Financial Product Awareness, Product Ownership and
Attitude in India
University of Groningen, Faculty of Economics and Business
Master Thesis in Finance
Abstract
This thesis deals with how financial literacy relates to financial behavior in India. Based on the 2014 NCFE (National Centre for Financial Education, in India) survey an empirical model is tested with three measures as a proxy for financial behavior: awareness and ownership of financial products and a score based on financial atti-tudes. In a representative sample of over 76,000 respondents I find that financial liter-acy relates positively to all three financial behavior proxies (awareness, ownership and attitudes). The same result is found in two additional tests. One of these tests also finds the relevance of location and social class / caste in the Indian environment. Based on existing literature I argue that these results are likely to be causal and are unlikely to suffer from reverse causality.
Student number: S3052559
Name: Karel Hruby
Study Program: MSc Finance (Faculty of Economics and Business)
Supervisor: Dr. M. M. Kramer
Key Words: Behavioral Finance, Household Finance, Financial Literacy,
Table of Contents
List of Tables ... III
List of Abbreviations ... IV
1. Introduction ... 1
2. Literature review ... 2
3. Data and Methods ... 6
3.1 Background ... 6
3.1.1 Background on financial markets in India ... 6
3.1.2 Caste system in India ... 8
3.2 Dataset: The NCFE survey ... 9
3.3 Variable Construction ... 10 3.3.1 Key Variables ... 10 3.3.2 Control Variables ... 11 3.4 Descriptive Statistics ... 12 3.5 Methodology ... 14 4. Results ... 17 4.1 Univariate results ... 17 4.2 Multivariate results ... 19
4.2.1 Financial Product Awareness, Product Ownership and Attitude ... 19
4.2.2 Financial Literacy and Financial Products: additional test ... 23
4.2.3 Class / Caste and Locations Categories: an extension ... 25
4.3 Endogeneity and possible limitations ... 27
5. Conclusion ... 28
Acknowledgement ... 30
References ... 31
Appendix ... IV
Appendix A: Correlation matrix ... IV Appendix B: Survey Questions ... V
Appendix B.3: Financial Literacy questions ... XII Appendix B.4: Time Preferences questions ... XIV
List of Tables
List of Abbreviations
BSE – Bombay Stock Exchange INR – Indian RupeeIV – Instrumental Variable
NCFE – National Centre for Financial Education NISM – National Institute of Securities Markets NSE – National Stock Exchange
NYSE – New York Stock Exchange OBC – Other Backward Class
OECD – Organization for Economic Co-operation and Development RBI – Reserve Bank of India
SC – Scheduled Caste ST – Scheduled Tribe
SCR – Socio-Cultural Region
SEBI – Securities and Exchange Board of India PPS – Probability Proportion to Size
1. Introduction
India is a quickly developing country and just as in developed countries, Indians are becoming more and more responsible for their own financial well-being. Given the vast economic development in the last decades and the abolishment of the rigid caste system, an increasing number of people belong to the middle class. With the Indian government slowly abolishing restrictions against foreign companies since the 1990s, Indians have more options in financial decision-making than ever. It is important, however, that in an increasingly complex world of finance, individuals have the nec-essary knowledge and confidence in it to make informed and sensible choices because it will have impact on their future. This may concern both the commonly researched
retirement savings1 and other matters in finance, such as investments in capital
mar-kets2.
The research on the relationship between financial literacy and the quality of financial decision-making (financial behavior) in India is quite scarce. In this thesis, my main research goal is to test this relationship empirically to see if the results for India are consistent with the existing literature from developed countries. To increase the un-derstanding of financial behavior I define it broadly: I look at awareness and owner-ship of financial products as well as on financial attitudes. Moreover, I will inspect what other variables influence financial behavior besides financial literacy. Following that I will conduct two additional tests, the first concentrating only on the impact of financial literacy on individual financial product categories and the second analyzing the influence of social class / caste and location category on financial behavior to emphasize the India touch of this thesis.
The research will be based on the representative 2014 NCFE (National Centre for Financial Education) survey that addressed over 76 thousand individuals and ques-tioned them on various financial behaviors and their financial knowledge.
The remainder of this thesis is structured as follows: section 2 reviews the previous literature on a few key topics related to this research. Section 3 has subsections on essential background on India related to this thesis, the dataset, variable construction,
descriptive statistics and methodology. Section 4 presents the results, which are divid-ed into univariate, and multivariate results and a short argumentation on endogeneity. Section 5 summarizes and concludes the findings of this thesis, discusses its implica-tions and outlines open quesimplica-tions for further research.
2. Literature review
In the field of studies on financial literacy there are several ways how to describe financial literacy, therefore let me begin with a review of the common measurements of financial literacy from the research. Huston (2010) has done a detailed summary of the conventional methods and criticized on this regard that the existing literature does not agree on a consistent definition of financial literacy. A majority of studies even fails to include a definition of their concept of financial literacy, making it difficult to compare the results with other papers. Huston (2010) has reviewed four distinct con-tent areas that are commonly used to measure financial literacy. First, money basics, that may include questions on time value of money, purchasing power and personal financial accounting concepts. Second, borrowing, which focuses on the understand-ing of brunderstand-ingunderstand-ing future resources into the present with the use of credit and loans. Third, investments, which focuses on the opposite, i.e. the understanding of saving present resources for the future with various savings and / or investment products. And fourth, protecting resources, that focus on the respondents’ comprehension of insurance and other risk management techniques. The most commonly implemented are investment concepts and money basics. Only one fourth of the studies reviewed by Huston (2010) implemented all four contents. Newer studies, such as Lusardi and Mitchell (2011b) and Van Rooij, Lusardi and Alessie (2011b) essentially implement the same things, they only frame or call them differently. In both there are questions on numeracy, interest, inflation and risk diversification. Van Rooij, Lusardi and Alessie (2011b) then differentiate between basic literacy questions that people with general education should be able to answer and advanced literacy questions.
be-tween groups of the population and discover the following: men are more financially literate than women, the middle-aged are more financially literate than the young and old and more educated people are more financially literate. Hilgert, Hogarth and Bev-erly (2003) report that in the U.S. high school seniors taking the Jump$tart Coalition's financial literacy tests answered only between 50% and 58% of questions correctly and that other studies have found a below average financial literacy among low in-come consumers, people with lower education, African Americans and Hispanics. Similarly, Lusardi and Mitchell (2007) argue that financial literacy among U.S. Americans is distressingly low but they mention that this problem arises in other countries as well. In Australia and New Zealand only 28% of respondents were rated as “good” when solving actual problems in finance. The aforementioned gender gap and connection of low education and low financial literacy is also found in Australia and New Zealand. In the U.K. Miles (2004) showed that borrowers exhibit a weak understanding of mortgages and interest rates. Lusardi and Mitchell (2007) moreover summarize that in Japan more than 71% of respondents had bad knowledge on equity and bond investments and up to 50% lacked any knowledge on financial products. Young Koreans taking the Jump$tart Coalition test in 2000 had similarly bad results as Americans: most of them failed. In the Netherlands Van Rooij, Lusardi and Alessie (2011a) and Van Rooij, Lusardi and Alessie (2011b) have found that most people (around 80%) get the first numeracy question on the survey right but only 40% of the respondents answer all five basic financial literacy questions correctly. On advanced financial literacy questions the respondents performed even worse, so sophisticated literacy is not widespread in the Netherlands as Van Rooij, Lusardi and Alessie (2011a) point out. In India, Agarwal et al. (2015), with three basic financial literacy questions, found around 80% of correct answers on an individual question but only around 60% of the respondents got all three questions correctly.
