A line in Jeff Singer’s DKos Elections Morning Digest caught my eye, then grabbed me by the throat and shook the beejeezus out of me. Jeff was discussing a report by Alan Abramowitz of Emory University that analyzed the impact of spending and other factors on the 2014 Senate election using data from the Center for Responsive Politics.
[The report] compares outside spending by Dem- and GOP-friendly outside groups in key 2014 Senate races, and shows little relationship between the ratio and each race's results.
Wait, what? You mean all that Koch brothers dark money actually had little impact on the election results? Well, as a matter of fact,
Each state's presidential-election partisanship, and the benefit of incumbency, explain almost all of the variance in the Democratic margins; party and outside spending both barely register.
OK, this bears some looking into.
I took a closer look at the Abramowitz analysis. The report provided a regression analysis of the factors that shaped the 2014 election results. Here is the methodology:
In order to measure the effects of outside spending and party spending on the outcomes of 2014 Senate races, I conducted a regression analysis with the Democratic Senate margin as the dependent variable and four independent variables: the difference between pro-Democratic and pro-Republican outside spending in millions of dollars, the difference between Democratic and Republican Party spending in millions of dollars, incumbency (coded as +1 for Democratic incumbents, 0 for open seats, and -1 for Republican incumbents), and the 2012 Democratic presidential margin.
In other words, the study attempted to find the coefficients of the following equation that would minimize the errors from the combined 34 Senate races, and then predict the margin of victory (or defeat) in each race.
DeltaDem = B1*Delta2012 + B2*IDem + B3*DeltaSparty + B4*DeltaSdark + B0,
where DeltaDem is the margin of victory or defeat (in percent) of the 2014 Dem Senate candidate, Delta2012 is the percent margin of victory or defeat for Obama over Romney in 2012, IDem is the incumbency factor (+1 for a Dem incumbent, -1 for a Rep incumbent, and 0 for an open race), DeltaSparty is the difference in spending between the DNC and the RNC in millions of dollars in each race, and DeltaSdark is the spending differential between liberal and conservative dark money groups in millions of dollars. The factor B0 represents the offset, which is the average percent differential between Dem and Rep candidates, all other factors being equal. The results are summarized in this table.
Table 2: Results of regression analysis of 2014 Senate election results
Source: Center for Responsive Politics and Alan I Abramowitz
If spending were a major factor, it would show up in terms B3 and B4. Let’s look at each factor one at a time and try to figure out what’s going on here.
The coefficient B1 means that for every one percent that Obama edged Romney (or vice versa) in 2012, the 2014 Senate candidate in the same state achieved on average a 0.781 percent margin. That makes sense, it’s the same statewide electorate voting just 2 years later. This turned out to be the strongest correlation in the study. Based on the standardized coefficients, it appears to have accounted for about 69 percent of the outcome (compared to the 3 other factors measured here).
The coefficient B2 means that the incumbent candidate in each race achieved on average a 9.1 percent advantage over his opponent. Again, not surprising, incumbents always have a big advantage, assuming they're not under indictment or caught visiting a paramour in another country. The standardized coefficients suggest that incumbency accounted for 34 percent of the outcome in 2014.
OK, what about spending. Term B3 represents the impact of party spending on the outcome of each race. Remarkably the coefficient is -0.050. In other words, on average, every million dollars in spending by the RNC or the DNC in excess of the opposition party's spending resulted in a LOSS of 0.05 percentage points, which essentially says that that money was ineffective. Granted, the standard error is large for this term, so in some states spending was helpful, but in many states, the impact was even more negative. The standardized coefficient for this term shows a 0.3 percent impact (i.e., almost none), and in the negative direction.
And finally we come to the dark money term, B4. The table shows a coefficient of -0.238: in other words, every million dollars spent by conservative (or liberal) groups in excess of the opposition actually cost them almost a quarter of a percentage point for their candidate. The standard error is significant, but not nearly as large as the RNC/DNC spending error. The standardized coefficient shows a 3.9 percent impact, and again in the negative direction.
The offset term B0 was -6.392 percent, which means that all other factors being equal, the Dem candidate lost on average by more than 6 percent, but since there were far more red-state races in play than blue, that also makes sense.
So what does all this mean? Did entrenched state-by-state political leanings trump every effort to move the electorate in either direction? Are Americans sick of campaign ads that insult their intelligence? Did the DNC and RNC choose the wrong places to spend? Did Citizens United have far less impact than feared (as I predicted here, but was roundly trashed for it)? Did the Koch brothers waste their money? Hell, I don’t know. But if I were Charles or David, I would be asking a lot of pointed questions right about now.
The Kochs are planning to spend almost $1B in the 2016 campaign. Let’s see, based on this analysis, that alone would result in about 1000*0.238 or 238 percentage points in gains for the Dems. If half of those funds were spent on the Presidential and Senate races (the remainder on Congressional and local races), then that cash should boost each Dem candidate by about 3.5 percentage points apiece. Unless of course, liberal PACs and the DNC also fund those same races, which would reduce the Dems’ margins accordingly.
Go figure.