, i.e. a financial institution and its customer.

Taking into account just the aspect of communication, from the perspective of a bank, the relationship is relatively less complicated. There are particular procedures that bank representatives have to follow to ensure the identity and needs of a client, follow know-your-customer guidelines, and assess the risk of engaging in a relationship with a given individual or entity. However, the consumer's perspective is far more complex. First, by putting personal funds in a banking account or a mobile money account, in some sense one loses control over those funds. People may worry about unauthorised transfers from their accounts. Hence, establishing trust between parties is crucial. Second, some clients experience concerns regarding data privacy and vulnerability of banks to cyberattacks (Png & Tan, 2020). Finally, consumers often have little or no familiarity with products and procedures and may have no experience with documentation. As noted by Cole, Sampson, and Zia (2011), people tend to not use products they are not comfortable or familiar with, and, given the length and complexity of the language used, banks’ contracts can be difficult to navigate.

The problem of language barriers is not limited only to those less financially savvy. FL is key to understanding the essence of the transaction, but the ‘language’ aspect enables understanding of terms and conditions, allows asking clarification questions, etc.—not to mention the fact that FL measuring questions cover rather simple aspects of finance. Their simplicity is accurate for the purpose they serve—they provide a framework for quantifying financial knowledge. However, what banks expect their clients to sign is much more complicated in a sense of transaction circumstances. Just by looking at papers to be signed, a person can feel overwhelmed. Documents are long, sentences are often written in the passive voice (less common in day-to-day conversations), the font is small, and the text contains a lot of finance-specific terminology. Such documents can be difficult to comprehend even by some bankers, should they not have a legal background (Plato-Shinar, 2014). Additionally, the impact of language may be different, depending on the complexity of the financial product.

The language barriers are also not limited to non-native speakers in a given society. The contracts are long and difficult to navigate, even for native speakers, simply because of the specific terms used in legal or banking language. The problem of language complexity has been stressed in previous research as potentially undermining the effectiveness of inclusion policies (Lusardi, 2015). An even broader issue of limited understanding of certain official texts has been addressed by The Plain Language initiative, which calls for simplicity and readability in all official documents. In some countries, providing parts of contracts in plain language is required by law, for example in the USA. Text in plain language should clearly communicate its purpose, allow for easy identification of key messages, and be fully understandable to the user (plainlanguage.gov, accessed 1 June 2021). On the other hand, whether the language of legally binding contracts should be simplified is questionable. Financial institutions operate under strict regulations and also aim at securing their position should any legal action be taken. Simplicity often comes at the cost of precision, which can be harmful to a bank's interest. In professional legal language, there is no place for ambiguity.

Having identified universal difficulties, it should be stressed that non-native speakers may experience yet another barrier when stepping into a legal relationship with a financial institution. Golding, Goodman, and Strochak (2018) show that limited English proficiency is a significant obstacle to homeownership and advocate distributing information materials in Spanish among Hispanic citizens in the United States to eliminate language barriers. Indeed, language homophily between the customer and service provider was found to increase intentions to use the service (Pezzuti, Pierce & Leonhardt, 2018). Bilingual customers also prefer to be offered service in their native language, especially when what they receive is personally involving (Holmqvist & Vaerenbergh, 2013). The majority, if not all, of transactions regarding one's finances can be treated as engaging personal experience. Providing high quality services in multiple languages would surely come at a cost to a financial institution, but may be a way to establish its competitive advantage by targeting a specific market niche. In terms of understanding, for some individuals the financial jargon may be similarly difficult to comprehend, just as a non-native language may be.