management. They find a strong positive relationship between financial literacy3 and daily financial management. Second, retirement savings, as Van Rooij, Lusardi and Alessie (2011a) point out, retirement saving are a good predictor of savings behavior in general. Since the retirement savings are also relatively easy to track / observe, it may be a reason why this proxy is so widespread in the literature. Lusardi and Mitch-ell (2007) were among the first ones to test that there is a positive relationship be-tween financial literacy and retirement savings, without proving the causality. Later studies prove this positive impact of financial literacy on retirement savings to be causal in the United States (Lusardi and Mitchell (2011a)) and in the Netherlands (Van Rooij, Lusardi and Alessie (2011a)). And third, Van Rooij, Lusardi and Alessie (2011b) have studied the impact of financial literacy on stock market participation in the Netherlands to find that a lack of understanding of economics and finance deters households from stock ownership (i.e. a positive influence of financial literacy on stock market participation). Essentially, to wrap this up, the researchers use either a “composite measure” of financial behaviors (or attitudes) from daily financial man-agement or ownership / behavior related to financial products. Or in some studies there is a combination of both.
Another important part of most studies are inferences on how the mechanisms that have been researched influence economic outcomes and what are the implications for policy makers. Huston (2010) suggests that mistakes in financial decision making have impacts on both individual welfare of the decision maker and on all other eco-nomic participants via negative externalities. The commonly researched retirement savings have a very clear impact: people who are underprepared for the retirement will have troubles to make ends meet when they reach higher age. According to Lu-sardi and Mitchell (2011b), most workers (even the older ones) have not planned or thought about retirement. However, according to Van Rooij, Lusardi and Alessie (2011a) this has implications not only for the retirement age but also for sudden in-come shocks (e.g. unemployment), since the people who fail to save for their retire-ment hardly ever have any capital buffer at all. Despite the fact that for many workers in the Netherlands, retirement savings are not a choice, the mentioned problem arises among the Dutch as well. The aforementioned paper of Van Rooij, Lusardi and
sie (2011b) aims to conclude that according to Cocco, Gomes and Maenhout (2005), the welfare loss from not participating in the stock market can be sizable and hence there is a chain of effects, where low financial literacy causes low stock market partic-ipation and the low stock market particpartic-ipation may cause a welfare loss having an effect on all of us. Because of the possible negative consequences that bad financial literacy may have on individuals and on the economy, there is a discussion about whether policy makers should implement obligatory financial literacy courses. How-ever, experimental results about the gains from these courses are mixed. While Man-dell and Klein (2009) do not find improvements in financial literacy of the treatment group after some time, Sayinzoga, Bulte and Lensink (2015) do and it is unclear whether these differences are due to different methodologies or environments.
To this point I have focused on the most well-known papers in the topic of financial literacy and behavior. Unfortunately, the literature is very scarce when I limit myself to papers from emerging markets or from environments that are somewhat similar to India. The briefly mentioned experimental study by Sayinzoga, Bulte and Lensink (2015) took place in rural Rwanda and the researchers conclude three important things. Firstly, that the individuals who took part in the financial literacy course have become more financially literate. Secondly, connected with their improved financial
literacy their financial behavior got better4. And thirdly, that there were no spillover
effects to fellow members of a bank, so the training benefited only the participants. In Chile Landerretche and Martínez (2013) have concentrated on pensions but in a dif-ferent way than we have seen above, they use pension literacy, which relates to knowledge of the Chilean pension system. The outcome shows that with higher knowledge about the pension system, Chileans are more likely to have additional savings outside of the pension system. This relationship is proved to be causal. More pension literate Chileans are also more likely to switch their pension fund type. To get to the studies that concentrate on India, Bönte and Filipiak (2012), have conducted an empirical research on the relevance of social interaction and caste affiliation for indi-vidual awareness of financial instruments and investment behavior in India. The re-sults suggest that the majority of Indian households are unaware of the financial in-struments available and out of the households who are aware, only a minority avails
those products. Consistently with the previous findings, Bönte and Filipiak (2012) conclude that this is caused by low financial literacy and suggest that financial litera-cy education could improve this phenomenon. A later study by Agarwalla et al. (2015) has also found a positive relationship between financial literacy and behavior, which they measured based on daily financial management. Agarwal et al. (2015) describe the following variables as determinants of financial literacy in India: educa-tional level, aggressiveness of the investor and the respondents’ gender. But a more important discovery for the topic of this thesis is that higher levels of financial litera-cy are associated with better financial planning. It is also surprising that the results on financial literacy in India found by Agarwal et al. (2015) are higher than those from developed countries. However, it is questionable, whether this is not due to the meth-od they have used, i.e. as mentioned above, their financial literacy score is based only on three questions.
3. Data and Methods
3.1 Background
3.1.1 Background on financial markets in India
open exclusively to institutional investors, so the only way a foreign retail investor can gain exposure to stocks listed in India is via institutional investors5.
A comparative study by Mukherjee (2007) has brought some interesting insights on how the Indian stock markets are doing compared to large world stock markets, such as the New York Stock Exchange (NYSE) out of which I would like to point out four. First, the Indian stock exchanges (referring to both BSE and NSE) have been very integrative with the global players (especially after 2002), resulting in similar move-ments of the Indian financial market with the global market. Second, the efficiency of trading on Indian stock exchanges is found by Mukherjee (2007) to be faster than that of NYSE and at par with the best in the world. Thirdly, the Indian stock market seems to be at the same level with most Asian counterparts and better than those of emerging economies. Its system even performs well when compared with Japan. And fourthly, a barrier that may hinder the Indian stock market to attain the global level is its low market capitalization.
A report by Dun & Bradstreet (2008) describes how the financial products were originally distributed in India and what has changed recently in this regard. In the past, financial products were offered only by Public Sector Banks (deposit, credit accounts), the Life Insurance Corporation and postal department (recurring deposit, National
Saving Certificate, Kisan Vikas Patra6). Nowadays there are more private and foreign
players in the financial market offering products in addition to the financial products traditionally offered by the public sector. The products put on the market by the pri-vate sector are the ones most of Europeans are familiar with, e.g.: debit and credit cards by banks, open-end and closed-end mutual fund schemes and both life and non-life insurance. The main criticism in the report of Dun & Bradstreet (2008) is a lim-ited reach of the private institutions and that there should be more focus to make at least some of them more accessible to all (social) classes, as well as to people in re-mote rural areas.
As a summary of this short introduction to financial markets in India I would like to point out that even though India is a large market for itself and has some differences from the European and American market, many of the differences are disappearing
5 SEBI grants investment accounts only to foreign institutional investors. An “exception” is high net worth individuals who may qualify for a sub-account of an institution and can construct their own portfolios of Indian stocks.
due to globalization and the Indian financial market is not as foreign to us as it may seem.
