Philadelphia Actual Value Initiative - Why it might have failed - The Interaction Effect between Land and Improvements

Two years ago Philadelphia passed the Actual Value Initiative (AVI) in an attempt to properly revalue the properties for taxation. I wrote about this earlier. Ever since, there have been two different stories. One side argues that the AVI was accurate and increased fairness, whereas the other side argues the opposite.

Previously my best guess as to the problem was that the two different sides used different data (different versions of the Office of Property Assessment data set).

My new guess is that the city was unable to separate a property's value into its components (improvements and land) because they ignored the interaction effect between the two. They also ignored the fact that most of the Philadelphia housing market is very depressed and that the value of improvements is generally less than the cost of building them.

For instance, even in my neighborhood (University City) which is in the top 10% of the housing market, there is only a modest amount of new residential construction. There are a small number of empty lots that are sitting around for years and nobody is trying to build on them. This is because the cost of building a new house is so high that it makes more sense to buy an existing house. Most of the construction projects in this area are for the universities (for their facilities or for housing) and they aren't interested in a single lot. As such, a single lot is worth relatively little (maybe $30,000).

An average quality (2600 square foot) house sells for $300,000 in my neighborhood. If you ignore the interaction effect, then you would assign $240,000 of value to the improvements and $60,000 to the land. In fact the city estimates my house improvements are worth $250,000 and the land $60,000. However, if you were to take one of these houses and move it a mile west (say to the Kingsessing neighborhood) - it would be worth a lot less. It might only be worth $150,000. If you ignore the interaction effect, your model might predict that the land in Kingsessing is worth NEGATIVE $90,000. However if you use the interaction effect, you will discover that a house of equal size and quality will have two different values based on its location AND this is not due to land value as the housing market is depressed and it doesn't make sense to build.

While there is a small amount of new construction in University City, there is even less new construction going on in the heart of low-income neighborhoods like Kingsessing (they do have some construction on the edges due to gentrification). In many low income neighborhoods the only construction is affordable housing developments which are costing $200/square foot (and possibly public buildings like police stations - Philadelphia is currently selling and demolishing schools, not building them). As such the land value in these areas is nearly zero.

While low income households have a great demand for quality housing, they do not have the money to back it up. So under capitalism there is very little demand for a $200,000 house in a low-income neighborhood. You could try to sell the house to a new person who would move into the neighborhood, but there are many people who are not willing to move into a neighborhood that is low income, 98% black (or 98% non-white), and has bad public schools. Consequently an average quality house will sell for less (per square foot of equal quality of living space) in a low-income neighborhood than a higher income one.

If there is very little new construction in 90% of Philadelphia, then there are very few sales of empty lots and no market mechanism for estimating the value of land. So instead of looking at land sale prices, the city may have tried to estimate the land value by comparing buildings of equal quality on a city-wide basis. The City may have falsely assumed that the difference in property valuations of two buildings of equal quality and size was due to a difference in land price.

Or the City might have used the market valuations of land and this could have let to improper house valuations.
For instance, the city might take two equal size and quality buildings: a $300,000 building in my neighborhood and a $100,000 in Kingsessing. They might assume that the land price is $60,000 in my neighborhood and $0 in Kingsessing. Then they could calculate the average improvement value to be ($300,000-$60,000 + $100,000) / 5200 (the total number of square feet = $65.4 per square foot. Then they might assess my house at $60,000 + $65 * 2600 square feet = $230,000 and the Kingsessing house at $0 + 2600 * $65.4 = $170,000.

The over-assessment of houses in low-income neighborhoods (and under-assessment in high income neighborhoods) is EXACTLY the problem that the Controller's report, the Crosstown Taxpayers Coalition, and my own research has found. It does seem crazy, but my research found that the AVERAGE case was a $50,000 house in low income neighborhood that was improperly assessed to $100,000 and that a $300,000 house in a high income neighborhood would be assessed at $150,000. When you combine this problem with a failure to properly identify foreclosure sales - it could explain a lot of where the AVI went wrong.

A better example would be to compare a $300,000 building in my neighborhood to two (or more) $100,000 buildings in a low-income neighborhood - as there are more houses in low income neighborhoods. Using two $100k buildings, this would give a lower square foot value ($56 / square foot) and assessments of $207,000 (for the $300,000 house) and $147,000 (for the $100,000 house).

The critics of the AVI argue that its estimates for land value are wrong. That the AVI under-estimated the land value in very high value areas like Center City and it over-estimated the land value in low value areas. This could be a result of ignoring the interaction effect.

Comments are welcome! Is this theory right?
Email me