A good question was asked about why I said that this might not be that statistically significant (note not saying not statistically significant, but not that significant). I explain my reasoning below the quote, but if to any who hate math, you might just want to ignore my post
.
Hi Bob,
Using your figures, last year's sales were 77% of the 5 year average you mention.
The average sales price is 72% of the 5 year average.
That seems statistically significant to me, however I am here to learn from others so what type of drop would you need to see to find it significant?
Thanks
So let's start with the data for sales of .me domain names. NameBio nicely compute for us both the mean (average) and standard deviation data. Here are the values:
Last 12 months: 134 sales $868 average price StDev=$1399
Last 5 years: 870 sales $1196 average price StDev=$2359
The fact the standard deviations are so high compared to average is simply telling us that while most of the sales are for lowish prices, there are some high ones, so a lot of scatter.
To compare two mean values for whether they are statistically different we use something called the
Standard Deviation Of the Mean (SDOM). It is obtained by taking the standard deviation which NameBio gave us and dividing by the square root of the number of observations that went into computation of that mean. For the last year period, that gives us $1399/11.2 = $121.
For the 5 year period you might think we would use the average number per year (174) but really the average is based on the total number for the 5 years, so we use the square root of the 870. When we do this we get a SDOM of $80.
No domain price data is certainly normally distributed in the statistical sense, but as a first approximation and lacking any better assumption, we will assume that it is. A commonly used significance test is that corresponding to about 95% likelihood (19 times out of 20 often see terminology in poll results).
For this the one year price will be between $868 (+/-) (2*121) or in other words we have about 95% chance that the actual value is between $626 and $1110.
Now looking at the
5 yr example, we have about a 95% chance that the actual value is between $1036 and $1356 (obtained from the 5 yr average and +/-2 times the $80 SDOM for this group).
Now the
critical thing is that the ranges overlap, i.e. the lowest possibility for the higher 5 yr average is lower than the higher possibility for the one year data. This tells us that
while it may well be true that .me sales prices are going down, it is to me not very statistically proven.
We could do something sort of similar with the number of sales. Here we often take the square root of a number as indicative of the uncertainty - i.e. 100 would be uncertain by 10, or 10% while 400 would be uncertain by 20 or 5%. If we do that here the 134 per year over the last year is probably between about 111 and 157. If we just take the 174 as the number per year it suggests it could be tween about 148 and 200 roughly. Near to significance but not quite. Really the way I treated the 5 yr over estimates uncertainty so it is almost on edge of significance.
We can convince ourselves that there is quite a bit of uncertainty by looking at the data. For example in the last 12 months there are 134 sales. If I eliminate just 7, the highest value 7, about 1/3 of the entire dollar volume is gone.
But even if the above had supported marginal statistical significance
we should be careful of systematic biases in the data. Let me indicate a couple.
(a)
Which venues report to NameBio has changed over the years. For example
in the early part of the 5 yr period Afternic reported, but now they do not. Sure enough in the early years of the 5 yr period there are 10 .me sales on Afternic and they are all between $1000 and $3500 per sale, and all are in the first 2.5 yr of the 5 year period. I am not sure why by Dynadot are not in the early period, but they certainly are recently. There are 18 Dynadot sales of .me all in the last year and a half and only one of which is as much as $1000 (supposedly because much more domainer to domainer on that platform).
(b) Even if we did not have the issues of biases introduced by different venues in different years, we might have the situation that
the average price is going down with time but that is just because the best names were all sold early on, and similar names are actually going up in price. In the overall .com I am pretty sure this is happening, maybe not with .me but it is at least a possibility.
One way to check (a) is to look at only sales data from a venue that has reported throughout the sales period and that is mainly retail sales. Sedo is the best option. They do seem to support .me going down in price as the
5 year average is $3026 and the one year is $2550 although if you use the SDOM the difference is not statistically significant. In fact, if I look at only Sedo data and instead of using mean
use the median (i..e. order the sales and take the one in the middle) the median .me sales price for
last year is $2350 while the median for the 5 year period is lower at $2000.
So I would find it
very possible that prices for .me are edging down, and number of sales too, but it is not dramatic and possibly simply effects from reporting venues altering the wholesale/retail mix.
Sorry this has been so long! I hope you don't get banned from NPs for using math!
Bob
ps By the way if you thought I had gone through all of this before I said it was not that statistically significant, I had not. I am way too lazy for that
. I had a hunch the data with a sample of just a bit over 100 would work out overlapped.