3.1.2 Caste system in India
The remains of the Indian caste system are still found in modern times and it is there-fore essential to outline the most important elements of it to comprehend the proce-dure later in this thesis. This sub-section is simplified only to serve a basic under-standing of the contemporary class system in India. Caste is actually a word that orig-inates from the Portuguese casta meaning “race” or “breed”. The (old) Indian caste system consists of Varna, which dates back to the Vedic Indian society and Jāti, a grouping from post-Vedic India. There were four Varnas defined and a fifth category, which was unmentioned in this system, so called Dalits, also referred to as untoucha-bles. The untouchables were people at the bottom of the society, living below the poverty level, containing also some indigenous tribes. Jāti is a more complicated grouping because it consists of thousands of “castes” with a pattern that is more com-plicated than Varnas. The segregation of the society based on castes was enforced in
the British Raj7, where castes determined job allocation8 and other discrimination to
“simplify” the administration. The current system is a result of the rigid caste system from the British Raj. It was in the 1920s when due to social unrests the government started reserving certain government jobs for the backward casts.
After India became Independent of the British Empire, in 1950 the Indian constitution defined the terms Scheduled Caste (SC), Scheduled Tribe (ST) and Other Backward Class (OBC), to protect the historically discriminated backward classes of the society. It is difficult to define here who are members of SC and ST, since this is in high detail in the 1950 constitution but for example most Dalits (“the untouchables”) belong to SC. ST refers to the tribes, i.e. the indigenous population of the Indian subcontinent. Other Backward Class are those members of the society who are educationally and socially disadvantaged and are defined neither as Scheduled Caste, nor as Scheduled Tribe. For members of the SC, ST and OBC there are still jobs reserved in the public sector and are quotas in colleges. Also, they are specially protected by the law against racial crimes.
To sum up, in contemporary India there are still remains of the caste system. However, today we may officially refer only to these four classes: Scheduled Caste, Scheduled Tribe, Other Backward Class and the General class (everyone else, those who is / was not disadvantaged in the society). It is not possible to rank the three “official back-ward classes” highest to lowest because in the Republic of India everybody is guaran-teed equality, so the backward classes were not designed to be superior / inferior to one another. It is commonly spoken of the General class and the backward classes.
3.2 Dataset: The NCFE survey
The survey data I am using for the research was provided to me by the National Insti-tute of Securities Markets (NISM). Before I get deeper into the data, let me briefly outline what this institute does and how it is peered to motivate its reach and re-sources. NISM is an educational initiative of the Securities and Exchange Board of India (SEBI), who is the regulator of securities markets in India. NISM conducts research and offers educational programs and master classes mainly for professionals and students in the field of securities markets with a focus on the Indian market. The National Centre for Financial Education (NCFE) is an organizational entity of NISM
and aims to improve financial literacy across India9. In 2014 the NCFE has
undertak-en a large survey that provides a represundertak-entative cross-sectional sample of the Indian population including over 76,000 respondents.
The sampling plan consists of a multistage cluster sampling with the following four primary steps: selection of district, selection of cluster (village or ward), selection of household and selection of individual respondent. A minimum of 20% of India’s
districts had to be covered (as required by NISM) and at least two districts10 in every
state or Union Territory (UT). To cover full cultural diversity, those 20% were dis-tributed across Socio-Cultural Regions (SCR) proportionately to their population based on the Census of India (2011). Within an SCR the districts were selected with Probability Proportion to Size (PPS) sampling. The clusters were then again selected in proportion to the population in the Census of India (2011) and PPS sampling. The selection of cluster worked in a similar way, i.e. following PPS sampling and
9 More about NCFE’s vision and mission can be found under: <http://www.ncfeindia.org/about-us> last access: November 29, 2016 .
tionately to the population11. The households were selected randomly and
inde-pendently of each other in the cluster12. In a household the surveyors had to choose a
member of the household, meeting the following three conditions: aged 18-80, resid-ing in India and residresid-ing in that household. All of this while the interviewers aimed at achieving a 1:1 male to female ratio and an age distribution mimicking the state pro-file of the whole sample surveyed. Only one eligible person per household was inter-viewed.
3.3 Variable Construction
3.3.1 Key Variables
With key variables I refer to the following three dependent variables of this thesis: Financial Product Awareness, Product Ownership and Attitude, as well as the main explanatory variable: the Literacy Factor.
Financial Product Awareness and Financial Product Ownership are closely related. In the survey, the respondent was first questioned whether he / she is aware of a certain financial product. If the answer was “yes”, a follow-up question was asked, whether the respondent has availed the product (ownership). The individual products asked in
the survey are clustered into these seven categories13: Banking & Savings Products,
Credit & Loans, Other Savings Products, Insurance Products, Capital Market Products, Pension Products and Commodity Futures. This is because there is a need to construct a score for Financial Product Awareness and Product Ownership and this way the measure should not be biased towards one category. In this design, the respondent can
earn a maximum of one point on the scale in each category14.
The third dependent variable is Financial Attitude. It is based on 8 different financial
attitudes that can generate a maximum score of 9 points15. These attitudes are: having
11 There is a sophisticated method, especially in the rural areas. All details can be found under section 1.4.2.2 in National Institute of Securities Market (2014).
12 Details, such as starting point of the route in a cluster, etc. can be found under 1.4.2.3 in National Institute of Securities Market (2014).
13 See Appendix B.1 for the exact financial products and their clustering.
14 It is likely that most people who are familiar with one banking product will be familiar with the other ones as well and there could be a too high score biased towards one category.
a household budget and being (jointly) responsible for it, behavior in financial distress, choosing of financial products, savings behavior, long-term financial goal setting, keeping a watch on personal financial affairs, paying bills on time and considered purchase (buying only things one can afford). The three just described dependent variables are meant to proxy for financial behavior of the respondents and will there-fore be referred to as “financial behavior proxies”.
The main explanatory variable in this thesis is financial literacy16. This score is made
up of 8 questions where the respondent can earn one point on each correct answer17.
The topics covered in the questions are: division, time-value of money, interest com-ponent on loan, knowledge of simple interest, knowledge of compound interest, risk-return relationship, diversification and knowledge of inflation. Each question was formed as multiple choice with a “don’t know option” to eliminate the bias of a lucky guess. Inspired by the practice of Van Rooij, Lusardi and Alessie (2011b) and Kramer
(2016), who also focus their study on financial literacy, I conducted a factor analysis18
on the underlying questions of financial literacy score and have generated one factor, which I will refer to as “Literacy Factor”. If only “financial literacy” is mentioned in this thesis, it will refer to the concept of financial literacy in general.
3.3.2 Control Variables
The first control variable is labeled Time Preferences19 and describes whether the respondent prefers to save up money to benefit from it in the long run. There are three
questions20 in the NCFE survey that make up the score for Time Preferences and the
higher the score, the more long-term oriented is the respondent.
The other control variables define the socio-economic profile of the sample. These are: age, gender, household composition, education, working status, occupation, an-nual income frequency of income, location category and caste / class of the
16 More information can be found in Atkinson and Messy (2012) pp. 16-22 and National Institute of Securities Markets (2014) section 3.3. It is necessary to mention those in papers the variable I describe as financial literacy, they call “financial knowledge”. Not to be confused with “their” definition of financial literacy that has additional components.
17 Exact wording of the financial literacy questions can be found in Appendix B.3. 18 More details on the factor analysis can be found in the Appendix C.
19 This variable in same design comes up in Atkinson and Messy (2012), pp. 33-35 and National Insti-tute of Securities Markets (2014), section 3.1; however, they describe it as “Financial Attitude”. (Not to be confused with Financial Attitude in this thesis.)
ent. All of them are of categorical nature and I have therefore created a dummy varia-ble for each category within a socio-economic dimension.
