Prove Me Right?

[Heathman914]

So a certified/coworker and I are having a discussion regarding market condition adjustments.

Example: 0-6 month data set reflects median value of $120,000 and 7-12 month data set reflects a median value of $100,000 or a difference of $20,000. Would you divide the difference of median values of $20,000 by 12 months or 6 months to arrive at a monthly rate of increase?

My opinion would be to divide the difference by the total data set of 12 months. Any other opinions or thoughts would be appreciated. Thank you in advance.

 

[Deleted member 134708]

I wouldn't call any of that a trend. That's only 2 captures of data. As far as I'm reading. You grabbed 500 bulk sales and took median, you grabbed another 500 bulk sales and took median. That's not the best way to go about observing market conditions. That's insufficient data sample to draw any accurate trends.

I do rolling average and median per month. And then group as 3 data sets. 0-6, 7-12, 1-12 So the rolling average and median goes thru the whole year. Then can go from there. You have 12 captures of bulk sales, that are then put into another 3 times frames.

Then you can just "have it your way" and lie all day long with stats. Median increasing too much for you? Use average, 6 months too high? Use 6-12 months. Don't like that, use 0-12 etc.

 

[Heathman914]

I wouldn't call any of that a trend. That's only 2 captures of data. As far as I'm reading. You grabbed 500 bulk sales and took median, you grabbed another 500 bulk sales and took median. That's not correct. That's insufficient data sample to draw any accurate trends.

I do rolling average and median per month. And then group as 3 data sets. 0-6, 7-12, 1-12 So the rolling average and median goes thru the whole year. Then can go from there. You have 12 captures of bulk sales, that are then put into another 3 times frames.

Then you can just "have it your way" and lie all day long with stats. Median increasing too much for you? Use average, 6 months too high? Use 6-12 months. Don't like that, use 0-12 etc.

Thank you for this response, I will certainly look at doing it this way in the future.

 

[Meandering]

With seasonal impacts,
I do it over a minimum of 3 years, quarterly.

That way you're not screwing up a traditional seasonal fluxation as a market trend.

.

 

[Gobears81]

While the reasons highlighted above indicate why better methods are available, if the median price jumps up by $20,000 in six months, you don't conclude that the $20,000 increase is the annual fluctuation. It would be an annualized increase of 44%, two 6-month increases of 20%, or a monthly increase of 3.09%. To calculate the change would be (1+((P1-P0)/P0)^(1/n) Where P1 is new price, P0 is original price, ^ is the exponential sign, and n is number of periods. So the monthly change would be calculated as (1+(($120,000-$100,000)/$100,000)) ^ (1/6)

 

[hastalavista]

So a certified/coworker and I are having a discussion regarding market condition adjustments.

Example: 0-6 month data set reflects median value of $120,000 and 7-12 month data set reflects a median value of $100,000 or a difference of $20,000. Would you divide the difference of median values of $20,000 by 12 months or 6 months to arrive at a monthly rate of increase?

My opinion would be to divide the difference by the total data set of 12 months. Any other opinions or thoughts would be appreciated. Thank you in advance.

It sounds like you are looking at 1004mc data. There is a potential flaw in using it the way you describe.
Unless the data is evenly distributed throughout the period, it can over- or under-state the a trend. Think of it like this:

20 sales occur in the oldest bucket in the first 3 months. Median Price is $100k. But, that reflects the median price from 10.5 months ago (approximately).
20 sales occur in the most recent bucket in the first 3 months. Median price is $100k. But that reflects the median price from 4.5 months ago. The data would indicate there was a $20k change over 6-months.

I see Gobears points out the issue as well.

 

[Elliott]

What's your sample size? What was your set of criteria in your MC selection? Are there other explanations for the change in median values? Making a 20% time adjustment will make you very popular with realtors.

Hard to prove something wrong without knowing more.

 

[TC]

You're a newbie, so I'll let it slide. PROVE

 

[Tom Woolford]

Using a rolling average or median is very useful. It smoothes out the peaks, minimizes the outliers and gives a much more reasonable and comprehensible evaluation of trends. 12 mo is a bare minimum period. 24 months is preferable. All my specific market analysis include a rolling average for sales price, Price per s.f. and DOM.

 

[residentialguy]

I wonder how those comps look if you were to adjust all the variances they had. Not uncommon to have over $20,000 adjustment on a comp. And as Marion said, seasonality. I've never been big on sale price comparisons...too many variables affect the price.