Interest Rate Changes and Borrower Search Behavior: An Experimental Replication with Extensions

1 Introduction

Financial markets, including mortgage markets and savings products, are essential components of the banking sector and crucial drivers of economic growth worldwide (Valickova et al. 2015). While mortgages generate revenue primarily through interest payments, banks also rely on savings deposits as a funding source, profiting from the spread between deposit and lending rates. These markets are often subject to regulatory oversight, which aims to balance consumer protection with financial stability. Effective regulation can support both objectives, benefiting consumers while allowing banks to maintain profitability (Battiston et al. 2016). Understanding consumer behavior in these financial decisions is crucial for policymakers, banks, and regulators alike.

These institutional aspects are not static. Mortgage and savings markets are shaped by regulatory interventions that affect competition, lending practices, and consumer protection (Benetton 2021; Nicoletti and Zhu 2023; Čermáková et al. 2023). Evaluations of the European Mortgage Credit Directive emphasize its role in strengthening transparency and comparability of mortgage offers across member states, thereby reinforcing the position of consumers in the market (European Commission 2020). At the same time, the rapid expansion of digital finance and fintech challenges existing frameworks and has spurred calls for more adaptive regulatory and supervisory tools (Alaassar et al. 2021; Gai et al. 2018). Taken together, these developments highlight that financial decision-making takes place in a dynamic institutional environment, where consumer behavior is influenced not only by psychological mechanisms but also by evolving regulatory conditions.

This study replicates the findings of Lukas and Nöth (2019), who examined consumers’ decision-making in seeking alternative remortgaging offers. Notably, the existence of search costs plays a role in this process, as finding a new offer requires consumers to invest at least some time and effort. Authors found that consumers are more likely to search for offers when market rates are rising than when they are falling, despite an experimental design that theoretically eliminates this difference. These findings align with prospect theory (Kahneman and Tversky 1979), which suggests that consumers’ strong aversion to losses makes them respond more intensely to potential losses.

Recent research has shown that consumer responses to changing financial conditions are shaped by reference dependence and loss aversion (Lukas and Nöth 2021); however, these effects are highly context dependent. Prospect theory posits that losses weigh more heavily than gains. Still, empirical evidence suggests that this asymmetry is not uniform and can be influenced by framing and task design (Ert and Erev 2013). In addition, reference dependence has been documented beyond monetary outcomes, including decisions involving timing under uncertainty, where the relative salience of losses and gains can even be reversed (Attema and Li 2024). Individual differences also play a role, as risk attitudes, loss aversion, and discounting tendencies systematically vary with demographic and cognitive characteristics (Meissner et al. 2023). Finally, experimental evidence suggests that switching behavior in financial services is heterogeneous across populations and that results obtained from student samples may not always generalize to the broader public (Proestakis et al. 2024). Relatedly, expectation-based loss aversion can generate scale-dependent psychological switching costs, whereby consumers may remain with an inferior default rather than explore better alternatives even when switching is easy and potentially beneficial (Karle et al. 2023). These insights underscore the importance of testing consumer responsiveness in both loss-oriented contexts such as mortgages and gain-oriented contexts such as savings.

Our study extends the framework of Lukas and Nöth (2019) to savings accounts, which represent a case where interest rate changes influence potential earnings rather than costs. Unlike mortgage borrowers, depositors may exhibit different search elasticities when interest rates fall, as they evaluate the trade-off between effort and forgone returns. The experimental design mirrors the original one but applies it to savings decisions. In line with our preregistered design, we formulate the following hypotheses:

2 Materials and Methods

The experiment was conducted in March 2024 at the Laboratory of Experimental Economics (LEE) at the Prague University of Economics and Business, with approval from the Faculty of Economics Ethics Committee, and was preregistered (https://osf.io/sb8r3/). A total of 252 subjects participated in the study. Following the model established by Lukas and Nöth (2019), participants were asked to self-assess their financial literacy, general risk preferences, and financial risk preferences. They then made a decision regarding whether to search for an alternative offer from a mortgage savings account provider.

