Monday, July 29, 2019

Boston Housing Market vs. Condo Market – Who Wins?

-- Intended for New College Graduates --

(Click on the image to enlarge)

Paula, a new college graduate with a major in Economics, is interviewing for a Research Analyst position. 

Question # 1
Interviewer: Take a look at the line graph and tell me why the monthly data points have been connected.

Paula: Because these are seasonally-adjusted monthly data points. When the monthly data points are seasonally adjusted, month-over-month comparison is fine, so a line graph makes sense.

Question # 2
Interviewer: If the data points were not seasonally adjusted, how would have presented and analyzed the data?

Paula: Via mutually exclusive bars and I would have compared the April, 2019 data with April, 2018 and April, 2017 data, respectively.  

Question # 3
Interviewer: Do you think the line graph should have comprised Y1 and Y2, rather than just one Y?

Paula: It's fine the way it is. You need Y1 and Y2 when the comparative values are too far apart. That's not the case here.

Question # 4
Interviewer: Study this scatter plot and explain to me the interactions between these market segments.

Paula: They are highly correlated and, therefore, are moving in tandem. 

Interviewer: So, can you make an universal generalization that Boston single family housing market is highly correlated with Boston condo market?

Paula: No. I meant during the period the data points are drawn from. Market segments within a broad group can easily diverge.

Question # 5
Interviewer: By looking at the scatter, can you tell me if the correlation coefficient would be higher or lower than this R-squared?

Paula: It would be higher because the linear trendline generating the R-squared is not the optimal fit here.

(The actual correlation coefficient is 0.9896)

Question # 6
Interviewer: If this linear trendline is not giving us an optimal fit, what trendline would you prefer in this case?

Paula: Given the distribution of the data, a polynomial trendline (of 4th/5th order) would produce much better results. 

Question # 7
Interviewer: How would that be an improved trendline? What is it that the linear trendline is not capturing that the polynomial would?

Paula: The linear trendline is not capturing the data points on either end of the curve. The proposed polynomial will take care of that inadequacy.

Question # 8
Interviewer: Let's assume that the software you are using does not allow experimenting with any other trendlines. So, holding the linear trendline constant, can you still improve upon the current stats?

Paula: Yes, by removing the outliers from the series. In this case, if we simply remove the three outlier data points from the two ends of the curve, the R-squared would jump.

Question # 9
Interviewer: Let's assume we transpose X and Y axes, meaning we graph housing values on Y and condo values on X. Will that change the R-squared? Finally, is there a winner here?

Paula: No. Swapping axes in this instance will not make any difference in R-squared. It will remain the same. As far as the winner is concerned, given the very high co-linearity, it's a tie.

Disclaimer - The author is not advocating the Case Shiller indices listed here. Consult your Financial Planner for an appropriate asset allocation model and/or trading strategies for different markets, including housing.

Thank you.

Sid Som, MBA, MIM
President, Homequant, Inc.
homequant@gmail.com
   
Coming soon: Sid's New Book: Modern Interviewing Techniques and Skills - Live Simulations with actual Market Data

No comments:

Post a Comment