Several months ago Stewart Fast, a new professor at Queens University in Kingston, Ontario, undertook a study of why southern Ontario was such a hotbed of anti wind energy sentiments. His conclusions were interesting, and I’ll be having more to say about them in a future posting. As part of his study he looked at property values and in particular he looked at MPAC (the Ontario real estate assessors), Wolfe Island and the property assessment reductions thereon.
As it happens, I had also looked at MPAC and Wolfe Island and posted on it about 18 months ago. It seems that Fast and I used the same FOIA-obtained spreadsheet. My main conclusion was that there seemed to be a large number of large reductions on Wolfe Island, but there wasn’t enough of a pattern to convincingly tie the reductions to the 86 wind turbines on Wolfe’s west end.
I’ve also posted on MPAC and property assessments in a 4-part series. My main conclusion, contained in part 1’s section, was that MPAC seemed to be hiding the reductions by lowering the values in neighborhoods that just coincidentally happened to be around wind turbines, but not formally incorporating distance to a wind turbine into their regressions.
What Dr. Fast’s work added to mine was that (1) he was able to group MPAC’s reductions on Wolfe Island by their distance to the nearest wind turbine, and (2) he reminded me of how to use chi-square to test the differences between the bands for statistical significance. The quick summary is that MPAC has been providing reductions to properties close to wind turbines significantly more often that those further away. And I’m not using the word “significantly” in some fuzzy qualitative manner – I mean “significantly” in the hard statistical quantitative manner. In other words, the odds of the getting a wind-turbine-centered pattern just randomly are vanishingly small. Wolfe Island provides a good hard-to-refute example of how MPAC is finessing the numbers to deny the obvious.
The raw data (i.e. the spreadsheet) is quite detailed, so to save space here’s the summary of it. There are 4 major areas in the municipality of Frontenac Islands: Wolfe West (where the turbines are), Wolfe East, Howe Island and Simcoe Island. The number of properties and total reductions are in the following table.
As both Fast and I have written, these numbers aren’t really indicative of anything having to do with wind turbine proximity. About the only thing that stands out is that Simcoe Island had a far higher rate than the other areas, which was at least partially due to reasons other than wind turbines.
The next step was what Fast added: he was able to use GIS software to group the reductions into buffers based on the distance to the nearest wind turbine. He had 5 buffers: < 1km, 1 – 2 km, 2 – 5 km, 5 – 10 km and > 10 km. He used chi-square to see if there were significant differences between the buffers and found that the 1 – 2 km and 2 – 5 km buffers were significantly more likely to have reductions than the other buffers. As he said, this may be suggestive but is not quite conclusive.
Dr. Fast graciously provided me the data that went into his buffers and I re-ran his chi-square calculations to make sure I could replicate his results. Initially I thought the non-significance of the < 1 km buffer (you’d expect the buffer closest to the turbines to show the most significant effect) was due to the income-producing nature of any land close to a wind turbine, plus the setback that rendered about 25% of that buffer unoccupied. While those could be important, I also noticed that by chance there were a lot of reductions just outside of the 1 km border. As an example I show the following picture of the reductions around Wolfe’s main city, Marysville:
The 1 km buffer ends about where Highway 95 T’s: to the left is inside that buffer while to the right is in the 1 – 2 km buffer. Since all the buffer borders are fairly arbitrary anyway, I decided to proceed with 4 buffers, with my results below.
As the buffers get closer to the wind turbines you can see that the ratios of reductions to properties generally gets higher. The chi-square is a test to see if these ratios could simply be due to chance. The “Chi-2 p” column provides the p (probability) that this buffer varies from the total Frontenac Islands ratio by random chance. The two buffers closest to the wind turbines have less than a 1% chance of having values that high by chance, while the buffer farthest from the wind turbines has a much less than 1% chance of having a ratio that low by chance. Note that BOTH close-in and far-away reductions are significantly different from the mean, and in directions that are BOTH consistent with the hypothesis that wind turbines are associated with property assessment reductions.
In my earlier posting on the MPAC 2012 study I predicted that they would lower assessments close to wind turbines while never explicitly recognizing wind turbines as the cause. I offered up Wolfe Island as an example of how this might proceed. Thanks to Dr. Fast and a fair amount of serendipity we now have a solid indication that MPAC is proceeding as predicted. While some aberrations in assessments and reductions would be expected (some chi-squares show significance where none really exists), the pattern shown above is just too consistent to be cast aside as coincidental or anecdotal. We proposed a hypothesis that there would be more reductions closer to the wind turbines and the data clearly support that hypothesis.
Dr. Fast does have a point that this is indirect evidence of lower values. After all, we don’t have the municipality’s “bufferized” assessed values, not to mention bufferized sales data. In the 2012 study, MPAC did provide bufferized assessments for the entire province and they show a 25% decrease within 5 km of the turbines, a result that somehow got lost in their summary. As for actual sales, there have been so few, especially on Wolfe’s west end, that any sort of statistically-valid testing would be difficult. In the meantime, reductions will have to serve. That MPAC seems to be going out of its way to hide this trend indicates that MPAC is being used to implement a political agenda – a problem greater than just wind turbine assessments.
Overall, some 22% of the properties in the Frontenac Islands were granted reductions by MPAC. I’d love to know if that is typical – it would be interesting to study the assessments and reductions in the municipalities with and without wind turbine projects. Unfortunately, as even Dr. Fast commented, MPAC has made getting their data just about impossible.
In his report Fast says that the evidence of reductions due to wind turbine proximity is “suggestive but not conclusive”. Given the numbers above you are of course free to come to your own conclusion, but to me they are more than “suggestive”. Perhaps there are confounding effects from the 2008 melt-down and following economic malaise, but I’d think they would affect the area as a whole. If it isn’t the wind turbines that produce this rather clear pattern, then what is?
2 thoughts on “MPAC and Wolfe Island, again”
The use of p-values for statistical analysis is coming under heavy fire. See http://www.nature.com/polopoly_fs/1.14700!/menu/main/topColumns/topLeftColumn/pdf/506150a.pdf
In your tables, you do not report the all-important magnitude of the price reductions…is it $1.00 or $1 million dollars?. Your p-valued statistical analysis would not discriminate between the two reductions in price, but everyone would agree that a $1.00 price reduction for a $300,000 house is nonsense. By contrast, a $100,000 price reduction would be troublesome. So you folks should report the raw data in terms of selling prices.
I note your tables also report price reductions for the eastern half of Wolf Island. What are the magnitude of those price reductions and how to they compare with the wind-farmed area of western Wolf Island?
Thanks for the comments and the link – it was an interesting article. The main reason I use statistics on property values is because the industry’s favorite property value apologists (i.e. Hoen, MPAC etc) use them. I much prefer, as do real estate professionals, individual comparisons. Read my Lansink postings, as examples. I didn’t include the magnitude of the reductions because I had posted on them earlier, and linked to that posting. I suspect the link was broken – it is now fixed – WordPress has been doing strange things to my permalinks. Just FYI, they range from 10% at the start of the period covered by the spreadsheet, and increase to 25% towards the end. You ask for raw data. I’d love to have it as well. You got $30,000?