# Thread: Cross-effect of two 'independant' variables - hedonic regression model (real estate)

1. ## Cross-effect of two 'independant' variables - hedonic regression model (real estate)

Hi MHF,

First post here (apart from my welcome post). I apologise if i have missed something basic here, I am not from a stats background, but am studying real estate.I have done a search here but nothing seems to address my precise issues. Any contributions would really be much appreciated.

Ok, Se here goes:

A paper i have been reading (on real estate economics) uses a hedonic regression model to determine the exact relationship between time on the market and price. It states that traditional attempts approach the relationship as follows:

ln(price) = a(property characteristics) + B(market conditions) + Y(time on the market) + E

However this brings an assumptions that market conditions will have no impact on time on market as both variables are considered entirely independantly. Thus an alternative model is suggested as follows:

ln(price) = a(property characteristics) + y(market conditions)(time on market) + B(market conditions) + E

They say that the traditional model gives the coefficient of 0.0220, but following Kennedy (1984) - interpreting dummy variables in semi-logarithmic regressions - they can calculate the change of price (g) for one month of extra TOM using the equation:

g=[exp(B(hat) - 1/2 VAR(B(hat))-1]

and thus g = 2.2%.

This confuses me as my understanding of dummy variables is that they are categorical as opposed to numerical. Does this not use a categorical interpretation method for a numerical variable?

Second problem is that they the go on to interpret the dataset to give multiple TOM (time on market) coefficients (and T stats) for different market conditions (rapid decline through to rapid growth). Is there another type of regression model which can be used to undertake such a regression? I fear i must be missing something obvious because i cannot understand how these multiple coefficients are achieved. I have tried multiplication of the y coefficient by the mean market conditions for each category but this is ineffective as it automatically gives negative values for a declining market whereas their results show that extended TOM still leads to a slightly higher price in a declining market.

finally if anyone can be bothered here is a link to the paper in question, although I really happy to answer all questions on the detail or anything i have missed.