Why do firms compete on price comparison websites? The impact on productivity, profits, and wages

ABSTRACT

A substantial literature indicates that competition on price comparison websites is fierce, leading to lower prices for products sold. As such, we want to answer the key research question: Why do firms compete on price comparison websites? Based on theory, we suggest that participation in these marketplaces leads to increased productivity, i.e., output increases when holding constant the level of inputs used. This, in turn, leads to increased profits, motivating firms to enter price comparison websites despite fierce competition. To find out if theory holds, we empirically investigate how firm entry into a price comparison website affects firm productivity, profits, and wages. Empirically investigating the impact of PriceSpy market participation on productivity, profits, and wages is not easy since firms are free to select whether and when to enter or exit the PriceSpy marketplace, and we use a two-step procedure to address this problem. In the first step, we control for differences in observables between entering firms and potential control-group firms. Then, in a second step, we use a within-firm difference-in-difference estimator on the matched data to investigate how entry into the PriceSpy marketplace affects output while holding inputs constant. Our results indicate that for the full sample of firms, PriceSpy participation increases output by almost 12% when holding the level of inputs constant. Also, an investigation of who gains from the increased productivity shows that, for entering firms, operating profits increase by 9% and gross wages by 14% when studying the full sample of firms. That labor gains more from PriceSpy participation is even clearer when studying the impact on wholesale and retail firms separately. For those firms, wages increased by 16–17% after entry, while no statistically significant impact was found regarding operating profits.

KEYWORDS:

• Online retailing
• e-commerce
• price comparison websites
• productivity
• value added

Acknowledgments

Research funding from the Swedish Retail and Wholesale Council, grant number 2018:773, is gratefully acknowledged. The authors would also like to thank Kenneth Carling, Lena Nerhagen, Siril Yella, and the participants in the Microdata Analysis Seminar (12 June 2020) for their valuable comments and suggestions. We also thank Anton Gidehag for research assistance regarding the CEM modelling.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/09593969.2022.2070773

Notes

1. We use the term ‘excess profits’ to represent all economic profits, i.e., all profits above a normal return on investment given in a competitive market. This separates the concept of economic profits from the operating profits found in annual reports and studied in the empirical part of the paper.

2. A related strand of the literature concerns the impact of price comparison websites on price dispersion. Numerous studies show that considerable price dispersion remains also in market with low search costs such as price comparison websites (e.g., Lach 2002; Baye, Morgan, and Sholten 2004; Haynes and Thompson 2008; Lin, Chen, and Song 2009; Menzio and Trachter 2018; Lindgren, Daunfeldt, and Rudholm 2020).

3. Swedish firms typically use outside carriers such as PostNord, Schenker, or DHL for delivery services.

4. We use a two-year lag to reduce the possibility that any pre-entry adjustments by treated firms might affect the results.

5. To obtain the change in output due to entry into the PriceSpy marketplace in percentage terms, the formula $100\times \left[\mathit{exp}\left({\beta }_6\right)-1\right]$ is used (Wooldridge 2010).

6. We have, however, also estimated a traditional difference-in-difference model using value added as the outcome variable. For all firms, the results indicate an increase in productivity of 8%, while for retail firms it was 13% and for wholesale firms 10%. For firms in other industries, the result was not statistically significant at conventional levels.

7. In Section 4 in the online supplemental material, we also present results of estimating a Cobb–Douglas production function specification. These results are similar to those presented in Table 2, indicating that our results are robust regarding the choice of production function, translog or Cobb–Douglas.

8. A production function model is not an option in this setting since using capital and labor when estimating profits or wages would create severe endogeneity problems.

Additional information

Funding

This work was supported by Handelsrådet [2018:773].