The Location category and Caste / Class are India specific variables, so let me outline what they stand for. Location category stands for whether the respondent lives in a rural or an urban area. In the NCFE survey this was determined based on the Census of India (2011). The Caste / Class has four categories: General class21, Scheduled
Caste (SC), Scheduled Tribe (ST)22 and Other Backward Class (OBC)23. The
back-ground of the Caste / Class variable was described above, under sub-section 3.1.2.
3.4 Descriptive Statistics
In Table 1 descriptive statistics are displayed. The final sample has 76,757
observa-tions24 (unique respondents). In variable construction I have mentioned that Financial
Product Awareness and Product Ownership have the same underlying categories of financial products and therefore they both of range from 0 to 7. In the sample we can see that on average Indians are aware of about 4 financial product categories but only own financial products from 1.6 different categories. The third variable in Table 1 is Financial Attitude. In the NCFE survey a mean score of 5.9 was achieved on Finan-cial Attitude and it ranges from 0 to 9 points.
The fourth variable in Table 1 is Financial Literacy. The “Financial Literacy Score” is only the number of questions on financial literacy that were answered correctly by a respondent. This is reported here because it is more straightforward to understand than the Literacy Factor. In the sample this was on average 4.7. The corresponding median (not reported in Table 1) is 5 points. Right under “Financial Literacy Score” we can see the Literacy Factor, which is more relevant for the regressions and is therefore here reported as well. The mean is very close to 0, and the factor ranges from -1.83 to 1.29 and is strongly correlated with the “Financial Literacy Score”. In Appendix A the correlation matrix reports a correlation coefficient between the “Fi-nancial Literacy Score” and the Literacy Factor of 0.995.
21 General class refers to those respondents who do not belong to SC, ST or Other Backward Class. 22 Who belongs into SC and ST is determined by the Indian Constitution (Scheduled Castes) Order, 1950.
23 Respondents in Other Backward Class are those who are educationally and socially disadvantaged and are not defined as CS or ST.
Table 1: Descriptive statistics Financial Behavior proxies are the dependent variables of the main regression: Financial Product
Awareness and Financial Product Ownership represent how many product categories the respondent is aware of and owns, respectively, Financial Attitude represents how well has the respondent scored on different financial attitudes. The rest are control variables. Time Preferences stand for long-term planning preferences of the respondent. Other control variables are categorical dummies. Number of observations, mean, standard deviation, minimum and maximum value of each relevant varia-ble are displayed.
(1) (2) (3) (4) (5)
VARIABLES N mean sd min max
Financial Behavior Proxies:
Financial Product Awareness (No. of product categories) 76,757 4.059 1.834 0 7 Financial Product Ownership (No. of product categories) 76,757 1.603 1.335 0 7
Financial Attitude (See Appendix B.2) 76,757 5.926 1.981 0 9
Financial Literacy:
Financial Literacy Score (No. of correct answers) Literacy Factor (from the factor analysis)
76,757 76,757 4.683 0.000 2.055 0.826 0 -1.83 8 1.29
Time Preferences (5 = long term financial preferences) 76,757 3.186 0.953 1 5
Age:
18 to 24 years 76,757 0.197 0.398 0 1
25 to 49 years 76,757 0.585 0.493 0 1
50 to 64 years 76,757 0.157 0.363 0 1
65 to 80 years 76,757 0.061 0.240 0 1
Gender (1 if male, 0 if female) 76,757 0.561 0.496 0 1
Household Composition:
Parents with Children 76,757 0.471 0.499 0 1
Parents with Adult Children 76,757 0.108 0.311 0 1
Joint Family (e.g. with grandparents) 76,757 0.398 0.489 0 1
Single Persons / Siblings 76,757 0.023 0.149 0 1
Disabled Person in the Household 76,757 0.025 0.155 0 1
Education: Illiterate 76,757 0.139 0.346 0 1 Primary 76,757 0.088 0.284 0 1 Upper Primary 76,757 0.134 0.340 0 1 Secondary 76,757 0.180 0.384 0 1 Senior Secondary 76,757 0.190 0.393 0 1 Diploma 76,757 0.024 0.152 0 1 Graduate 76,757 0.245 0.430 0 1 Working Status: Not Working Part Time 76,757 76,757 0.420 0.084 0.494 0.278 0 0 1 1
Full Time (4 or more hours a day) 76,757 0.495 0.500 0 1
Occupation:
Self Employed (Agriculture) 76,757 0.075 0.263 0 1
Agricultural Labourer 76,757 0.047 0.212 0 1
Self Employed (Non-Agriculture) 76,757 0.172 0.378 0 1
Casual Labourer 76,757 0.077 0.267 0 1 Salaried (Private) 76,757 0.138 0.344 0 1 Salaried (Govt.) 76,757 0.056 0.231 0 1 Student 76,757 0.113 0.317 0 1 Housewife / Homemaker 76,757 0.258 0.438 0 1 Retired person 76,757 0.044 0.205 0 1 Others 76,757 0.020 0.139 0 1
Annual Income (in INR; 10,000 INR ≈ € 122)
No Income 76,757 0.398 0.489 0 1 Under 10k 76,757 0.062 0.241 0 1 10k to 50k 76,757 0.206 0.404 0 1 50k to 200k 76,757 0.235 0.424 0 1 200k to 500k 76,757 0.080 0.272 0 1 500k to 1M 76,757 0.016 0.123 0 1 Over 1M 76,757 0.003 0.054 0 1 Income Confidential 76,757 0.000 0.011 0 1 Frequency of Income: Daily 76,757 0.082 0.274 0 1 Weekly 76,757 0.041 0.198 0 1 Monthly 76,757 0.291 0.454 0 1 Irregular 76,757 0.186 0.389 0 1 Others 76,757 0.002 0.042 0 1
Location (1 if rural, 0 if urban) 76,757 0.523 0.499 0 1
Class / Caste:
General 76,757 0.501 0.500 0 1
Other Backward Class 76,757 0.251 0.434 0 1
Scheduled Caste 76,757 0.116 0.320 0 1
The rest of the variables reported in Table 1 are controls. The score of Time Prefer-ences ranges from 1 to 5 and the mean score is 3.2, meaning that the sample is bal-anced, i.e. the respondents neither too myopic nor too long term oriented. The most represented age group is 25-49 years (58.5% of the respondents), followed by the young 18-24 years (19.7%). Just over a half of the respondents are male (56.1%). India’s households seem to be dominated by so-called “nuclear families” (parents with children), as the survey sample suggests. Education is relatively evenly distribut-ed across the categories and most respondents have a graduate distribut-education (24.5%). Up to 13.9% of the respondents are illiterate. Almost half of the sample is working full time (49.5%) and a large fraction is reported as “not working” (42%), however this is largely due to students, retired people and housewives / homemakers, which can be observed in the next socio-economic variable: occupation. Up to 25.8% are house-wives / homemakers, the second most occurring occupation is self-employed outside of agriculture. In terms of annual income, we can see that most of the respondents earn generally said below 200 thousand Indian Rupees a year, which was is an
ap-proximate equivalent of 2,440 Euros25 in 2014. In terms of frequency of income, most
respondents have a monthly income, followed by irregular income. And lastly the “India specific variables”, over a half of the respondents (52.3%) are from rural areas. An approximate half of the sample is defined as General class. Other Backward Class accounts for a one fourth of the respondents, which is a similar part of the sample like Scheduled Caste and Scheduled Tribe jointly.
