We analyze how FinTech adoption improves consumer financial decision-making. Using a regression discontinuity in time design, we exploit the exogenous introduction of a mobile application for a financial aggregation platform. In response, individuals accessed information about their transactions and bank account balances more often, which led to significant reductions in high-interest unsecured debt and bank fees. The magnitudes are economically significant: for the overall population, one additional monthly login reduced consumer debt by 14 percent over a 2-year period. Additionally, we complement our within-individual identification with cross-sectional evidence using a difference-in-differences estimation strategy. Our empirical finding that additional information about account balances helps to reduce consumer debt sheds light on the widespread use of such debt throughout the developed world, which has been a long-standing puzzle in the household-finance literature.
How does access to information affect individual decision making? For decades, this has been one of the most fundamental questions in economics (e.g., Stigler, 1961). With the advent of the information age, and the surge of FinTech products consumers use (e.g., mint.com, personalcapital.com, YNAB.com), we would presume that people are better informed and equipped to make good choices. However, beyond measuring the adoption of new technology, actually quantifying its economic impact is challenging. We know that people of different generations and demographic backgrounds incorporate new technology into their lives at different rates (Anderson, 2015). But we know very little to date about how this affects actual outcomes, and whether any effects vary cross-sectionally in the population. Sorting this out in a robust and careful way is challenging because it is typically impossible to deal with endogeneity, reverse causality, and omitted variables without making some leaps of faith. While it is intuitive that less costly access to information may increase welfare, it is also likely that higher wealth increase either the incentives to acquire information or the ease of accessing it.1 In this paper, we address these issues by using a unique data set from Iceland. A substantial fraction of the citizens in the country use a common on-line platform that consolidates all of their bank account information and transaction histories in one place. Before 2014, access to this personal financial information only occurred via the Internet on a desktop or laptop computer. On November 14, 2014, a mobile application was exogenously released, which gave users easier and remote access to bank account information. Figure 1 shows the propensity to log in to the financial aggregation platform before and after the mobile app introduction, documenting that consumers indeed increased their information access in response to the availability of new technology. Before turning to the effects this had, it is important to note that the mobile application did not offer consumers either notifications or financial advice before they actually logged in and it did not have a functionality to execute transactions. As such, any observable change in consumer outcomes that we document occurred because of less costly access to information and thus more frequent information acquisition, not because of notifications, financial advice, or more convenient transactions.
In our data set, we have time-series information about the frequency and method of access to bank information (desktop vs. smartphone), which we can analyze together with demographic data, economic decisions (e.g., consumption and savings), channels through which consumers access credit (credit cards versus debit card overdrafts), and the resultant financial outcomes (consumer debt and bank fees). One key economic outcome that we focus on is the tendency for people to pay penalties in the form of interest on short-term uncollateralized debt and other fees, such as late and non-sufficient fund fees. No matter what an individual’s utility function might be, paying lower bank fees in response to voluntarily acquiring more frequent information should improve their welfare. We use a regression discontinuity in time design, where the time of the exogenous introduction of the mobile application is used to instrument logins. This allows us to isolate the causal impact of more information on economic outcomes. In the first stage, we estimate the change in an individual’s propensity to log into their financial accounts, which characterizes how new technology affects access to information. In the second stage, we use the predicted jump in logins at the time of the app introduction to identify the per login effect of the app introduction on financial fee payments and other economic outcomes. In both stages, we include individual fixed effects to control for all time-invariant omitted factors or individual characteristics that could affect the economic outcomes we measure. Thus, we estimate a causal within-individual local average treatment effect (LATE). However, since the app was broadly adopted relatively quickly by all age groups and we look at a long time period before and after the introduction of the app, we believe our results can be generalized. Because the regression discontinuity in time design relies on time-series variation for identification, one potential challenge is a higher likelihood that important covariates are also discontinuous to the forcing variable over time. If other confounding events take place at the same time and affect adopters differently from non-adopters, our identification would be challenged. We undertake a number of steps to address this problem. First, we carefully analyze whether any other confounding institutional changes or new regulations occurred during our time period of the app introduction.2 Second, we include controls for the month of the year and for concurrent economic conditions (e.g., interest rates, inflation, and unemployment). Third, we use different functional forms and bandwidths of time from the app introduction to alleviate concerns about time-series trends and time-varying confounds that we assume to change smoothly across the date of the experiment.3 Finally, as a robustness check and to complement our within-individual time-series identification approach with cross-sectional identification, we also employ a difference-in-differences estimation strategy to document the effects of the app introduction on adopters versus non-adopters. The mobile application helped consumers pay less financial fees. Based on just the raw data, Figure 2 shows the total bank penalties paid during our time series. Up until the introduction of the app, financial fees increased. But, once the app was introduced, there was a trend reversal and the amount of fees paid grew at a lower rate. Based on our regression results for the entire population, each added login was associated with approximately $2.24 lower bank fees per month and $1.77 lower overdraft interest. Accounting for the frequency of individual monthly logins, logging in at least once in response to the app introduction was associated with a decrease of $19.62 in bank fees and $15.47 in overdraft interest. Given that the average individual who adopted the app rolls over $1,356 in overdraft debt ($4,698 conditional on having overdraft debt) and pays around $13 in interest expenses per month, when they look at their overdraft balance one more time per month, they reduce their overdrafts by approximately 14 percent over a 2-year period.
