Quantifying search and switching costs in the US auto insurance industry

Abstract

I estimate demand for auto insurance in the presence of two types of market frictions: search and switching costs. I develop an integrated utility-maximizing model in which consumers decide over which and how many companies to search and from which company to purchase. My modelling approach rationalizes observed consideration sets as being the outcomes of consumers' search processes. I find search costs to range from $35 to $170 and average switching costs of $40. Search costs are the most important driver of customer retention and their elimination is the main lever to increase consumer welfare in the auto insurance industry.

Power to Choose? An Analysis of Consumer Inertia in the Residential Electricity Market

“…4 Our paper is also related to a recent literature developing methods to estimate preferences in settings where decision making is influenced by search frictions/ inattention. Hong and Shum (2006), Hortaçsu and Syverson (2004), and Moraga- González and Wildenbeest (2008) are early attempts that utilize aggregate market level data, and more recent efforts by, e.g., Kim, Albuquerque, and Bronnenberg (2011); De los Santos, Hortaçsu, and Wildenbeest (2012); De los Santos, Hortaçsu, and Wildenbeest (2013);Honka (2014); Koulayev (2014);and Honka, Hortaçsu, and Vitorino (2014) utilize detailed consumer level data on both choices and the search process/consideration sets (as obtained from website clicks) to test and estimate models of consumer search. Our empirical setting is one where we observe the choices of consumers but do not observe their search process/consideration sets.…”

Failing to Choose the Best Price: Theory, Evidence, and Policy

“…27 Typically, empirical models include either a switching cost or a search cost to capture inertia, but not both, leading to overestimation of the included cost (Wilson, 2012). Exceptions include two recent papers that utilize rich data 28 to separately identify search costs from switching costs, finding that search costs are both larger and more important sources of inertia than switching costs in the US auto-insurance market (Honka, 2014) and the Chilean pension-fund-administrator market (Luco, 2014).…”

Demand Estimation with Availability Variation

“…In addition to simulation, feasible computation may require numerical techniques to smooth the likelihood function so that it can be optimized. Honka (2014) uses kernel smoothing as an alternative to the prohibitively large number of draws that would be required to avoid lumpiness in the simulated probabilities.…”

“…In structural models of simultaneous search consumers choose a set of products for which they will search for particular information (e.g., prices), based on search costs and expectations of the outcome of the search. The information obtained through search could be prices (e.g., Honka (2014), Seiler (2013), and Mehta, Rajiv, and Srinivasan (2003)), unobserved utility errors (e.g., Moraga-González, Sándor, and Wildenbeest (2012), or both (e.g., Spence (2014), Muir, Seim, and Vitorino (2013), and Pires (2012)). The chosen set of products constitutes the consideration set, from which the consumer then makes a choice after obtaining information.…”

“…The choice probabilities must be corrected to account for this selection. Structural search methods of consideration set formation have been applied to a variety of data structures, including aggregate markets (e.g., Moraga-González, Sándor, and Wildenbeest (2012)), household surveys (e.g., Honka (2014)), household scanner data (e.g., Mehta, Rajiv, and Srinivasan (2003), Seiler (2013), and Pires (2012)), online search records (e.g., Koulayev (2014)), online productrank data (e.g., Kim, Albuquerque, and Bronnenberg (2010)), and regulatory administrative data (e.g., Muir, Seim, and Vitorino (2013)). …”