3.5 Methodology
Financial behavior has been defined in different ways, as was summarized in the literature review. In this thesis there are three proxies for financial behavior: Financial Product Awareness, Financial Product Ownership and Financial Attitude. Firstly, Financial Product Awareness may seem more like knowledge than a behavioral proxy, however, it is a crucial determinant of Product Ownership. Also, a person may be aware of multiple financial products because he / she has actively sought information about financial products available, making this without doubts an essential behavioral proxy. Secondly, Financial Product Ownership has a strong base in the existing
ture (as summarized in the literature review), such as pension savings (e.g. Lusardi and Mitchell (2007)) or stock market participation (Van Rooij, Lusardi and Alessie (2011b)), and is therefore a very adequate second proxy for financial behavior. Third-ly, Financial Attitude also has a foundation in the existing literature (e.g. Mandell and Klein (2009)) and is based on daily management of personal finance. The diversity of financial attitudes in this variable makes this a good proxy for financial behavior in the third regression.
As a main explanatory variable I will use the Literacy Factor (from the factor analysis on financial literacy scoring). In the main model I will control for Time Preferences and for the socio-economic profile of the respondent, namely: Age, Gender, House-hold Composition, Education, Working Status, Occupation, Annual Income and Fre-quency of Income. These controls are in line with Landerretche and Martínez (2013) or Van Rooij, Lusardi and Alessie (2011b), who have similar or in some cases identi-cal control variables. Hence, this procedure should make my results interpretable in a contrast with the exiting studies. To sum up, my main model will look like this:
𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝐵𝑒ℎ𝑎𝑣𝑖𝑜𝑟! = 𝛼 + 𝛽!𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝐿𝑖𝑡𝑒𝑟𝑎𝑐𝑦! + 𝛽!𝑇𝑖𝑚𝑒 𝑃𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒𝑠! + 𝛽!!!𝐴𝑔𝑒 (𝐷𝑢𝑚𝑚𝑖𝑒𝑠)! + 𝛽!𝐺𝑒𝑛𝑑𝑒𝑟 (𝐷𝑢𝑚𝑚𝑦)! + 𝛽!!!"𝐻𝑜𝑢𝑠𝑒ℎ𝑜𝑙𝑑 𝐶𝑜𝑚𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛 (𝐷𝑢𝑚𝑚𝑖𝑒𝑠)! + 𝛽!!!!"𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛 (𝐷𝑢𝑚𝑚𝑖𝑒𝑠)! + 𝛽!"!!"𝑊𝑜𝑟𝑘𝑖𝑛𝑔 𝑆𝑡𝑎𝑡𝑢𝑠 (𝐷𝑢𝑚𝑚𝑖𝑒𝑠)! + 𝛽!"!!"𝑂𝑐𝑐𝑢𝑝𝑎𝑡𝑖𝑜𝑛 (𝐷𝑢𝑚𝑚𝑖𝑒𝑠)! + 𝛽!"!!!𝐴𝑛𝑛𝑢𝑎𝑙 𝐼𝑛𝑐𝑜𝑚𝑒 (𝐷𝑢𝑚𝑚𝑖𝑒𝑠)! + 𝛽!"!!"𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑜𝑓 𝐼𝑛𝑐𝑜𝑚𝑒 (𝐷𝑢𝑚𝑚𝑖𝑒𝑠)! + 𝑒!
squares (OLS) with robust standard errors. An additional test, ordered probit (oprobit), will be completed, to check the consistency of the results with the OLS model.
Nevertheless, the above-mentioned procedure is for the main model of this thesis. I will conduct additional two tests after the main model. In the first additional test,
Financial Product Awareness and Product Ownership categories26 will be used as
dependent variables to compare the effects of the Literacy Factor on individual cate-gories. In other words, there will be one regression for each product category of Fi-nancial Product Awareness and one regression for each product category of FiFi-nancial Product Ownership, in total 14 linear probability regressions in this test. Even though in the case of a probability model it would be econometrically more correct to apply a probit regression, other papers (e.g. Van Rooij, Lusardi and Alessie (2011b)) have also applied a linear probability model because the results are more straightforward to interpret and understand. I will conduct an additional probit regression aside to ensure consistency of the signs, significance levels and magnitude of the effect. The second additional test will be more an extension of the main model of this thesis. I will exe-cute the same regression as the main model (see equation above) again, with two additional India-specific control variables: Location Category and Class / Caste. Con-trol variables of such nature are very rare in the existing literature on financial literacy and financial behavior but they are likely to be significant in the case of India, since these two matters often determine the quality of life of most Indians.
As was seen in the literature review, the relationship between financial literacy and various financial behavior proxies was found in all cases to be positive, therefore, I hypothesize that in my findings the relationship between the Literacy Factor and all of my proxies for financial behavior (Financial Product Awareness, Product Ownership and Attitude) will be positive.
4. Results
4.1 Univariate results
Table 2 reports univariate statistics. Since displays mean values within a given cate-gory and the Literacy Factor is not a dummy variable, I created the categories as (ap-proximate) quartiles, ranked from lowest to highest. These quartiles have a very simi-lar number of observations, yet not equal, because there is a simi-large number of respond-ents having an exact same Literacy Factor (therefore approximate quartiles). When focusing on the columns exhibit the means, there is a pattern observable: with a high-er Lithigh-eracy Factor of the respondent, the mean scores of all three dependent variables, Financial Product Awareness, Product Ownership and Attitude, increase. Because of the grouping that has a similar number of observations across the groups, the inter-preted result should not be biased by discrepancies in number of observations. This finding is in line with my hypothesis that there is a positive relationship between financial literacy and each of the three proxies for financial behavior.
Table 2: Univariate statistics. The second, third and fourth column of the table depict the mean values of Financial Product
Awareness, Product Ownership and Attitude, respectively in a category. The fifth column reports the number of respondents given category. The Literacy Factor is grouped into quartiles from low to high literacy. Since many respondents have the same Literacy Factor a completely equal distribution of the respondents within the quartiles cannot be achieved.