The theoretical reflection on language's role in finance puts us at an impasse. On one end is a consumer, who—in the most extreme case—understands very little of the transaction he is engaging himself in, and on the other end is a supplier—a financial institution securing its interest through a legal contract. The question remains: how to establish a viable communication channel that serves mutual interests? We argue that the indirect impact of language on FI is mediated through trust. Struggling to understand the nature of a transaction reduces trust because of unfamiliarity and lack of comprehension. It is worth emphasising that language diversity itself is not believed to be the reason for lower FI. As discussed before, financial documents and legal language are unreadable for the majority of consumers, even native speakers. What matters, however, is the possibility of meaningful communication between the financial institution and its customers. This communication happens via media advertisements, website content, informational materials, and finally, via personal contact in a bank's offices. This relation can be different should we consider a group of people moving to a foreign country. Barcellos et al. (2012) find that lack of language proficiency is a significant barrier to the use of finance and retirement planning. Financial participation of immigrants is yet another aspect of FI, which we do not cover in our analysis. Should immigrants be included in such a study, the methodology ought to be different, especially because immigrants tend to remember and stick to the characteristics of their country of origin, even after moving to a different country (Barcellos et al., 2012; Davoli & Rodriguez-Planas, 2020).

In this study, we use the Linguistic Diversity Index published by the Summer Institute of Linguistics. The measure is based on the language typology by Joseph H. Greenberg and reflects the probability of two people randomly chosen from the population to have different mother tongues (Ethnologue). An alternative may be the index proposed by Alesina et al. (2003); however it is based on older data, obtained from different sources and collected between 1979 and 2001. Cultural characteristics are considered to be generally stable across time, so these values can still be treated as relatively recent ones. The factors that may undermine their validity are mass migrations, which cause movement of large groups of people speaking their local languages into territories dominated by another language's speakers. However, studies have shown that immigrants tend to chose their destination countries minding the language proximity to their mother tongue, meaning how similar the languages are, as language proficiency significantly influences their chances of getting a well-paid job (Adsera & Pytlikova, 2015; Chiswick & Miller, 2015).

Quantifying cultural characteristics has always been a puzzle for researchers and policymakers. The serious concern is that measures of linguistic heterogeneity do not account for the knowledge of foreign languages. Also, the differences between different indices are very subtle, and in some cases, the classification of a given language (or in fact a dialect) is ambiguous, as argued by Desmet et al. (2009). Depending on the social problem analysed, sometimes the concept of diversity is more accurate and sometimes it is the distance that plays an important role. We argue that in the context of our analysis, it is the diversity that matters. Trust is a social phenomenon, which is stronger between a unified group of people. Language barriers negatively influence trust formation in multilingual teams (Tenzer et al., 2014). A society, in fact, can be understood as an extremely large, multicultural, and multilingual ‘team’.

The variable of interest—FI—can be represented by different measures. Usually, a binary variable is employed, and the probability of a given person having an account, a credit card, etc., is estimated. Such an approach is limited by the fact that there are numerous financial services on offer, and someone may simply favour one type of product over another. It is also important to distinguish between having an account and making actual use of it. Simply being an account owner changes little or nothing but the national statistics; there is no benefit to its holder, especially because maintenance costs are quickly imposed. To give an example, as reported by Karpowicz (2016), less than 13% of account holders (or 5% in rural areas) in Colombia made at least three deposits per month in 2011, yet a steady growth of the number of financially included citizens has been observed. However, people who hold an account are more likely to use other services, like saving or investment accounts or insurance products (World Bank, 2018). The difference in access and use is accounted for in our study by developing separate measures for either activity.

Following Aslan et al. (2017), individual FI scores were constructed by performing multiple correspondence analyses on answers to chosen categorical questions. Two indices were calculated: an access score and a use score, related to access to various products and use of each of them in the past 12 months, respectively. Table 1 summarises components of each index, values for which were initially taken from the Global Findex 2017 Microdata Repository.

Before calculating the scores, we delete answers: ‘don’t know’ and ‘refuse’ and unify answers’ values (1=yes, 0=no). The results indicate that in both cases, much more of the variation is explained by the x axis than by the y axis. Hence, x coordinates were used as predicted FI scores. The bigger the distance from the (0, 0) point, the further from an ‘average’ response of the whole observed population.

