Background
This assignment is based on the paper “From Mad Men to Maths Men: Concentration and Buyer Power in Online Advertising” (2021) by Francesco Decarolis and Gabriele Rovigatti published in the American Economic Review. The authors motivate their analysis as follows:
Online advertising sales are the main fuel of all of the major digital platforms. In the in- ternet era, advertising means capturing the attention of consumers who are browsing the web and this requires both detailed data to effectively target the ad to the right customers and algorithms to bid in the online auctions where ad space is sold. These needs have led to a major, but understudied, shift in the industry: rather than bidding individually, advertisers increasingly delegate their bidding to highly specialized intermediaries. This concentration of demand within a few large intermediaries raises the question of coun- tervailing buyer power. Can the emergence of intermediaries counterbalance the highly concentrated supply of online ads? (pages 3299-3300)
The authors establish causality by leveraging a novel data set, natural language processing, and instrumental variables.
1By “run”, we mean that the TA can click “do” in the do-file editor and the whole do-file runs through and produces the desired results and produces all the requested output.
The third ingredient is an instrumental variable (IV) strategy. Instruments are needed for two reasons: measurement error in the proxy for demand concentration and potential omitted variable bias. For instance, there might be unobservable shocks to the popularity of some keywords that drive changes in both revenue and demand concentration. Similar to Dafny, Duggan, and Ramanarayanan (2012), we address this problem by exploiting the variation in intermediary concentration driven by changes in network ownership of MAs. In our sample period, there were 21 acquisitions and 2 divestments, affecting 6 out of the 7 agency networks. These merger and acquisition (MA) operations, especially the larger ones involving a multiplicity of markets, are a useful source of variation in demand concentration as the revenue dynamics in each local market are too small by themselves to cause the MA operations. We extensively discuss this empirical strategy and evaluate its robustness. (page 3301)
Now your task is to replicate some simple aspects of the author’s study and try to understand how IV helps to resolve the measurement error and omitted variable bias problems.
Questions
1. The dataset analysis1.dta2 contains the main data used in the paper. The outcome of interest is Google’s estimated log revenue logr hat (measured in log dollars). This variable is an estimate for Google’s revenue from advertising on searches with a specific cluster of keywords. Think of a cluster of keywords as a “market” or industry.
The covariate of interest is a Herfindahl-Hirschman index (HHI) HHI hat. This variable measures concentration on the demand side of the market. Firms want to advertise their products beside Google search results. If these firms are competitive, then HHI is low. If firms are oligopolistic or not competitive, then HHI is high. HHI ranges from 0 to 1.
Merger and acquisition operations mean that firms combine or break apart. When this happens, the market concentration changes. The instrument variable is sim, the simulated change in market concentration (HHI) that occurs from merger and acquisition operations.
Load the dataset into STATA, get an overview over the dataset (STATA: describe and sum).
2. Create a scatter plot showing the correlation between HHI hat and sim. Include a linear predic- tion in your graph. Explain the intuition of this figure.
3. Now consider the model:
logr hatmt = α + βHHI hatmt + ϵmt, (1) where m indexes markets and t indexes years.
2Decarolis, Francesco, and Rovigatti, Gabriele. Data and Code for: From Mad Men to Maths Men: Con- centration and Buyer Power in Online Advertising. Nashville, TN: American Economic Association [publisher], 2021. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2021-09-29. https://doi.org/10.3886/E130502V1
(a) Run an OLS regression of revenue in the market-year on market concentration (HHI) using robust standard errors.
(b) Interpret the coefficient β. Is it statistically significant? Is it economically meaningful? (Answer in max. 2-3 sentences)
(c) According to the paper, what is the potential problem that could cause OLS to be incon- sistent? (Hint: Read the Identification Strategy section on page 3316.) (Answer in max. 2-3 sentences)
4. As explained before, the authors use mergers and acquisitions as an instrument for changes in market concentration.
(a) Run TSLS manually (i.e., run the first stage, and then run the second stage, without using the automated IV command) and compare the TSLS estimate to the OLS estimate.
(b) Run TSLS using the automated STATA command ivregress 2sls.
(c) According to the authors, what is the reasoning for the validity of this instrument? (Hint: Read the Identification Strategy section on page 3317.)
(d) Should we worry about weak instruments in this application? Conduct a formal test.
(e) The authors consider the more general model:
CS 340 Milestone One Guidelines and Rubric Overview: For this assignment, you will implement the fundamental operations of create, read, update,
Retail Transaction Programming Project Project Requirements: Develop a program to emulate a purchase transaction at a retail store. This
7COM1028 Secure Systems Programming Referral Coursework: Secure
Create a GUI program that:Accepts the following from a user:Item NameItem QuantityItem PriceAllows the user to create a file to store the sales receip
CS 340 Final Project Guidelines and Rubric Overview The final project will encompass developing a web service using a software stack and impleme
