One of my holy grails lately has been to figure out how many housing units were built in Minneapolis in each decade of the postwar era. This grew out of my long-languishing Potential Population Project and the need to find a basis for assuming the average density of new multifamily construction. For over a year now I’ve been compiling a spreadsheet of buildings and their build year and unit density, adding them manually as I come across them in my job (which is so boring that no more need be said of it). I now have on my list 39,392 units in 580 buildings built since 1947, which is 60% of the 65,912 units in structures of 5 or more units in the city.
You may be thinking, “what kind of idiot wastes his time on something like that, especially since someone else has already compiled that information and he would have found that by now if he’d only spent some time looking for it.” Well, I agree with you now, since I’ve figured out an easy and effective way of estimating number of units built per decade using that all-powerful database, the US Decennial Census.
It all started a few weeks ago, when in a post about apartment construction that rambled into Minneapolis population change, I mentioned that the massive drop in population in the 70s was probably due to an after-dinner hiccup of freeway construction and renewal. I was thinking of 35W through Northeast, which was built in the early 70s, but Froggie pointed out that most of the clearance for that project had been completed by the end of the 60s.
This leads us to an interesting digression surprisingly early in this post. Froggie’s correction led me to actually fact-check one of my assertions for once, and after spending some time looking at old aerials, I realized that even more freeway clearance was done in the 60s than I’d thought. Besides the Northeast portion of 35W, which was built in the 70s but cleared in the 60s, much of 94 on the Northside had been cleared by 60s as well, although it would not be built until the early 80s.
So it turns out that the only freeway clearance that may have extended into the 70s was for Hwy 55, which ended up not being a freeway of course. Borchert library has an image (5 MB) from 1969 that shows partial clearance:
And Historic Aerials has one from 1979 which shows a bit more gone, although obviously I don’t know how much if any of that clearance happened in the 70s:
(The above two photos show the intersection of Hiawatha and Lake, with the majority of both images showing the Corcoran neighborhood)
So freeway clearance doesn’t seem to have been a major contributor to population loss in Minneapolis in the 70s. A 1971 map indicates renewal was probably more of a contributor, with activity in the 70s happening in Seward, Holmes, and perennial HRA punching bags Cedar-Riverside, Hay, and Near North. It seems likely that some of Plymouth Ave was cleared in the 70s in the distant wake of the riots, and also that a portion of the enclave of suburbs in North and Northeast were built in the 70s, at half the density of the urban fabric they replaced.
For some reason, all these units being destroyed and built up again led to a wall in my brain finally crashing down, allowing understanding to spring through: The Census tracks building age every 10 years. If I want to have a good idea how many units were built in any given decade, just look at the Census for the last year of that decade and look at what it reports for units built in the decade prior. This isn’t the exact number for two reasons, but it should be “close enough” as we say here at horseshoes&handgrenades.com. Reason #1 is that some of the units built in the prior decade could also have been destroyed that decade. My guess is that rarely happens but you never know. Reason #2 is that censuses are never consistent in the time periods they report. Here is my compilation of year built data from the 1950 to the 2010 censuses:
The table above gives you a sense of the varying time periods reported by the different censuses. Actually I’ve cheated a bit by cramming decade categories that are off by a year or less into one category, which you can see in the 1970 column, where I added the 1960 to 1964, 1965 to 1968, and 1969 to 1970 categories into the 1960 to 1969 category. That explains why the number of units built between 1970 and 1979 grew by 1,706 between 1980 and 2000, but not why the number of units built between 1980 and 1990 grew by 1,233 between 2000 and 2010. That last anomaly is more likely due to another cheat: the structure age data was actually from the ACS rather than the decennial census, as I’ll complain more about later.
So census data suggests that more units were built in the 60s than in any other postwar decade (the 1970 census reports 20,184 units that had been built between 1960 and March 1970), but the city’s total number of units dropped by almost 6,000. But not enough was done, apparently, to make a significant impact on total population. You smart people probably figured this out long ago, but in a city as big as Minneapolis, it’s really hard to add or remove a significant percentage of units.
Again using building age data from the Census, I’ve estimated the number of units built each decade, and extrapolated from that and the change in total units to get the number of units destroyed. In the postwar era, the net change in dwelling units reached a maximum of 7% increase between 1950 and 1960, when some greenfields were still being developed in the far north and south of the city. Interestingly, the next greatest change in units was the 6% increase between 2000 and 2010 – I’ll get to that in a moment. In the intervening decades, the change in units has fluctuated between 1% and 3% plus or minus, so that by 2000 the total number of units was only 1500 more than 1970.