Furthermore, Figures 3 and 4 separately display the evolution of overdraft interest and late fees paid during our time series. We can see in the raw data that most of the reduction in overall fees is driven by overdraft interest, i.e., individuals roll over less highinterest consumer debt in response to accessing information about their account balances more often.
Relative to the monthly expenditures during the sample period, the effect we document represents an economically meaningful change, especially for lower income households.5 Moreover, our findings help to shed light on why high-interest consumer debt exists in such magnitudes, even though it is not consistent with standard preferences in life-cycle consumption models (Laibson et al., 2000). Indeed, the 2015 American Household Credit Card Debt Study estimates the total credit card debt owed by an average U.S. household to be $15,762, which amounts to a total of $733 billion. Our empirical finding that consumers manage to reduce their debt holdings by paying more attention to it speaks to non-standard preferences and overconsumption problems as a likely explanation for the initial use of consumer debt (Laibson et al., 2007), rather than rational consumption smoothing in response to permanent income shocks when funds are tied up in illiquid savings (Kaplan and Violante, 2014) or rational consumption smoothing in response to transitory income shocks (Keys, 2010; Sullivan, 2008). Beyond making more responsible consumption-savings decisions after paying more attention to bank account balances, the observed drop in financial penalties could be explained by changes in how people used consumer credit. We find this to be the case as well. In the total population, adoption of the technology was associated with a 10.6% growth in credit card use relative to debit cards in managing short-term liabilities. Increasing credit card use is a rational response to having better information. Since credit cards offer a 30-50 day float to avoid paying interest for convenience users, compared to overdrafts where interest is incurred immediately, they are superior to consumers for very short-term debt holdings. However, such better liability management can only explain a fraction of our reduction in overdraft interest. We thus conclude that paying greater attention to finances itself causes more prudent consumption and savings decisions, which is also consistent with the fact that we observe most of the fee reductions in overdraft interest rather than late fees or non-sufficient fund fees. In other words, the savings appear driven by more prudent spending decisions instead of being relatively mechanical reductions in avoidable fees. The academic profession has only hit the tip of the iceberg in characterizing the potential benefits and costs of technology on consumer financial decision-making. This is a nascent and growing literature. Agarwal et al. (2018) show that individuals learn to avoid late fees after having paid them initially. Stango and Zinman (2014) document that individuals respond to surveys about overdrafts by paying greater attention to account balances and incurring fewer fees. Levi (2015) shows that individuals respond to information about their net worth by increasing their savings in certain conditions.(2016) finds that reminders for timely payment reduce the credit card late fees that are paid. Karlan et al. (2016) show that text message reminders help consumers to avoid penalties.6 Generally, these studies involve some sort of reminder or stimulus that induces people to change their behavior towards one specific form of avoidable financial fees. In contrast, in our study, we show that simply because individuals access information more frequently without receiving reminders, messages, or other stimuli, they reduce their high-interest consumer debt. Furthermore, a defining feature of our study is that we have information not only on individual financial standing at very high accuracy and frequency, but also on individual access to that information via logins to the financial aggregation app. Consistent with our results, several previous studies have shown that the quality and channel through which information is accessed affects consumer knowledge and decision making. Fernandes et al. (2014) show that just-in-time access to on-line advice improves financial decision-making. Lusardi et al. (2015) show that on-line videos are more effective than standard materials like written disclosures when consumers make choices. Carlin et al. (2016) show that video content is beneficial in helping consumers to both choose better opportunities and avoid falling prey to deceptive advertising in retail financial markets, and are also drivers of social learning in these settings. On the other hand, in some circumstances, it is possible that having more access to information may lead to worse outcomes. As Bernhardt and Cuevas (2016) show, better access to information led to the Felices y Forrados consumer financial fiasco in Chile. The recent rise of FinTech has certainly piqued the interest of the academic community. For example, Buchak et al. (2017) study the effect of FinTech on shadow banking. The authors show that because of different regulatory requirements, FinTech lenders account for the majority of the growth in shadow bank mortgage lending. Philippon (2016) studies the impact of FinTech on the financial industry and how it affects access to financial services. A review of the evolution of FinTech is provided by Arner et al. (2016). The remainder of the paper is organized as follows. In Section 2, we describe the data and offer summary statistics. In Section 3, we explain our identification approach. In Section 4, we report our main results and discuss their robustness. Finally, Section 5 presents concluding remarks.
Watch This : https://www.youtube.com/watch?v=Z5vxRC8dMvs