Measuring Time Inconsistency by Using Bank Account Data

Projekt Start:10/2016
Forscher:Andrej Gill, Florian Hett
Bereich: Financial Intermediation, Experiment Center
Finanziert von:LOEWE

This project was part of the team project "The Dynamics of Finance, Competition and Information Production ".

Topic and Objectives

Present biased preferences are one of the cornerstones of behavioral economics. An important potential consequence of such preferences is that they induce time-inconsistent behavior in intertemporal decision problems. Arguably, one of the most important domains of intertemporal decision making is choosing the structure and financing of consumption patterns. This includes the timing of purchasing particular goods as well as the decision whether, how much, and in which form to save and invest money. From a theoretical perspective, individuals with present biased preferences should behave systematically different in these areas than individuals with “standard” intertemporal preferences that feature exponential discounting. This raises the question about the prevalence and severity of time-inconsistent behavior in this area.

Collaborating in this project with a fintech start-up based in Berlin, we can access detailed financial account data from individual households. The data set is transaction-based, e.g. it lists every transaction that takes place on an individual’s bank account. This allows identifying individuals whose consumption, financing, and savings behavior can be classified as present biased by looking for systematic differences in transaction patterns within a consumption period, e.g. a month. In particular, present bias corresponds to excessive shares of consumption spending in direct response to the reception of income within a short period. Identifying this kind of response to income flows explicitly requires transaction data detailed by area of consumption. Our project relies on exactly such data, therefore, allowing to assess the role of present bias in explaining deviations from optimal intertemporal consumption, financing, and savings behavior.

Key Findings

  • We find that the average household in our sample appears present-biased, as it spends 18.9% more on immediate consumption goods in weeks following paycheck receipt than in other weeks on average. 
  • The extent of the present bias according to this measure, displays substantial heterogeneity across households, with some being severely present-biased and others not at all.
  • Paycheck sensitivity is systematically associated with overdraft usage and the estimated effects are economically meaningful: a change from the 25th to the 75th percentile of paycheck sensitivity increases the probability of using overdrafts on average by 2.7-6.5 percentage points. 
  • Paycheck sensitivity affects overdraft usage both at the extensive and the intensive margin: Not just are stronger paycheck sensitivities associated with a higher probability to use overdrafts at all, but also the volume, frequency, and length of overdraft periods increase the stronger immediate consumption reacts to paycheck receipt.

Policy implications 

The substantial extent of present-biased consumption behavior and its robust association with actual real financial “mistakes” in the form of overdraft usage enhance our understanding of financial and economic decision-making. In particular, it relates to distinct between information/education-based and preference-based (present bias) explanations for financial mistakes: Depending on the relative weight of these plausible mechanisms potentially underlying suboptimal financial behavior, the according policy response might differ widely: While preference-based explanations rather point to policies providing commitment devices to be a promising approach, information-based explanations rather call for educational and regulatory responses. In this sense, our results rather point to the former than the latter.

Related Policy Publications

Andrej Gill,
Florian Hett,
Johannes Tischer
Measuring Time Inconsistency Using Financial Transaction Data
White Paper No. 55