I must digress. The first house Shirley and I bought was in the Detroit area. After looking at several gorgeous, ridiculously cheap, mansions in the City of Detroit, we decided to look for more modest digs in the suburbs. We wanted to identify the “good” neighborhoods near the GM Tech Center, so on a map I shaded 2-mile radius circles around every K-Mart. The closest hole in the shading was the eastern half of Lathrup Village, a small municipality completely surrounded by the City of Southfield. Lathrup Village, known for its exclusionary deed restrictions, had one house on the market suitable for a budding GM executive. We bought it, becoming some of the first residents of Lathrup Village who were not from northern European stock. We lived on Sunset Drive, several doors down from Gordie Howe’s house. As long as we kept our ample lawn mowed, nobody much cared where our grandparents were born.
Our move to Wisconsin was sudden, and we had almost no time for house hunting. I had been to the Milwaukee area only once before (see https://ajhassoc.com/index.php/2020/01/22/the-elusive-concept-of-delay-in-urban-traffic/). I did not have much time for research, but I did have a large stack of results from a land-use model of the Milwaukee area.
There is little apparent rationale for my possessing a land-use model of Milwaukee while living in Detroit. Perhaps it was David Boyce, an occasional consultant to GM, who convinced them to build such a model. GM was interested in the urban impacts of tiny cars, a land-use model seemed to be a good means of investigation, and we were sitting on a treasure trove of SEWRPC (Southeastern Wisconsin Regional Planning Commission) data: the latest highway network, zonal demographic information, …, the works. SEWRPC gave GM these data, once GM promised never to disclose any of it.
I wrote the code, calibrated the model as best as needed for GM purposes, and then ran myriads of future scenarios, with and without tiny cars. The model was unconstrained as to housing supply, so it could not be used for short-range analysis.
One of the most obvious findings was that without any new influences to slow things down, the Milwaukee area would undergo decades of suburbanization due to freeways. Of course, our interest in tiny cars led us to consider scenarios with high fuel prices. One of these scenarios guided my house-hunting decisions.
I believed in an inevitable rise in fuel prices, so I focused on those particular scenarios. Among the housing markets seeming to do best was Whitefish Bay. Putting all our money where my mouth was, we bought there. Today, Whitefish Bay does not have anywhere near the highest average home prices in the region, but it is the place where you get the least house for your investment.
I must digress again. Detroit was hollowed-out by the impact of freeways. Maybe some of the early rush to the suburbs can be attributed to the riot of 1967, but the extreme decentralization cannot be explained by racial issues alone. Consider that the population of the Detroit metropolitan area is almost exactly the same today as it was in 1960 (about 3.5 million people), but the City of Detroit has declined in population from 1,670,000 people to 670,000 people. The rate of decline has been slowing lately, perhaps indicating suburbanization trends are finally playing themselves out. The beginning of the decline in the City of Detroit population actually predated the riot by several years. When I was just a lad, I set up a residential location model for the Detroit metropolitan area. The pattern of change was predicted well by the model at the time, and as I can best recall, the predicted pattern of change has roughly continued through subsequent decades.
So I have often asked the rhetorical question to students and colleagues, if I can build a well-functioning land-use model from scratch for the Milwaukee region in just a couple months using SEWRPC’s data, why couldn’t SEWRPC do it also in say 2 years or 20 years?
Alan Horowitz, Whitefish Bay, October 18, 2020
Good question! Maybe they need to hire you or someone like you? And look for Targets, rather than K-marts? (Were you trying to avoid those or the zoning they imply?)
These days I would think dollar stores would be a good measure of neighborhood quality.
Thanks for the article & I’m keen to understand more. When you said that Whitefish Bay was forecast to be one of the best, what metric(s) were you using – is it growth in the price of housing? I guess today what you are observing is the house prices are not the highest, but in terms of price per square foot it is the highest in Whitefish Bay?
It is interesting, & I think in a good way, that your model predicts the suburbanization trends that are playing out in the real world now. My guess is the demand for suburbanization has been met by sufficient supply and now there are demands for other types of living that the city needs to provide?
Has the model been used for other cities, or could it be? Here in Auckland, New Zealand and largely across different metros in the country there has been a severe shortage of land for suburbanization over the last 30 years which has resulted in extremely unaffordable housing because that is where most of the demand is. The city is responding by further restricting land supply which will exacerbate this. I believe we have yet to find our equilibrium that your model may be able to predict.
Thanks for your article, I believe it is relevant.
As for “best”, at the time I used forecasted, unconstrained population increase. The model did not have price an an output. It would have been possible to derive a shadow price by imposing additional costs on access to land, but I needed to use the model outputs I had on hand at the moment.
The highest home prices in the Milwaukee area are in River Hills, which has estate-type zoning. Both lots and houses are quite large there. Whitefish Bay is a distant second. Whitefish Bay has many modest homes, but none of them are cheap by Wisconsin standards.
The central city in our region, Milwaukee, has stopped losing population and has many desirable neighborhoods. Areas close to downtown are doing especially well. However, parts of Milwaukee are still depressed.
The model is of a Lowry-Garin type. I actually created a standalone piece of software (HLFM) which was a much fancier version of what I did at GM. However, newer concepts came along, and I stopped trying to compete. At some point, I incorporated HLFM fully into QRS II, so you can actually try running your own land-use model with the QRS II demo edition for free. You can download the demo edition with documentation from this site.