Financial Product Awareness Financial Product Ownership Financial Attitude Number of respondents in the category Mean
Literacy Factor (grouped in quartiles):
1 (low literacy) 3.17 1.06 4.73 19,313 2 3.91 1.42 5.85 19,356 3 4.39 1.76 6.37 19,777 4 (high literacy) 4.79 2.21 6.79 18,311 Age: 18 to 24 years 4.15 1.27 5.58 15,112 25 to 49 years 4.08 1.72 6.09 44,913 50 to 64 years 3.96 1.69 5.92 12,019 65 to 80 years 3.79 1.36 5.52 4,713 Gender: Female 3.76 1.31 5.66 33,662 Male 4.29 1.83 6.13 43,095 Household Composition:
Parents with Children 4.05 1.64 5.99 36,170
Parents with Adult Children 4.10 1.55 5.85 8,300
Joint Family (e.g. with grandparents) 4.06 1.58 5.87 30,537
Single Persons / Siblings 3.97 1.44 5.84 1,750
Disabled Person in the Household 4.15 1.62 5.66 1,894
Education: Illiterate 2.85 0.78 4.74 10,655 Primary 3.38 1.11 5.39 6,770 Upper Primary 3.59 1.27 5.64 10,253 Secondary 4.07 1.57 5.94 13,830 Senior Secondary 4.22 1.63 5.99 14,610 Diploma 4.73 2.19 6.48 1,819 Graduate 5.05 2.38 6.83 18,820 Working Status: Not Working 3.81 1.19 5.50 32,271 Part Time 3.73 1.42 5.93 6,459
Full Time (4 or more hours a day) 4.32 1.99 6.29 38,027
Occupation:
Self Employed (Agriculture) 3.83 1.50 5.82 5,717
Agricultural Labourer 3.09 1.02 5.23 3,625
Self Employed (Non-Agriculture) 4.35 1.84 6.37 13,229
Casual Labourer 3.59 1.14 5.41 5,911 Salaried (Private) 4.65 2.31 6.73 10,554 Salaried (Govt.) 5.18 3.51 7.17 4,328 Student 4.28 1.23 5.61 8,676 Housewife / Homemaker 3.56 1.08 5.38 19,838 Retired person 4.45 1.87 6.16 3,368 Others 3.59 1.17 5.45 1,511
Annual Income (INR; 10,000 INR ≈ €122):
No Income 3.77 1.12 5.44 30,553 Under 10k 3.13 1.16 5.53 4,769 10k to 50k 3.85 1.51 5.85 15,784 50k to 200k 4.49 2.03 6.39 18,074 200k to 500k 5.17 2.88 7.17 6,160 500k to 1M 5.47 3.43 7.34 1,186 Over 1M 5.49 3.71 7.46 221 Income Confidential 4.10 2.60 7.00 10 Frequency of Income: Daily 3.93 1.51 5.84 6,266 Weekly 3.96 1.69 6.18 3,150 Monthly 4.56 2.30 6.61 22,345 Irregular 3.98 1.56 5.88 14,306 Others 4.45 1.77 6.56 137 No Income 3.77 1.12 5.44 30,553 Location Category: Urban 4.35 1.85 6.28 36,599 Rural 3.80 1.38 5.60 40,158 Class / Caste: General 4.38 1.79 6.20 38,468
Other Backward Class 3.98 1.50 5.59 19,266
Scheduled Caste 3.84 1.34 5.47 8,873
Aside from the four consistent patterns, it is also worth pointing out observations about the Class / Caste variable. General class respondents achieve the highest score on all three financial behavior proxies. Other Backward Class achieves higher score on all three measures than Scheduled Caste but only on Financial Product Awareness and Product Ownership than Scheduled Tribe. So even though Scheduled Tribe seems to be the lowest from these three “backward classes”, people from Scheduled Tribe have on average better Financial Attitude than Other Backward Class and Scheduled Caste. It may be that the Indian tribes are skeptical towards financial products but in fact have good financial attitude.
One last thing I would like to point out is that salaried government workers have significantly higher means on all three financial behavior proxies than any other oc-cupation. All government workers in India have a banking & savings related product and are covered in the Indian National Pension System27, which explains Financial Product Awareness and Product Ownership. The high Financial Attitude could be a peer or learning effect from high Financial Product Ownership.
4.2 Multivariate results
4.2.1 Financial Product Awareness, Product Ownership and Attitude
Table 3 shows the output of the main regressions of this thesis with three different proxies for financial behavior: Financial Product Awareness (1), Product Ownership (2) and Attitude (3). The regressions follow the equation from the methodology sec-tion. The most important finding is that in line with the expectations that emerged from the literature review and the univariate results, there is a positive and highly significant relationship between financial literacy (represented by the Literacy Factor) and all three proxies of financial behavior. In terms of economic significance, a one
standard deviation28 increase in the Literacy Factor leads to an increase in Financial
Product Awareness by 0.36 points (categories); in Financial Product Ownership by 0.21 points (categories) and Financial Attitude by 0.6 points.
Time Preferences are significantly and positively related to Financial Product Aware-ness and Product Ownership, hence more patient respondents are, on average, more
27 See: <https://india.gov.in/spotlight/national-pension-system-retirement-plan-all> last access: De-cember 6, 2016 .
financially aware and own more products than respondents who have myopic views on personal finance. Interestingly, Time Preferences are insignificant with Financial Attitude, which can be explained by the fact that the questions in the survey on Finan-cial Attitude were designed to complete the questions on Time Preferences, rather than overlap unnecessarily29. The variable Age shows a high significance across all age groups. The omitted age dummy “18-24 years” makes the base group. The highest coefficients across all three regressions are in the age group 50-64 years, which could suggest that people are most experienced just before retiring. Since Lusardi and Mitchell (2011b) found that the middle-aged are more financially literate than the young and old, I can augment this finding with the insight that the middle-aged also exhibit a better financial behavior than the young and old. Some gender effects are observable as well in the results. Men score, on average, 0.06 on Financial Product Awareness and 0.02 on Product Ownership more than females. This is consistent with the univariate result. Nevertheless, when we look at Financial Attitude, women score, on average, 0.14 points higher than men, which is puzzling because the univariate statistics showed otherwise.
Next is the variable Household Composition. It is difficult to interpret this variable or rather draw any conclusions about it because there are mixed results in the literature. Van Rooij, Lusardi and Alessie (2011b) do not find significance in their household variables (married, number of children) but Landerretche and Martínez (2013) do. My results in Table 3 report that some household dummies are significant, so there seem to be peer effects in Indian households. In the case of Education, as expected, the results are highly significant. All coefficients are not only positive, but also increasing with each higher level of education achieved, which is consistent with the pattern seen in the univariate statistics in Table 1. A Graduate respondent scores more than 1 addi-tional point across every financial behavior measure over Illiterate (for Financial Product Awareness even 1.52, according to the estimation). The next variable, Work-ing Status, has a base group “Part Time workWork-ing”. The dummy “Full Time” is signifi-cant but “Not Working” seems to be less relevant. It is slightly puzzling why the coef-ficient in the third regression (Financial Attitude as a dependent variable) for Full Time is negative (while the first two are positive). In terms of Occupation, the varia-ble Self Employed outside of agriculture is insignificant for Financial Product
ness and Product Ownership and only 10% significant (with a positive sign) for Fi-nancial Attitude. However, the base group is “Self Employed (Agriculture)”, which implies that Self Employed people have similar financial behavior, regardless of whether they are in agriculture or not. The highest coefficients over the base group in Financial Product Ownership and Attitude are in the group “Salaried (Government)”. As mentioned earlier, people working for the government all have a Banking & Sav-ings product(s) and Pension related product(s), so this explains the huge gap they have in Financial Ownership over other groups. Retired respondents have the highest coefficient for Financial Awareness and are second highest in the other two proxies for financial behavior. This is likely thanks to their life experience and having to manage their financial matters very carefully because of not receiving a salary form an employer anymore. The socio-economic dimension “Annual Income” is not signif-icant for Financial Product Awareness and Attitude. Only two income groups show relatively low significance in Financial Ownership and that with a negative sign. This means that the amount of the respondents’ income has hardly any impact on financial behavior. Frequency of income seems to have a consistent impact only in the sense of Financial Product Ownership. And lastly, the constant is relatively high compared to other coefficients and highly significant, which implies that the respondents have some unexplained positive financial behavior per se.
In order to make sure that these results are not biased by the design of the OLS model,
I replicated the model with an ordered probit regression30. In the replicating ordered
probit regression the Literacy Factor is still significant on the 1% level and has a positive effect on all three financial behavior measures. The results are consistent also for the control variables: same signs and significance.
