Welfare gains from the introduction of new goods. Hausman, Valuation of New Goods Under Perfect and Imperfect Competition (NBER Volume, 1996)

Transcription

1 Welfare gains from the introduction of new goods Hausman, Valuation of New Goods Under Perfect and Imperfect Competition (NBER Volume, 1996) Suggests a method to compute the value of new goods under perfect and imperfect competition. Looks at the value of a new brand of cereal Apple Cinnamon Cheerios. Intuitively the expected welfare gains are small, because several other brands of Cheerios already exist. Basic idea: Estimate demand; Compute virtual price the price that sets demand to zero; Use the virtual price to compute a welfare measure (essentially integrate under the demand curve); Under imperfect competition need to compute the effect of the new good on prices of other products. This is done by simulating the new equilibrium; Data monthly (weekly) scanner data for RTE cereal in 7 cities. Note: the frequency of the data is higher than before. Also no advertising data.

2 Model Multi-level demand Lowest level : the Almost Ideal Demand System where: s jct is the dollar sales share of total segment expenditure, y gct is overall per capita segment expenditure, B gct is the price index; p kct is the price of the kth brand in market t. Middle level: log-log demand: where: q gct is the quantity of the gth segment in market t; Y Rct is total category (e.g., cereal) expenditure; B kct are the segment price indexes; Top level: demand for the category is where: Q ct is the overall consumption of cereal in market t; I ct is real income; A ct is the price index for the category; Z ct are variables that shift demand;

3 Estimation Done from the bottom level up; IV: for bottom and middle level prices in other cities. Note: the frequency of the data is different from before; uses prices in all other cities not just regional average (large dof might be a problem if IV are weak) Results Elasticities: overall versus conditional Table 5.6 overall elasticities for family segment Welfare Compare to estimates we previously saw: Own-price elasticities are roughly comparable Cross price elasticities: tend to be somewhat larger tend to be less significant several negative (and sig) Raisin Bran s The difference is not driven by the different data sets. Value of AC-Cheerios Under perfect competition approx. $78.1 million per year for the US. Imperfect competition: needs to simulate the world without AC Cheerios assumes Nash Bertrand; ignores effects on competition; finds approx $66.8 million per year; Extrapolates to an overall bias in the CPI 20%-25% bias.

4 Comments Most economists find these numbers too high. Demand estimates raise some issues; In his discussion Bresnahan raised a couple of concerns IV s mainly complains about national advertising Use of Nash Hausman keeps competitors prices unchanged (citing Nash logic ) Hausman s reply was: that the results do not seem to change if advertising is included computing the correct Nash equilibrium also did not matter (not surprising given the small price elasticities)

5 Consumer Welfare Using the Discrete Choice Model assume the indirect utility is Equivalent Variation (EV): As shown by McFadden (1981) and Small and Rosen (1981) for our model where. The mean EV in the population is given by where M is the size of the market. Under the assumption of no wealth effects we can integrate analytically the extreme value distribution of g; in which case the above integral becomes (McFadden 81): With wealth effects the integral has to be computed by simulation. Trajtenberg (JPE, 1989) estimates a (nested) Logit model and uses it to

6 measure the benefits from the introduction of CT scanners. Petrin (JPE, 2003) uses the BLP data to repeat the Trajtenberg exercise for the benefits from the introduction of mini-vans. Petrin finds that the most of the gains are coming from the introduction of another error term. The welfare is measured as the max utility, but adding another product we take the max over more products. He claims that Logit will always over-estimate the welfare effects. He deals with this by trying to reduce the variance of the Logit errors using additional moments. The analysis has 2 steps: 1) Simulate the world without\with minivans, depending on the starting point; 2) Summarize the simulated and observed prices and quantities into welfare. Note: if we observe pre- and post- introduction data we might be able to avoid step 1. This does not allow us to isolate the effect of the introduction, but it does give us a ballpark estimate. I claim that the Logit model fails in the first step but might get the right answer in the second step.

7 Example (red bus/blue bus problem): t=0 t=1 predicted actual option share share share car red bus log(0.5) blue bus log(0.5) welfare log(2) log(3) log(2) In principle, Logit can get the right answer. What fails is the prediction for t =1. This is a well known problem of the Logit model and is related to the other problems we discussed.

8 The Petrin result generated much interest and several solutions have been offered: Berry and Pakes suggest the Pure Hedonic model: drop the Logit term. Nevo (ReStat, 2003) suggests using data to avoid the problem in order to compute price indexes. He points to several assumptions that need to be made. Ackerberg-Rysman (2004) suggest changing the distribution of the Logit error as more products are added. For example by adding a function to the utility. They propose using past changes in the number of products to estimate how the distribution should be changed (i.e., estimate f()). The key assumption is that past changes are similar to a brand introduction that is evaluated.