BLP data and models

• I am first uploading the data from an R package called hdm.

require("hdm")

## Loading required package: hdm

data('BLP')

• I preview the data.

data=BLP$BLP
head(data)

• The variable names are:

names(data)

##  [1] "model.name" "model.id"   "firm.id"    "cdid"       "id"        
##  [6] "price"      "mpd"        "air"        "mpg"        "space"     
## [11] "hpwt"       "trend"      "share"      "outshr"     "y"

• I extract and preview the data for year 1990.

data1990=data[(2217-130):2217,]
head(data1990)

• In BLP Econometrica p. 871: The least expensive car in the sample is the 1990 Yugo at $3393 (1983 dollars). I check this information.

x=data1990$price[which.min(data1990$price)]
x

## [1] -8.368152

• This looks in log but I cannot recover this price.
• We need to find the prices elsewhere (see below).

Check the data.

• BLP have a Table II p. 870.

• They compute weighted means.
• The variable Size is called space here.

round(weighted.mean(data1990$hpwt, data1990$share),3) # Hp/wt

## [1] 0.419

round(weighted.mean(data1990$space, data1990$share),3) # Size

## [1] 1.27

round(weighted.mean(data1990$air, data1990$share),3) # Air

## [1] 0.308

round(weighted.mean(data1990$mpg, data1990$share),3) # MPG

## [1] 2.27

round(weighted.mean(data1990$mpd, data1990$share),3) # MP$

## [1] 2.852

• They coincide perfectly

Prices

• I have another data set.

require(BLPestimatoR)

## Loading required package: BLPestimatoR

dataB=productData_cars
dataB1990=dataB[(2217-130):2217,]
names(dataB1990)

##  [1] "firmid" "cdid"   "id"     "const"  "price"  "hpwt"   "air"    "mpg"   
##  [9] "space"  "share"

• I will check the means

round(weighted.mean(dataB1990$price, dataB1990$share),3) # price

## [1] 10.337

• It also coincides perfectly.

• I will use the price as the correct one.

data1990$price=dataB1990$price

• I rename the variables

names(data1990)[10]="Size"

write.csv(data1990,"Newdata1990.csv")

Checking the estimates

data1=read.csv("BLPdata.csv")
data2=read.csv("BLPdata2.csv")
data=cbind(data1,data2)
data=data[,-c(1:4,6:7,13,15:16,17:21,23:27)]
data$cdid=data$cdid+1970
names(data)[1]="year"

calculating the endogenous variable

We calculate

s0=as.data.frame(cbind(1971:1990,1-tapply(data$share, data$year,sum)))
names(s0)=c("year","s0")

data=merge(data, s0, by.x="year")
data$y=log(data$share)-log(data$s0)
names(data)

##  [1] "year"  "mpd"   "air"   "mpg"   "space" "hpwt"  "share" "price" "s0"   
## [10] "y"

lm(y~hpwt+air
   +mpd+space+price
   ,data= data)

## 
## Call:
## lm(formula = y ~ hpwt + air + mpd + space + price, data = data)
## 
## Coefficients:
## (Intercept)         hpwt          air          mpd        space        price  
##   -10.07159     -0.12431     -0.03434      0.26502      2.34209     -0.08864

• BLP obtain this