Cognitive abilities were measured using both numerical and verbal versions of the cognitive reflection test (CRT), with the verbal version included as a control due to the increasing familiarity of the standard numerical version. Age, gender, and field of study served as additional control variables. Overall, the sample can be considered similar to the original paper (Lukas and Nöth 2019). The biggest difference seems to be the higher financial literacy of our sample. Summary statistics are presented in Table 1.

The experimental design was taken from the original paper by Lukas and Nöth (2019). The experiment followed a within-subject design, where each participant completed both the mortgage and the savings conditions. This design enables us to compare behavioral responses to losses and gains within the same individual. It was randomized whether the subject started with the mortgage or the savings scenario. The number of periods per condition was kept identical to the original study. In contrast to Lukas and Nöth (2019), all participants were paid based on the sum of their realized payoffs across all rounds, which should increase their incentives. The average payment was 215 CZK (approximately 10 EUR at the time of the experiment). The experiment was implemented in z-Tree (Fischbacher 2007). Full instructions and used parameters can be found in the Appendix.

Overall, our sample shows a similar composition in terms of gender and CRT test, and higher scores for financial literacy and financial risk preferences.

3 Results

Figure 1 presents the proportion of subjects who chose to seek an alternative offer, comparing their decisions relative to the interest rate from the previous round (Δr _t,t−1) and to the rate from the first (Δr _t,1). In the mortgage context, the results replicate the pattern identified by Lukas and Nöth (2019), showing that as interest rates rise and payment obligations increase, consumers are more likely to search for an alternative offer.

Figure 1:

The rate of contacting the provider for an alternative offer. Note: Bar heights show the share of decisions in which an L2 offer is solicited. Numbers printed above the bars are category counts (n) and do not determine bar height. In total, 1,008 observations are included. The paths of interest rate changes follow the design of Lukas and Nöth (2019) and are provided in the Appendix (Tables A1 and A2).

For savings accounts, the results exhibit a symmetrical and opposite pattern, where decreases in the interest rate lead to reduced interest earnings, prompting consumers to increasingly search for alternative offers.

To better illustrate the dependencies influencing the decision to seek an alternative offer, Figure 2 shows the relationship between this decision and changes in interest rates, focusing on the univariate effects of the difference between current and previous interest rates. They display the probit and logit curves capturing the univariate relationship between the decision to seek an alternative offer and changes in interest rates across rounds.

Figure 2:

Probit and logit models of alternative offer decisions based on interest rate differences. Note: The figure shows the relationship between changes in interest rates and the probability of requesting an alternative offer. Each point represents the observed proportion of subjects making this choice for a given interest rate difference, with fitted probit and logit curves displayed. Numbers printed above the points are category counts (n) and do not determine point position. In total, 1,008 observations are included in each scenario. The exact paths of interest rate changes follow the design of Lukas and Nöth (2019) and are provided in the Appendix (Tables A1 and A2).

Each point on the graphs represents the proportion of participants who chose to request an alternative offer for each given difference value, providing a probability estimate that the probit and logit models aim to capture. The y-axis represents the observed probability of an alternative offer request, while the x-axis depicts the values of these differences.

The curves demonstrate that the core trends of our hypotheses are consistent across both probit and logit models, suggesting a robust directional effect of interest rate changes on consumer behavior. While the general trends align, slight differences between the probit and logit paths underscore model sensitivity and offer insights into variance within the data.

