When to haggle, when to hold firm? Lessons from the used‐car retail market

• Published date: 01 July 2020
• Date: 01 July 2020
• Author: Guofang Huang
• DOI: http://doi.org/10.1111/jems.12385

J Econ Manage Strat. 2020;29:579–604. wileyonlinelibrary.com/journal/jems © 2020 Wiley Periodicals LLC

|

579

Received: 5 December 2017

|

Revised: 13 February 2020

|

Accepted: 25 May 2020

DOI: 10.1111/jems.12385

ORIGINAL ARTICLE

When to haggle, when to hold firm? Lessons from the

used‐car retail market

Guofang Huang

Krannert School of Management, Purdue

University, West Lafayette, Indiana

Correspondence

Guofang Huang, Krannert School of

Management, Purdue University,

403 W. State Street, West Lafayette,

IN 47907.

Email: huan1259@purdue.edu

Abstract

Though haggling has been the conventional way for auto retailers to sell cars,

the last two decades have witnessed the systematic adoption of no‐haggle

prices by many large dealerships, including the largest new‐and used‐car

dealership chains. This paper develops a structural empirical model to estimate

sellers' profits under posted price and haggling, and investigates how market

conditions affect sellers' optimal pricing formats. The model incorporates a

simple class of bargaining mechanisms into a standard random‐coefficient

discrete‐choice model. With the extension, the product‐level demand system is

estimated using data with only list prices, and the unobserved price discounts

are also recovered in the estimation. The counterfactual experiments yield a

few interesting findings. First, dealers' adopted pricing formats seem superior

to the alternative ones. Second, dealers enjoying larger market power through

vertical differentiation and carrying a large number of models are more likely

to have posted price as their optimal pricing format.

1|INTRODUCTION

In the auto retail market, negotiation, or ‘haggling’, is the traditional way to reach a transaction price. The

conventional wisdom is that negotiation allows dealers to reap higher profits from consumers not interested in or

not good at negotiation and, at the same time, to get more business from the price‐sensitive buyers. Yet, starting in

the early 1990s, some dealerships began to adopt haggle‐free posted prices. For example, CarMax, founded by

Circuit City in 1993, sells mainly used cars at no‐haggle prices. AutoNation, the largest new‐carretailerinthe

United States, started experimenting with no‐haggle pricing in its Denver stores in early 1999, and has recently

adopted no‐haggle pricing throughout its stores. Many other large auto dealership chains have also converted to no‐

haggle pricing.

Why do some dealerships choose haggling while others choose posted prices? Are there conditions in which

onetypeofpricingformatwouldbefavored?Thispaperseeks to answer these important questions related to

pricing strategy in the car market. To address these questions, I estimate a structural model of demand and

supply using a new data set of car sales at five Baltimore‐area dealerships. The model explicitly incorporates

sellers' pricing formats. I then use the estimated structural model to simulate and compare sellers' profits under

various pricing policies, and investigate the connection of the results with the sellers' competitive positions in the

market.

The above research strategy relies on two key steps. The first involves the ability to estimate the actual profits for all

dealerships. The second one is a consistently estimated demand system, from which the profits under alternative

pricing policies can be simulated. The main difficulty with implementing this approach is that I observe only the list

prices for cars and not the actual transaction prices, which is a common problem with aggregate‐level data in many

markets.

1

With the transaction price data missing, a direct application of the random‐coefficient discrete‐choice model

developed by Berry et al. (1995; hereafter referred to as BLP) would lead to inconsistent estimates, especially of the price

coefficient and actual profits. This is because it needs a more inelastic demand to justify the observed sales with the

observed list prices when some consumers actually were able to obtain discounts through negotiation. By explicitly

incorporating seller's pricing formats into the standard BLP model, I am able to correct the estimation bias due to the

missing data problem and recover the haggling dealers' unobserved price‐negotiation rules and, thus, the actual profits.

