Bsrgvty

Within the same city, as the market rises and falls, do cheap, average-priced, and expensive home values typically rise and fall at a similar percentage?

For instance, all homes in the city rise about 20%, across the price range:

 2010 value 2020 value 
cheapest $30,000 $36,000
average $100,000 $120,000
most expensive $5,000,000 $6,000,000

Or do houses at the higher end of the market increase at a greater percent?

 2010 value 2020 value 
cheapest $30,000 $31,000
average $100,000 $150,000
most expensive $5,000,000 $10,000,000

Or is there another pattern typical? I know that there can be a lot of factors, such as gentrification of a neighborhood, but is there an overall, fairly common pattern?

I used to work with a team that produced a pretty accurate housing prices model so I know a little about this. There is no simple relationship between average house prices for a valuation cluster (cheapest, middle, highest value). You cannot even say that rising prices at the lower strata imply rising prices at the upper end because there are a lot of complex variables that go into any model.

Like most prices house prices are purely a function of supply and demand. This sounds simple therefore; all you need to do is estimate those two numbers and you are fine surely? The problem comes when you realise that the affordability of housing at a given price is related to the cost of borrowing for a mortgage on the house and local wages.

Estimating the cost of borrowing in the future to make predictions from the model is an art about which thousands of books have been written. Even so the best models use Monte Carlo simulation to find the prevailing rates since the factors involved are even more complex. Remember that even if you don't need a mortgage or can afford a big one at a low interest rate, not everyone can. If the interest rates are right it will also attract more speculators.

If that sounded hard, modelling future wage growth and economic forecasting are even more complex. If you want to understand how hard economic forecasting is look at how often national statistical bodies restate their country's GDP for past periods. Since the measures of GDP for the past are inaccurate at best this kind of forecasting is fraught with pitfalls.

Add into this local demand the demand from overseas investors that depends on their country's economy and the legal risk in those countries and estimating demand is very hard. An example of the legal risk is that China is currently cracking down on investors buying assets outside of the country.

Supply is somewhat easier to model since you have a fixed number of already built dwellings and it is hard to demolish or build more quickly. In larger cities and some other areas, however, it is relatively easy to get change of use to turn commercial buildings to housing and back. It is hard to work out at what price this is viable and just as hard to work out at what price expanding the housing stock, by building more becomes profitable. This is further complicated by building regulations and government schemes and funding aimed at changing the housing stock.

Remember that supply and demand are different for each of your strata named above!

The models also rely on location data as input which tells us about gentrification and changes in area popularity. Sometimes we could use our price estimates to estimate crime levels in areas as small as a single postal/zip code.

So, how do the professionals get anywhere with this? The professional models use multi-tier regression models to model the underlying drivers such as GDP, interest rates, returns on house building etc. with an overarching regression model that estimates the prices given those factors. In our simplest model there were at least 28 historic economic factors being modelled that I can remember and a few hundred variables total. We used gigabytes of house price data from various sources. Our complex model used a lot of the same data but added a Markov chain model as the highest tier because a lot of the underlying factors feed into each other.

A single small change in government policy could easily render this model mostly obsolete and a lot of the calculations would have to be completely redone.

Therefore, for a long list of economic and government policy reasons there is no simple rule-of-thumb way to calculate how house prices will react. In fact we are even pretty bad at estimating the present let alone the future!

A city is multiple markets. Neighborhoods can be hot or cold depending on what things are going on around them. The type of housing stock can also act independently. If the local university adds a new campus in the suburbs then the neighborhoods near the new campus will change, but the demand for apartments and townhouses will react differently than single family houses.

If the amenities near a neighborhood change, then undesirable locations can become desirable. That can mean that apartments geared for singles and young families can become hard to find, or maybe people turn small houses into large houses.

Also averages can be influenced by hidden trends. In 2019 in Northern Virginia the debate is what impact will Amazon's HQ2 have on home prices. Initial analysis shows prices near the future project jumped 17%, but detailed analysis showed that wasn't the case. The mix of properties in 2019 that sold was different than 2018, so even though the prices of multifamily, townhouse, and single family houses didn't move much the mix changed so the average jumped.

The reason the median price is up so much isn’t because owners are
actually asking that much more for homes, it’s because the number of
homes listed from January-May 2019 vs January-May 2018 for under $700k
is down nearly 27% compared to a decrease of just over 9% for homes
over $700k.