Next, probit regressions are used to examine the factors influencing the decision to seek an alternative offer. Table 2 presents the factors affecting the decision in the mortgage scenario. The effect of a lower interest rate compared to the previous round (Δr _t,t−1<0) is statistically significant at 1 %, with a coefficient value comparable to that reported in the previous literature by Lukas and Nöth (2019). Quantitatively, the marginal effects indicate that when the current interest rate is higher than in the previous round, the probability of searching for an alternative mortgage offer increases by 0.222 in Model 1 and by 0.255 in Model 2. These values correspond to 22.2 and 25.5 % points, respectively. Given the mean of the dependent variable, these effects represent relative increases of approximately 39.4 % in Model 1 and 45.2 % in Model 2. However, our results do not show a significant effect when comparing the first round, as the coefficients remain insignificant even at the 10 % level. A possible explanation is that our sample exhibits higher financial literacy compared to Lukas and Nöth (2019), which may reduce their susceptibility to anchoring on the initial interest rate. Anchoring refers to the tendency to rely heavily on an initial reference point when making decisions (Tversky and Kahneman 1974). If participants in our study were less anchored, their decisions might be more independent of the initial interest rate, leading to the observed non-significance.

The current interest rate, financial literacy, and positive prior experience all show significant effects. In contrast, general and financial risk preferences, along with both the numerical and verbal cognitive reflection test scores, are not significant.

The biggest difference here is the insignificance of the interest rate differential compared to the first round. There is no evidence of anchoring compared to the first round. In our case, financial literacy is then significant, which is higher on average in our data. In contrast, CRT is significant in Lukas and Nöth (2019).

Similarly, probit regressions were conducted for the savings account scenario. It is important to note that the main variables, Δr _t,t−1 and Δr _t,1, take the value of 1 when the consumer is at a disadvantage, specifically when the interest rate has declined. The results are statistically significant when the interest rate falls compared to the previous round (Δr _t,t−1<0). Quantitatively, the marginal effects show that when the current interest rate is lower than in the previous round, the probability of searching for an alternative savings offer increases by 0.309 in Model 5 and by 0.337 in Model 6. These correspond to 30.9 and 33.7 % points, respectively. With the dependent-variable mean at 0.601, these effects imply relative increases of approximately 51.4 % in Model 5 and 56.1 % in Model 6. Additionally, the marginal effects in the probit regression are larger for savings accounts than for mortgages, indicating that consumers are even more responsive to changes in this context.

As with the mortgage scenario, the results remain statistically insignificant when compared to the first round. The current interest rate shows significance in relation to the previous round, though the previous round comparison itself is not significant. Financial literacy, general risk preferences, and both numerical and verbal cognitive reflection test scores are insignificant. In contrast, financial risk preferences are weakly significant, and positive past experience is strongly significant (Table 3).

Considering that individuals were exposed to both loan and savings settings, we also control for this factor. While order effects appear in the savings scenario, where subjects request alternative offers more often when encountering it later, no such effect is found for loans. Importantly, this does not affect the statistical significance of other variables, as all retain their significance and comparable marginal effects.

4 Discussion and Conclusions

This study successfully replicates Lukas and Nöth’s (2019) findings on the effect of interest rate changes on borrower behavior, particularly in the decision to seek alternative offers. Our results confirm that consumers exhibit heightened sensitivity to rising interest rates in a mortgage context, prompting them to search for competitive offers, as anticipated.

However, several differences emerged in the results of the current experiment. Increased searching behavior is confirmed when conditions worsen compared to the previous round, as is a significant positive effect of prior positive experience with alternative offers. In contrast, there is no observed effect of interest rate deterioration compared to the first round, and some differences appear in the significance of control variables.

Extending the original study, we introduced a savings account context to explore whether consumer responses align symmetrically in a gain-oriented financial setting. The results reveal that consumers are similarly sensitive to decreases in interest rates, actively seeking alternative offers when potential interest earnings diminish. This behavior supports a broader application of prospect theory (Kahneman and Tversky 1979), particularly loss aversion, as consumers respond comparably to losses or reduced gains, whether they are in debt or savings situations.

With clear evidence that consumers respond to interest rate changes, institutions might consider structuring product offerings to anticipate and address this sensitivity, especially in volatile rate environments. Banks could achieve this by offering products with interest rate guarantees, step-up or step-down interest options, or loyalty incentives for customers who stay despite rate fluctuations. Additionally, they may enhance digital tools that allow customers to track and compare interest rates in real time, helping them make informed decisions while fostering customer retention.