As you can see, the change in units in any particular period is relatively small and apparently unrelated to the change in population. Two notes about the chart above – first, the units destroyed isn’t calculated for 1950 because I didn’t bother to find the total units for 1940 (a brief digression, though – the 1950 census suggests 12,425 units were built in the 40s, which is about 2,500 units more than were built in the 00s). Second, my method calculated only 298 units destroyed in the 00s, which is almost certainly too low. I think that is due to the switch to ACS for housing data such as structure age – because the total units recorded in the 2010 census is lower than the total units estimated in the 2005-2009 ACS, I probably should compare it to the ACS report of the units built in the 00s. Hmm, should I choose consistency or results?
Another factor in play in the interface between population and dwelling units is the vacancy rate. However, the vacancy rate in Minneapolis has been remarkably steady through the postwar period (at least in the 7 years in which a census was conducted), staying at 4% every year except for 1950 when it was 2%. Well, there are two more exceptions, and they’re doozies.
The 2010 census found an 8% vacancy rate, which I believe is mostly explained by the foreclosure crisis, although overbuilding or overconverting of condos likely played a part. As I mentioned, the Northside was the only sector of the city with significant population loss between 2000 and 2010, losing 7,704 residents during that period. Assuming the average household size of 2.23 persons/unit, that’s the equivalent of 3,455 units, which is about 2% of the total units in 2010. That doesn’t sound like a lot, but it means that this one sector that represents 15% of population was the locus of half of the increase in vacancy from the typical level of 4%. (Or it would be, if I hadn’t conjured the number of vacant units there out of other figures. Still, glancing over maps of foreclosures makes it feel real.)
The other census that found a high vacancy rate was 1990, when it reached 7% – my knowledge of this period doesn’t go much beyond the Bartman, but it might have something to do with the building crime wave that gave us the nickname Murderapolis. Another explanation may be the Savings & Loan Crisis that peaked just before this census (and eventually led to our current housing crisis due to inept and/or corrupt legislation). Anyway it seems to have had a similar effect as the 2010 vacancy rate – the population dropped slightly despite an increase in dwelling units.
The above chart shows a flat vacancy rate during the biggest population drops in Minneapolis history, in the 60s and 70s. And the next chart up shows that change in total units likely played only a small role in the 60s and none at all in the 70s. So what caused those drops?
As Jon pointed out in his comments on Apartments go boom!, the answer is “in the details of who moves in and out.” Families moved out and singles moved in, causing the average household size to plummet and the population to plummet with it. This is borne out by charting average household size against the change in population, which tracks remarkably close:
Of course you can make charts prove anything, but it is just as convincing when you compare the rates of change of the two metrics:
I have to admit that I dropped out of statistics and ended up with a math credit from Maps and Geographical Reasoning, but those numbers are pretty convincing to me. The only problem is that average household size, more than most metrics, invites more questions than it answers. It’s more of an indicator of other trends. What caused the massive drop from 1950 to 1980? Whatever it was, it wasn’t unique to Minneapolis. Here are charts of Richfield and St Louis Park:
Starting with 1960, these towns follow a similar trajectory in average household size as Minneapolis did. (1950 was probably a peak in average household size, not to mention total population, for Minneapolis due to the postwar housing shortage.) In fact, the four other Hennepin County cities for which I have data showed similar or greater decline in average household size in the 60s and 70s. These suburbs kept growing through the 60s with greenfield development, but as soon as they ran out of land, they mostly ran out of growth.
Kind of interesting to see that the county as a whole lost population in the 70s. Anyway, the best my feeble brain can do to explain this widespread drop is to blame the boomers. That swollen generation would have come of age in the 60s and 70s, presumably creating smaller households than they came from. I would add that the boomers and their smaller households likely created demand for smaller units, which fed the construction of housing units in the 60s and 70s seen in an above chart. This construction trend was metro-wide, and identified at an early stage in some maps that I included in an earlier post.
So while it may seem that size is, in fact, everything, data grail seekers must choose carefully. The notion that there is one holy grail may be illusory, and the truth may be that the grail can be found in many seemingly disparate measures. But that doesn’t mean the quest is not worth pursuing, as seekers will encounter many intriguing charts, graphs and maps along the way.