Why do Consumers not Switch?

An Experimental Investigation of a Search and Switch Model

Irene Maria Buso and and John Hey

Abstract

Search and switch costs are two market frictions that are well known in the literature for preventing people from switching to a new and cheaper provider. Previous experimental literature has studied these two frictions in isolation. However, field evidence show that these two frictions frequently occur together. Recently a theoretical framework has been developed (Wilson, 2012) which studies the interplay between these two costs. We report on an experiment testing this theory to see if individual behaviour with search and switch costs is in line with the theoretical predictions derived from the optimal choice rule of Wilson. The results show that the crucial role of the search strategy: not only, according to Wilson model, the search cost has a greater deterrent impact on search than the switch costs, but also the sub-optimality of the search strategy is the major source of sub-optimality in the switching behaviour.

Keywords: Search Cost, Switch Costs, Choice Under Risk

JEL Codes: D83, D90, D43, C99, G28

The main experiment

This involved 124 subjects and 5 firms.

The average payment to these subjects was £13.50 including the show-up fee. The subjects spent on average 2 hours in the laboratory, including reading the instructions and the payment.

There were 80 problems in total, chosen to give a rich quantity and quality of data. They key parameters of interest were c, the search cost and s, the switch cost. In the main experiment, we had four values of each (c: 0, 0.25, 0.5 and 1; s: 0, 1, 2 and 4) and we implemented all 16 combinations of these (giving the total of 80 problems), each repeated 5 times, and presented in a random order.

The extension experiment

This involved 39 subjects and 10 firms (this latter being the extension).

The average payment to these subjects was £14.50 including the show-up fee. The subjects spent on average 2 hours in the laboratory, including reading the Instructions and the payment.

In the extension experiment, as we wanted to keep the setup of the experiment as close as possible to the main experiment, we kept the range of the offer between 8 and 22, and we have a similar but not identical set of c and s parameter values,(c: 0,0.125,0.25 and 0.5; s: 0,0.5,1 and 2). Indeed, we did not include the extremely high values of the search and switching cost parameters (that is, c = 1 and s= 4) because it would have implied the possibility of negatives payoffs. In this setup the subjects are presented with the same number of different combinations of the parameters and the same number of repetitions of each combination both in the main experiment and in the extension experiment.

The Display Cost treatment

It seems that telling the subjects their accumulated search costs increased the deviations from optimality (perhaps because it it suggests to them not to search at the margin) but the differences are minor.

Experimental Instructions

The instructions (PDF , 954kb)‌ can be found here.

Software

Experimental Software ( 27kb download)‌ written in Z-tree can be found here.

and the questionnaire ( 4kb download)‌ is here.

Matlab Estimatoon programs

Estimation RN July 2019 ( 1kb download)

The Presentation (PDF , 173kb) can be found here.

The Manuscript (PDF , 721kb)can be found here.

Data

You can find the data first treatment (MS Excel , 1,042kb) and data second treatment (MS Excel , 499kb)‌ here.