Table 3: Main multivariate analysis to illustrate the relationship between financial literacy and financial behavior. The first
column, Financial Product Awareness, proxies for how many kinds of financial products the respondent is aware of (range 0-7). The second column has Product Ownership of those same kinds of products as a proxy for financial behavior. Ownership is preconditioned by awareness. The third column has a score based on nine behavioral questions (score / scale 0-9) in the NCFE survey as a dependent variable.
(1) (2) (3) VARIABLES Financial Product Awareness Financial Product Ownership Financial Attitude Literacy Factor 0.43*** 0.25*** 0.72*** (0.01) (0.01) (0.01) Time Preferences 0.07*** 0.01*** 0.01 (0.01) (0.00) (0.01)
Age to the base group: “18 to 24 years”:
25 to 49 years 0.04** 0.32*** 0.48*** (0.02) (0.01) (0.02) 50 to 64 years 0.15*** 0.46*** 0.56*** (0.02) (0.02) (0.03) 65 to 80 years 0.16*** 0.37*** 0.42*** (0.03) (0.02) (0.04)
Gender to the base group “Female”:
Male 0.06*** 0.02** -0.14***
(0.02) (0.01) (0.02)
Household Composition to the base group “Parents with children”:
Parents with adult children 0.10*** -0.04*** -0.06***
(0.02) (0.01) (0.02)
Joint Family 0.01 -0.03*** -0.09***
(0.01) (0.01) (0.01)
Single Persons / Siblings -0.00 -0.16*** -0.01
(0.04) (0.03) (0.04)
Disabled Person in Household 0.29*** 0.13*** -0.10**
(0.04) (0.03) (0.04)
Education to the base group “Illiterate”:
Primary 0.33*** 0.17*** 0.36*** (0.03) (0.01) (0.03) Upper Primary 0.47*** 0.28*** 0.52*** (0.02) (0.01) (0.03) Secondary 0.82*** 0.47*** 0.67*** (0.02) (0.01) (0.02) Senior Secondary 0.95*** 0.58*** 0.76*** (0.02) (0.01) (0.03) Diploma 1.32*** 0.97*** 1.07*** (0.04) (0.03) (0.04) Graduate 1.52*** 1.01*** 1.20*** (0.02) (0.02) (0.03)
Working Status to the base group “Part Time”:
Full Time 0.17*** 0.10*** -0.09***
(0.02) (0.02) (0.02)
Not Working -0.07 -0.05* -0.22***
(0.05) (0.03) (0.04)
Occupation to the base group “Self Employed (Agriculture)”:
Agricultural Labourer -0.30*** -0.21*** -0.28***
(0.04) (0.02) (0.04)
Self Employed (Non-Agriculture) 0.01 -0.01 0.07**
Annual Income to the base group: “No Income”: Under 10k -0.27 -0.99** -0.79 (0.60) (0.46) (0.56) 10k to 50k 0.24 -0.81* -0.71 (0.60) (0.46) (0.56) 50k to 200k 0.50 -0.61 -0.61 (0.60) (0.46) (0.56) 200k to 500k 0.71 -0.27 -0.32 (0.60) (0.46) (0.56) 500k to 1M 0.82 0.12 -0.39 (0.60) (0.46) (0.56) Over 1M 0.76 0.38 -0.35 (0.61) (0.47) (0.57)
Frequency of Income (no base group):
Daily -0.06 1.09** 0.88 (0.60) (0.46) (0.56) Weekly -0.18 1.12** 1.14** (0.61) (0.46) (0.56) Monthly -0.16 1.08** 1.03* (0.60) (0.46) (0.56) Irregular -0.09 1.07** 0.87 (0.60) (0.46) (0.56) Others 0.36 1.23*** 1.56*** (0.62) (0.47) (0.58) Constant 2.56*** 0.35*** 4.87*** (0.06) (0.04) (0.06) Observations 76,757 76,757 76,757 R-squared 0.24 0.38 0.26
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
4.2.2 Financial Literacy and Financial Products: additional test
As a starting point of this sub-section, let me recap the categories that make up the score for Financial Product Awareness and Product Ownership. These are: Banking & Savings Products, Credit & Loans, Other Savings Products, Insurance Products, Capi-tal Market Products, Pension Products and Commodity Futures. In this additional test I will estimate probability models to show that there are differences between the cate-gories of financial products. Table 4 reports the summarized regression outputs of these probability models. The control variables are not reported, as their effect was addressed in the main model.
It is also essential to mention that in regressions (1), (6), (8) and (13) the (unreported) control dummy “Salaried government worker” had to be omitted from the probability models because it predicted success perfectly. This is because of the aforementioned Indian National Pension System that covers workers of the Indian government.
Next I would like to point out some observations that can be taken away from the results in Table 4. We can see in Panel A (Financial Product Awareness) that the Literacy Factor coefficients are significantly lower in regressions of Banking & Sav-ings (1) and Insurance (4), while their constants are higher than in regressions with more complicated products, such as Other Savings (3) or Capital Market Products (5). This means that a majority of people is aware of the most common financial products per se and financial literacy is more important for complicated products in terms of Financial Product Awareness. The only product category that does not follow this pattern are Commodity Futures (7), however, they have both very low awareness and ownership.
When we look at Panel B (Financial Product Ownership) the same pattern is observa-ble only with the constants (relatively low for “complicated products”, high for “basic products”) but an opposite pattern with the Literacy Factor. The highest coefficients of the Literacy Factor (in Financial Product Awareness) are for those basic products, Banking & Savings (1) and Insurance (4).
The conclusion of these observations is that financial literacy has a lower effect on Financial Product Awareness of basic products than on their Product Ownership. The
opposite is true for more complicated products. A replicating probit model31 was
conducted that showed consistent signs, magnitude and significance levels with the linear probability model (OLS).
Table 4: Additional multivariate test on Financial Product Awareness and Product Ownership to illustrate the relationship
between the Literacy Factor and the individual categories of financial products available on the Indian financial market. Panel A displays the output for Financial Product Awareness, regressions (1) – (7), Panel B for Product Ownership, regressions (8) - (14). The individual product categories are abbreviated in the following way: B&S = Banking & Savings, C&L = Credit & Loans, OtS = Other Savings, Ins = Insurance, CapM = Capital Market Products, Pen = Pension, Com = Commodity Futures. The control variables are not reported here, see Table 3 for their list. In regressions (1), (6), (8) and (13) the control dummy “Salaried (Government)” (Occupation) is omitted because it predicts success perfectly.