2

The data that I use in my estimation are constructed from the daily listing data from August 2008 to August 2009 of

five dealerships in a suburb of Baltimore, MD. I scraped the daily listing data from Cars.com, a leading automobile

classified website. Two of the five dealerships use the no‐haggle pricing policy and the other three haggle. The listing

data include the list prices and detailed characteristics for all cars in the inventory of the five dealerships during the

period. I infer the sale date of a car as the date when the car's listing disappeared from the website,

3

and compute the

market shares by the car model and dealership for each quarter.

For estimation, I follow BLP and Petrin (2002) in using the Simulated Method of Moments to estimate the para-

meters in the structural demand model. My proposed price‐negotiation process has two dealer‐specific structural

elements: the negotiation cost and price discount. The assumption I use to identify these additional parameters in the

structural model is that there is no unobserved heterogeneity across dealerships other than the unobserved discounts

and negotiation costs.

4

This assumption together with the existence of nonhaggling dealers in my data allows the

identification of the parameters in the negotiation process and in the buyer's utility function. Intuitively, the variations

in the market shares across car models sold by nonhaggling dealers identify the standard parameters in the buyer's

utility function; and, for a given model and the observed list prices, the difference in the market shares between the

haggling and nonhaggling dealers help identify the parameters in the price‐negotiation processes for the haggling

dealers.

Overall, the estimates show good face validity. The estimated unobserved discounts are of similar magnitudes to

those found in previous field experiments and in the actual buying experiences at the haggling dealerships. For the

medium‐sized dealerships in my data, the estimates of their price–cost margins are close to the market average

price–cost margins that I derived using external sources of information. From the modeling perspective, my estimates

show that ignoring sellers' actual pricing policies and using list prices as the transaction prices in the estimation would

lead to significantly smaller magnitude for the estimates of the price coefficient and price elasticities.

For the empirical investigations of this paper, I find that, first, the relative performance of posted price and haggling

varies across dealerships. On the one hand, the dealerships using posted prices would see their profits increase only

slightly if they haggled, and the increase in profit could disappear if the haggling cost is accounted for. One the other

hand, the haggling dealerships' profits would drop significantly if they switched to posting prices. Second, I find that

market power derived from two sources, the vertical differentiation in qualities and the variety of models carried by a

dealership, has significant impact on dealerships' choices of pricing policies. In particular, sellers selling used cars with

rare qualities and/or carrying a large number of models are more likely to find posted price as their optimal pricing

format.

This paper is first related to the literature of pricing strategy in the auto retail market (e.g., Chen, Yang, &

Zhao, 2008; D'Haultfoeuille, Durrmeyer, & Février, 2019; Scott Morton, Zettelmeyer, & Silva‐Risso, 2001,2003;

Sudhir, 2001; Zettelmeyer, Scott Morton, & Silva‐Risso, 2006; and references thereon). This paper is distinguished from

the papers in this literature by its substantive focus on sellers' pricing format choices. In addition, this paper is also

different in its modeling approach that incorporates a simple negotiation structure into an aggregate demand system for

all the used cars in the market.

A closely related paper from the literature is D'Haultfoeuille et al. (2019), which faces the same methodological

challenge caused by the deviations of transaction prices from observed list prices when applying the BLP framework in

estimating the demand for new cars. Their paper obtains identification by exploiting supply‐side optimality conditions

and focuses on showing the extent to which the price discounts vary by consumer characteristics. In contrast, my

identification approach exploits the existence of nonhaggling dealers in the market, and my substantive focus is on how

the relative performance of posted price and haggling for individual dealers depends on their competitive positions in

the local market.

This paper also relates to the empirical literature on bargaining and price discrimination, especially in the

automobile market (e.g., Ayres & Siegelman, 1995; Goldberg, 1996; Larsen, 2018; Scott Morton, Silva‐Risso, &

Zettelmeyer, 2011). Papers in this literature show significant impact of bargaining on transaction prices and market

580

|

HUANG