If you’re a regular Daily Kos Elections reader, you’re probably familiar with our media markets-to-congressional districts spreadsheet. It’s most helpful for when you see news of “$x million ad buy in x media market!!” in a Politico-type publication without any discussion of what congressional district it’s intended for; with the help of the spreadsheet, you can quickly ascertain what race (or races) that money is being spent on. But not only does the database help you be a more informed consumer of campaign strategy news, it can also help you make smarter use of your own contribution dollars, by letting you pinpoint the races that give you more bang for the advertising buck.
Now we’re adding a new component to our array of media market-related information. We’ve calculated presidential election results for each media market, in much the same way that we’ve calculated presidential results by congressional districts and legislative districts. This may not be as directly useful to you as presidential results-by-CD or LD, because CDs and LDs are the precise geographies that elect the members of the legislative branch, and the results of those elections tend to closely track presidential elections, so those numbers have a lot of predictive value. While knowing the lean of a media market can be important to the advertising arm of a campaign (for example, a campaign running statewide in Pennsylvania might want to run differently-themed ads in the Philadelphia market versus in the Johnstown media market, given the different priorities of the residents of those parts of the state), no elections are actually decided along media market lines.
But what’s very interesting about these presidential/media market numbers is that, unlike congressional districts or legislative districts, media market boundaries don’t change every decade. Unless you somehow have access to precinct-level data from the 1980s or 1990s, you couldn’t go back and track the presidential results for a particular CD over the decades. Media market boundaries follow county lines and stay stationary, so you can track those numbers longitudinally and watch how different parts of the country evolve politically, at a more detailed level than states but at a less granular level than individual counties. I’ve included data going back to the 1980 presidential election.
One thing you’ll notice, for instance, is that the nation’s most populous media markets are now among the bluest parts of the country: the New York media market (which includes not just New York City but all of its suburbs on Long Island, in New Jersey, and in Connecticut) went for Hillary Clinton 62 to 35 in 2016, while she won the Los Angeles media market 62 to 32. That may not seem surprising to you given the growing urban/rural split in political preferences, but what may be surprising was that the Democratic nominees were losing those media markets as recently as 1988; Michael Dukakis lost the New York market 50-49 and lost the Los Angeles market by a whopping 54-45.
The flipside, of course, is that there are hundreds of much smaller media markets around the country where the Democratic brand has declined precipitously; one very typical example is the Ottumwa, Iowa, media market, where Dukakis outright won, 51-48, but where Clinton lost 64 to 31 in 2016. That may not matter that much at first glance when the New York market has nearly 150 times more votes than the Ottumwa market (8.4 million versus 57,000 in 2016), but when that same thing happens in dozens of similar rural markets, that’s the stuff that “death of a thousand cuts” is made of.
To be better able to see this phenomenon at a glance, take a look at the second tab on the spreadsheet, where I’ve consolidated all the numbers down to a one-year Partisan Voting Index number. “PVI,” as it’s known, is a concept created by Charlie Cook that measures the relationship between the two-party vote share in a CD (or any other geography), and the two-party vote share nationwide. (For instance, if Hillary Clinton got 50 percent in a market, while getting 48 percent nationwide, that would be D+2; if Barack Obama got 53 percent in that same market in 2012 while getting 51 percent nationwide, that would still be D+2. The advantage of PVI, compared with results expressed as simple percentages, is that it makes a time-series comparison much easier between elections that were close and elections that were blowouts, as well as with elections where there was a large third-party effect.)
As an example, let’s take a quick look at some of the major markets in Pennsylvania, as a way of illustrating the vastly different directions the various parts of the state have been taking. Here are the Philadelphia and Pittsburgh markets over the years since 1980, as well as the somewhat smaller Harrisburg market in the middle of the state, which has stayed pretty much the same during that whole period, and the even smaller Erie, Johnstown, and Wilkes-Barre markets, which, while not starting out as blue as Pittsburgh, have taken a similar downward trajectory. Taken together, they tell the story of what happened in Pennsylvania in 2016; the Philadelphia market, while staying dark blue, didn’t become even bluer in the way that the Clinton camp had hoped, in a way that would offset the rest of the state becoming even redder:
| MARKET |
2016 |
2012 |
2008 |
2004 |
2000 |
1996 |
1992 |
1988 |
1984 |
1980 |
| PHILADELPHIA |
D+11 |
D+11 |
D+11 |
D+10 |
D+10 |
D+7 |
D+5 |
D+2 |
D+5 |
D+3 |
| PITTSBURGH |
R+6 |
R+4 |
R+3 |
D+2 |
D+3 |
D+1 |
D+9 |
D+12 |
D+14 |
D+7 |
| HARRISBURG |
R+14 |
R+12 |
R+10 |
R+13 |
R+14 |
R+15 |
R+14 |
R+12 |
R+11 |
R+12 |
| ERIE |
R+8 |
D+1 |
D+2 |
D+2 |
D+1 |
D+1 |
D+3 |
D+4 |
D+4 |
D+1 |
| JOHNSTOWN |
R+20 |
R+15 |
R+9 |
R+9 |
R+9 |
R+6 |
R+4 |
R+0 |
R+0 |
R+2 |
| WILKES-BARRE |
R+14 |
R+4 |
R+3 |
R+3 |
R+2 |
R+3 |
R+5 |
R+3 |
R+1 |
R+2 |
You might be thinking “wait a minute, Pittsburgh is still a dark-blue city.” Well, the unit of analysis here is the entire media market; it doesn’t incorporate just Pittsburgh, or just Allegheny County (which is still very much a blue county, though somewhat less than where it was a few decades ago), but all of the market: the entire Pittsburgh metropolitan area, plus some of the entirely rural counties that are situated outside that metropolitan area but still watch Pittsburgh TV affiliates. That R+6 is largely because of quickly-Republican-trending places in the exurbs like Westmoreland County and Fayette County. So you’ll see other instances where a major city reads as “Republican-leaning,” because it includes not just the city but its dark-red surroundings (like, for instance, St. Louis, which is currently R+4, thanks to the rural parts of eastern Missouri, or Nashville, which is R+14, because of the large and deeply-evangelical rural swath surrounding its metro area).
Finally, one additional benefit to presidential results-by-media market data is that sometimes you’ll see statewide poll data broken down by media market in the crosstabs. Looking at presidential results-by-media market data can help you assess whether the crosstabs pass the smell test. To zoom in on that, check out the third tab on the spreadsheet which breaks the 2016 results down by each state’s share of each media market.
In other words, many media markets cross state lines (you can find a full rundown of what county goes where on Wikipedia), and in some of those cases, you find conservative suburbs across the state line from a major city on the other side of the river. One prime example is the Wisconsin portion of the Minneapolis media market; the exurbs on the other side of the state line in Wisconsin are much more conservative than the Minneapolis market as a whole, so you wouldn’t want to assume that the Minneapolis market numbers as a whole apply to a statewide poll of Wisconsin that included a subsample of the small portion of Wisconsin that’s in the Minneapolis market. Instead, you could look at the third tab to see that while Hillary Clinton won the Minneapolis market as a whole by a 47 to 44 margin, she lost the Wisconsin portion 37 to 57.
