Mapping My Christmas Card List
Because Why Not?
My wife and I are Christmas-Letter-Afficianados. We send a full-page letter to about 150 people every year, updating them on all the exciting details of our life. Having done this for 3 years now, we have address lists cataloging our changing social circles and family ties over the years.
But that also means I have 3 years of standard-format address lists for a fairly large group of people! Being who I am, I started to wonder: what would the migration rate look like for this group if I track them across years?
So I did. I have added together everybody who has an address in the same municipality, and obviously will be providing no info which could be used to identify the otherwise top-secret recipients of the much-coveted Stone Family Christmas Letter. So here’s a map of our 2017 recipients, with localities smaller or larger based on how many adult recipients reside there:
As you can see, we’ve got a big cluster in central and northern Kentucky. There are hubs around Lexington where we went to college and where I grew up, a hub in Ashland where Ruth grew up, a hub in Cincinnati where Ruth has lots of close-and-extended family, and a hub in Louisville, Kentucky’s biggest city, which thus exerts a kind of gravity on our peer social circles. Then we have a hub in Saginaw, Michigan, where Ruth’s dad’s family are from, and a hub around DC, where we have spent our married life. Then there’s a St. Louis cluster because we are part of the Lutheran Church-Missouri Synod, so have “natural gravity” there… and then everything else is scattered about! Many of the southern dots are my extended family.
This map surprised me a bit, as our social circles were (1) far more clustered and (2) far more Ohio-Valley-centric than I mentally pictured. That’s due to the heavy weighting towards Ruth’s extended family in our Christmas card list.
So what did 2016 look like?
Much the same!
But how did they change? This is where it gets interesting. For one thing, we reduce the number of cards we sent from 170 in 2016 to 149 in 2017, so many places will show a decline… but on net, we can see how the geography of our life-connections changed year-to-year:
What you see here is called “growing up.” Big red bubbles in our childhood homes and college spot, gains out in other places further afield. We lose touch with old friends, aged relatives go to sleep in Christ, college friends move out of Lexington, etc. Our regional social network grows more diffuse.
But how much of this was actually migration? Well, I can look at change of address, ZIP, municipality, and state.
Nearly 20% of the households who appeared in both the 2016 and 2017 address lists had a change of address. This was disproportionately one-adult-households, so the migration rate for people is lower than for households. Of the 16.9% of people who changed addresses, 3 percentage points did not change ZIPs, so made very-local moves. Then 6.3% of people changed ZIP codes, but did not change municipalities: so that’s still pretty local, but they’re not in the same neighborhood anymore. 7.6% of people changed municipalities… of which, 6.3 percentage points also changed states or countries! That’s the thing that is odd to me: fully 83% of people who changed municipalities in my sample changed states as well. But I suspect that’s partly demographic: our list is disproportionately very young families not yet suburbanizing for space, or older families nearing retirement who may be looking for a golden-years destination. Short-range-inter-muni migration probably peaks during the school-age years of children, so 30s and 40s.
Our social circles have much higher gross migration rates than society on the whole. I can’t compare ZIP and muni, but I can compare any move and state.
As you can see, I have very migration-prone social contacts! Their interstate migration rate is something like 2–3 times the national average, and their total migration rate is 25–50% higher!
How much of this is due to demographic differences?
Well, I don’t have full demographic data for my Christmas card recipients… but I can easily tell whether they are married or not, and I can guess their broad age range: under 30, 30–60, and over 60. I can then compare migration rates for each group (married/unmarried for each age class).
As you can see, for any address change, my Christmas card list has pretty similar demographic-specific migration rates. The unmarried under-30 crowd has higher mover rates, which probably relates to the fact that it is heavily composed of Ruth and my college friends, thus has an extremely strong educational selection bias. The sample size of unmarried retirees is extremely small so the 0-mover-rate may not mean much. The middle-age unmarried crowd is exactly what the ACS would predice.
For married under-30s, it’s again a higher-than-predicted-rate, but it’s closer to the ACS prediction as this includes more people included for family reasons. The married middle-age crowd has a higher migration rate too, but not wildly higher. The married retiree crowd has a lower rate.
If we composite these together and use the rates I have for each group to estimate a migration rate for the whole adult population based on ACS population composition, we get 16.2% migration instead of the 16.9% in the raw sample. Meanwhile, the adult-sample of the ACS has 14.7% movement, vs. 13.9% for the full sample. So adding very rough controls for age and marital status reduced the gap from 2.94 percentage points(that is, the Card-estimate was 2.94 percentage points above ACS, or 21.1%) to 1.5 percentage points, or 10%. Basically, these extremely rough controls sliced the gap in half. Add in an educational control and I suspect that the Stone Family Christmas Card list has an entirely-typical mover rate.
But what about interstate migration?
The sample size for interstate movers is a lot smaller; under 20 cases. So the demographic sample sizes are really small. And as you can see, the estimates range widely. I do think that these seem high; even with standard error ranges and a guesstimate of educational controls, the unmarried under 30 migration rate is almost certainly higher than the ACS would predict, and the married rate likely is too. So I think that the Stone Christmas Card recipients may be disproportionately likely, even controlling for demographics, to migrate across state lines, despite not being more likely to move on the whole: so their average migration distance is higher.
But demographic controls do reduce the estimated differences! Rather than being 6.33% vs. 2.36%, the difference comes to 5.3% vs. 3.1%. Still a big gap, but, again, demographic controls reduced the size of the gap by around half.
Conclusion
People often have a hard time believing it when I tell them migration is very low: and the Christmas Card exercise gives a possible explanation why. Very few people have “demographically typical” social circles, and it is almost certainly the case that the choices of young people are more salient than of old people in terms of perspective-forming. And if you just sit down and update an address list, boom, you get 25% to 300% higher migration than official surveys! It looks like high migration, but it’s really just selective perception. So next time you run across “natural data” about migration, whether updating an address list, car registry data, or other sources, remember: it’s not a representative sample! You need to make appropriate demographic adjustments!
Merry Christmas everybody! (Remember: Christmastide doesn’t end until Epiphany!)
Check out my Podcast about the history of American migration.
If you like this post and want to see more research like it, I’d love for you to share it on Twitter or Facebook. Or, just as valuable for me, you can click the recommend button at the bottom of the page. Thanks!
Follow me on Twitter to keep up with what I’m writing and reading. Follow my Medium Collection at In a State of Migration if you want updates when I write new posts. And if you’re writing about migration too, feel free to submit a post to the collection!
I’m a native of Wilmore, Kentucky, a graduate of Transylvania University, and also the George Washington University’s Elliott School. My real job is as an economist at USDA’s Foreign Agricultural Service, where I analyze and forecast cotton market conditions. I’m also a Research Fellow at the Institute for Family Studies and a Senior Contributor at The Federalist. I’m married to a kickass Kentucky woman named Ruth.
DISCLAIMER: My posts are not endorsed by and do not in any way represent the opinions of the United States government or any branch, department, agency, or division of it. My writing represents exclusively my own opinions. I did not receive any financial support or remuneration from any party for this research.