In the case of the both scores, positive values were assigned to negative responses (‘no’ or ‘not possible’) and negative to the positive ones (‘yes or ‘possible’). For example, a person who answered ‘yes’ to all four (or six) component questions scored −0.56 (−0.81), while the one who answered ‘no’ got a value of 9.24 (3.57). In other words, a person who owns an account, has a debit and credit card, and has access to some kind of emergency funds got an access score value equal to −0.56. An individual who regularly uses credit and debit cards, uses a mobile money account for checking their balance and transferring money, and has both saved and borrowed in the past 12 months achieved −0.81 in value of the use score. In an opposite scenario, he or she would have scored 9.24 in the access score and 3.57 in the use score. The scores will be later included as alternative measures of FI on the microlevel. To incorporate the aforementioned indices in the cross-country analysis, we calculate the mean access score and mean use score for every country included in our study.

Figure 1 presents the distribution of mean access and use scores against the GDP's logarithm. Clearly the values follow an expected trend. In the upper left corner, countries such as Afghanistan, Tajikistan, Chad, and other less-developed countries are located. In such countries, due to noticeable inequalities in the society, huge differences can be found also in terms of access to and use of financial products between citizens. Also, both indices are significantly and negatively correlated with the country-level FI variable (inclusion_inst) and therefore can be used as appropriately constructed measures.

Figure 1

We performed the first step of our analysis on a sample of 120 countries observed in 2017. The following model was formulated: (1) inclusion_insti=b0i+b1i credit_infoi+b2ifin_literacyi+b3ilang_divi+b4ibank_zscorei+b5ilgdpi+b6iinflationi+b7ieasy_langi+eI, \matrix{ {inclusion\_{inst_i} = {b_{{\it 0}i}} + {b_{{\it 1}i}}\,credit\_{{info}_i} + {b_{{\it 2}i}}fin\_literac{y_i} + } \hfill \cr {{b_{{\it 3}i}}lang\_di{v_i} + {b_{{\it 4}i}}bank\_zscor{e_i} + {b_{{\it 5}i}}{lgdp_i} + {b_{{\it 6}i}}{inflation_i} + } \hfill \cr {{b_{{\it 7}i}}easy\_lan{g_i} + {e_{I,}}} \hfill \cr } where -

inclusion_inst is the percentage of people above 15 years old who hold an account at a formal financial institution;

The model specification made it impossible to observe the impact of trust in FI; there were only 42 observations with all relevant information provided. Due to infrequent publication of indicators (stemming from the institutional character of cultural traits) it was not possible to construct a panel model.

5.1.2

Quantile Regression: Country Level

Next, quantile regression was carried out to capture the distribution of the dependent variable rather than to estimate its mean value. Hypothetically, in countries with higher FI, the influence of information factors and the general economy's stability may be more significant than in countries with lower FI, as the demand for such information is likely greater (variables bank_zscore, inflation, credit_info). On the other hand, in countries with lower FI nonfinancial aspects may play a role in encouraging or discouraging people to use banks’ services (for example, issues with understanding the procedures (variable easy_lang)).

Specification of the quantile regression model was identical to the OLS model; however, this time the parameters were estimated with regard to FI quantiles. The results of our estimation are presented in Figure 2.

Figure 2

As expected, in higher quantiles the impact of bank_zscore and credit_info is rising, but the differences between quantiles in case of this variables, though visible, are not big in size. In case of lgpd, starting at the 0.8 quantile, the prediction values decrease, meaning that the impact of GDP is declining and other factors are important among countries with high FI. No significant effect of easy_lang was found. Considering inflation, people may spend their money faster, and, in order to maintain asset value, store it rather in durable goods, without bothering to keep it in an account. This effect is visible in countries within 0.2 quantile of FI, which are consequently more reliant on cash.

Next, the difference in FL significance between 0.2 and 0.8 quantile was tested. On the 5% significance level, this difference cannot be rejected (p = 0,0104). On the graph, the negative slope across subsequent quantiles is noticeable, yet the size of the difference was too small to be precisely captured.

