Bmw Sustainability

[pic 1][pic 2][pic 3][pic 4]

[pic 5][pic 6]

Question 1

• gen ln_price = ln(price)
• gen dist_mile = dist/5280
• gen proximity = (dist_mile<3)
• reg ln_price proximity if year == 1981, r

[pic 7]

There are two conditions under which the above stated regression would be a good estimator for the impact of incinerators construction on prices:

• (C1) Before the contract was signed (e.g. in 1978) there was no difference in prices of houses near the future construction sites and of houses far from them
• (C2) The proximity to the construction site is the only source of difference in prices (i.e. price of house does not depend on house age, size, etc.)

If we believed these two assumptions to be true (which is obviously not the case), we could conclude that after signing the contract houses near to construction sites became 33% cheaper than houses far from construction sites^[1].

Question 2

• reg ln_price proximity if year == 1978, r 

[pic 8] 

The results of this regression prove that our assumption regarding price differences before signing of the contract, made in the first part, is wrong (i.e. prices of houses near to and far from construction sites differed even before the official approval of the project). Thus, the main conclusion we can derive from here is that our previous estimation of the impact on prices (from question 1) is wrong.

The presence of gap in prices in 1978 suggests that we have to use difference-in-difference approach in order to have an accurate estimation of the impact.

Question 3

• gen post = (year == 1981)
• gen inter = post*proximity
• reg ln_price proximity post inter, r

[pic 9]

The results of the regression are summarized in the table above and Exhibits 1 and 2 in the Appendix.

As we can see, the effect of the construction of incinerators is smaller than one we got before (houses near construction sites are on average cheaper than those far from construction sites by 6%^[2]).

There are 3 assumptions that we should make in order to use this model:

1. If the contract was not signed in 1981, the prices (logarithms of prices) for houses near to construction sites would grow by the same amount as prices for houses far from construction sites grew by. (parallel trend assumption)
2. There were no changes in other factors of the house prices between 1978 and 1981
3. The proximity of houses to construction sites is the only source of price difference

While the first 2 assumptions seem reasonable, the last one is most likely to be wrong (many factors that determine prices for real estate are omitted: age of building, its size, surrounding infrastructure, etc.). (this serves as part of the answer for the question 4)

The conclusions that we can make now are:

• We cannot use the current estimation for the impact. Control variables should be included in the current model to get more accurate result.
• We do not reject DiD approach because of the statistically insignificant coefficient for the interaction variable at this stage of the analysis. Not only coefficient’s value, but also the level of its significance may change once we add control variables.

Question 4

• gen age2 = age^2
• gen dist_ln = ln(dist)
• gen land_ln = ln(land)
• gen area_ln = ln(area)
• gen cbd_ln = ln(cbd)
• reg ln_price proximity post inter age age2 dist_ln land_ln area_ln rooms baths cbd_ln, r

[pic 10]

As we observe from the results of the regression:

• The coefficient for the interaction variable is now statistically significant and its value changed to -0.13, which represents the difference in prices (near and far from construction sites) of 12.2%.
• Coefficients of most of the control variables are aligned with our anticipations regarding their sign (the more rooms a house has, the more it is expensive; the more the area of the house, the more expensive it is, etc.)

[pic 11][pic 12][pic 13][pic 14][pic 15][pic 16][pic 17]

[pic 18]