Panel A: Financial Product Awareness
(1) (2) (3) (4) (5) (6) (7)
VARIABLES B&S C&L OtS Ins CapM Pen Com
Literacy Factor 0.03*** 0.07*** 0.08*** 0.05*** 0.09*** 0.08*** 0.04***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Other Controls YES YES YES YES YES YES YES
(see Table 3)
Constant 0.75*** 0.58*** 0.22*** 0.67*** 0.06*** 0.38*** 0.08***
(0.01) (0.02) (0.02) (0.01) (0.01) (0.02) (0.01)
Observations 76,757 76,757 76,757 76,757 76,757 76,757 76,757
R-squared 0.076 0.106 0.106 0.090 0.201 0.136 0.048
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Panel B: Financial Product Ownership
(8) (9) (10) (11) (12) (13) (14)
VARIABLES B&S C&L OtS Ins CapM Pen Com
Literacy Factor 0.08*** 0.05*** 0.03*** 0.08*** 0.01*** 0.01*** 0.00***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Other Controls YES YES YES YES YES YES YES
(see Table 3)
Constant 0.35*** 0.15*** -0.03*** 0.12*** -0.04*** 0.19*** 0.01***
(0.01) (0.01) (0.01) (0.02) (0.01) (0.01) (0.00)
Observations 76,757 76,757 76,757 76,757 76,757 76,757 76,757
R-squared 0.181 0.100 0.080 0.183 0.088 0.384 0.012
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
4.2.3 Class / Caste and Locations Categories: an extension
The main equation that I presented in the methodology section was used in Tables 3 and 4. As an extension, there will be two additional control variables that are specific to India: Location Category (Urban vs. Rural) and Class / Caste (General Class, Scheduled Caste, Scheduled Tribe and Other Backward Class). Class or caste is a thing that cannot be influenced, usually people are born into them. Extensive litera-ture (e.g. Fan, Hazell and Thorat (2000); Borooah and Iyer (2005); Thorat and Neu-man (2012)) has shown that the caste often determines important matters that influ-ence the life standard, such as education. Bönte and Filipiak (2012) found that people from backward castes are much less aware of financial products that are available. The location problem is a common thing not only in emerging markets but also in general in large countries with high population. The rural areas are less developed in many ways than urban areas.
highly significant and positive. On the other hand, the R-squared has increased by a little bit (in the output table with rounded values it can be observed mainly for Finan-cial Awareness). The variable Location uses the base group “Urban” and the coeffi-cients of “Rural” are negative and significant. This supports the outlined theory that rural areas are less developed than urban ones. On average, people in urban areas are more financially aware, own more products and have better financial attitudes.
The Class / Caste variable shows a surprising result, that in regression (3) the category “General” has a negative sign. This is surprising because the base group of Class / Caste in this regression is Scheduled Tribe (ST) and in the univariate results, we have seen that the General class has the highest mean score on Financial Attitude of all classes / caste. Hence, Scheduled Tribe, despite performing badly on Financial Prod-uct Awareness and ProdProd-uct Ownership, has relatively good results on Financial Atti-tude. This is less so the case of Financial Product Awareness and Product Ownership, here the General class achieves the highest (and positive) coefficients. Scheduled Caste (SC) and Other Backward Class have their coefficients quite close to each other and they are both positive over ST (referring to regressions (1) and (2)).
Table 5: An extension to the main multivariate analysis with two additional socio-economic variables: Location Category and
Class / Caste. These variables are specific for the Indian environment.
(1) (2) (3)
VARIABLES Financial Product Awareness Financial Product Ownership Financial Attitude
Literacy Factor 0.40*** 0.24*** 0.71***
(0.01) (0.01) (0.01)
Location Category to the base group: “Urban”:
Rural -0.09*** -0.12*** -0.22***
(0.01) (0.01) (0.01)
Class / Caste to the base group "Scheduled Tribe":
Scheduled Caste 0.76*** 0.20*** -0.33***
(0.02) (0.01) (0.03)
Other Backward Class 0.75*** 0.23*** -0.38***
(0.02) (0.01) (0.02)
General 0.83*** 0.28*** -0.11***
(0.02) (0.01) (0.02)
Other controls YES YES YES
(see Table 3)
Constant 1.95*** 0.23*** 5.29***
(0.06) (0.04) (0.06)
Observations 76,757 76,757 76,757
R-squared 0.26 0.38 0.26
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
literacy, is an essential determinant of financial behavior, which is consistent with previous literature, such as Bönte and Filipiak (2012); Thorat and Neuman (2012); Agarwal et al. (2015).
4.3 Endogeneity and possible limitations
As a last point of the results section, I would like to address endogeneity. It is not possible to conclude the current results with certainty as causal. Lusardi and Mitchell (2014) provide some insight into how to interpret the relation between financial liter-acy and financial behavior. They show that IV regression estimates are consistently larger than OLS estimates. Since in their review, among others, they mention Van Rooij, Lusardi and Alessie (2011b), whose method was partly inspirational for this thesis, it is likely that the results I interpreted are lower than the actual ones.
The whole thesis was a little bit limited by the scope of the NCFE survey. The last test (on castes and location categories) has then “exhausted” the last variables availa-ble. Endogeneity may be caused by the problem of omitted variables. By including a plethora of controls, I attempted to circumvent that problem as good as possible. The question is, whether the significant constant term that I found in my regressions is because of a “benchmark financial behavior”, i.e. a good financial behavior that all respondents have per se, or whether there are additional factors that were not captured. Nevertheless, the 2014 NCFE survey does not provide additional socio-economic variables.
Another questionable point may be, if the results are not influenced by reversed cau-sality, namely if high financial literacy is not caused by a high level of Financial Atti-tude, for example. Based on the reasoning of Allgood and Walstad (2016) , I would like to argue that reverse causality is unlikely. In their sample, Allgood and Walstad (2016) find that people with high financial literacy do not demonstrate a consistently good financial behavior and people with low financial literacy do not demonstrate a consistently bad financial behavior. Hence, reversed causality should not be a prob-lem.
5. Conclusion
This thesis aimed to identify the relationship between financial literacy with the awareness and ownership of financial products as well as with financial attitudes of Indians. To that end, I used the NCFE survey from 2014 that covers a representative sample of over 76,000 respondents across India. Financial Product Awareness, Prod-uct Ownership and Attitude have served as proxies for financial behavior, each of them capturing a slightly different aspect of it. Consistently with the earlier findings in this field, I find that financial literacy is significantly and positively related to fi-nancial behavior. This result is consistent for all three proxies and was found after controlling for time preferences and socio-economic characteristics. There is also a significant and positive relationship between financial literacy and each of 7 financial product categories individually (both in awareness and ownership), which ensures consistency with the previous papers that proxy for financial behavior with financial product ownership. A test on India specific variables (location category and class / caste) has confirmed the consistency of the aforementioned results and showed the high relevance these India specific controls. Even though the above-mentioned results have not been tested for causality empirically, other papers on the same topic have found causality in the effect and as Lusardi and Mitchell (2014) argue, the coeffi-cients are usually larger than in OLS when tested for causal effects. Reverse causality should not be an issue based on earlier literature.
might be a good idea to include a few additional questions that address self-assessed financial literacy in future surveys to better understand financial behavior. Other pa-pers have found self-assessed literacy to be a significant variable in this regard. And fourth, this promotion of financial education should be focused on the backward clas-ses / castes of the society. Bönte and Filipiak (2012) also reported a very low financial product ownership in the backward castes in India but suggested that this is unlikely to be “caused by the castes” but by low financial literacy. The Indian government has limited reach in terms of integrating the backward castes with the general class, so both educational institutes and financial institutions should address this part of the society more extensively.
Acknowledgement
I would like to extend my huge thanks to the National Institute of Securities Markets in India for providing me the dataset on the 2014 NCFE survey on which they have worked very hard. Special thanks goes to Mr. Sandip Ghose, director of NISM; Mr. G.P.Garg, registrar NISM and head of NCFE; as well as to Mr. Sandeep K. Biswal, who had put a lot of effort into this survey and gave me some insights on it during my internship at NISM; and last but not least Ms. Nandini Dubey for giving me guidance at NISM during my internship and stay in Mumbai in the summer of 2016.