As a second step of our analysis, we observe individuals and their financial decisions. In models with individual-level data, two categories of right-hand side variables are included. The first group contains personal features (age, gender, level of education, income quantile), while the second one consists of country characteristics. The latter were divided into: information availability and understanding (FL, depth of credit information, easy language), trust (confidence in banks), and controls (GDP, inflation, bank Z-score). Unfortunately, the microdata data set does not include information on individual FL scores. In this case, aggregated values were assigned to each individual from a given country.

To compare results across different measures of FI, the following dependent variables were employed: access score, use score, and simply having an account at a financial institution (binary variable, 1=yes, 0=no). Later we tested the possible explanation of lack of trust and the binary value ‘if does not have an account: because of lack of trust’ (1=yes, 0=no) served as left hand-side variable. We estimated the following models: (2) Model (I):financial inclusion measureij=b0i+b1iagei+b2i femalei+b3i education levels+b4i income quantilei+b5i countrycharacteristicsi \matrix{ {{\rm{Model}}\,({\rm{I}}):} \hfill & {financial\,inclusion\,{measure_{ij}} = {b_{{\it 0}i}} + {b_{{\it 1}i}}} \hfill \cr {} \hfill & {ag{e_i} + {b_{{\it 2}i}}\,femal{e_i} + {b_{{\it 3}i}}\,education\,leve{l_s} + } \hfill \cr {} \hfill & {{b_{{\it 4}i}}\,income\,{quantile_i} + {b_{{\it 5}i}}\,country} \hfill \cr {} \hfill & {{characteristics_i}} \hfill \cr } (3) Model (II):Model (I)+b6i financial literacyi+b7icredit information indexi+ b8i easy languagei \matrix{ {{\rm{Model}}\,\left( {{\rm{II}}} \right):} \hfill & {{\rm{Model}}\,\left( {\rm{I}} \right) + {b_{{\it 6}i}}\,financial\,{literacy_i} + {b_{{\it 7}i}}} \hfill \cr {} \hfill & {credit\,information\,{index_i}} \hfill \cr {} \hfill & { + \,{b_{{\it 8}i}}\,easy\,{language_i}} \hfill \cr } (4) Model (III):Model (I)+b6i bank trusti \matrix{ {{\rm{Model}}\,\left( {{\rm{III}}} \right):} \hfill & {{\rm{Model}}\,\left( {\rm{I}} \right) + {b_{{\it 6}i}}\,bank\,{trust_i}} \hfill \cr } (5) Model (IV):Model (I) +b6i financial literacyi+b7icredit information indexi+ b8i easy languagei+b9i bank trusti, \matrix{ {{\rm{Model}}\,\left( {{\rm{IV}}} \right):} \hfill & {{\rm{Model}}\,\left( {\rm{I}} \right)\, + {b_{{\it 6}i}}\,financial\,{literacy_i} + {b_{{\it 7}i}}} \hfill \cr {} \hfill & {credit\,{information}\,{index_i}} \hfill \cr {} \hfill & { + \,{b_{{\it 8}i}}\,easy\,{language_i} + {b_{{\it 9}i}}\,bank\,{trust_i},} \hfill \cr } where i identifies an individual, j = {access score_i, use score_i, account dummy_i} and b₀ is a constant component. Model (I) is a basic specification and includes individual and country controls. The model is later extended by gradually adding variables of interest.

In total, twelve models were estimated – four specifications for each out of three FI measures. The first model (I) contains only control variables: characteristics of an individual and those of macroeconomic environment (country characteristics), i.e. GDP, inflation, and bank Z-score. In the second model (II), the influence of information—its availability and comprehension—was captured together with control variables. FL reflects the general level of understanding financial concepts and is treated as the consumer-side knowledge factor, while depth of credit information provides information for suppliers—the financial institutions. Easy language possibly makes the exchange of information (and products) easier and more transparent. Availability of credit information is a source of additional knowledge for financial institutions. The third model (III) focuses on the importance of trust in making financial decisions. Only variable bank_trust was added to the initial, control equation, so that the effect of trust and only trust is estimated. However, we hypothesise that trust can complement the knowledge or substitute for the lack of it, and this idea is captured in the last model (IV), where both information and trust variables are included.