NEW YORK (TheStreet) -- "It's the economy stupid." It was the 1992 presidential election, and Clinton adviser James Carville was exhorting the campaign to focus on what mattered most to voters: their pocketbooks.
The economy may not be at the top of voters' minds in every election, but it is close. This is the principle underpinning the Moody's Analytics presidential election model. The model predicts whether the Democratic or Republican presidential nominee will win the popular vote in each state and the District of Columbia. In our Electoral College system, the candidate who wins the plurality of votes in enough states with a total of at least 270 electoral votes wins the election.
The model is constructed based on presidential election results since the 1980 Reagan-Carter contest, and captures the impact on voting decisions of the health of state economies in the lead-up to the election as well as the party affiliation and political realities in each state.
The model is similar to the one we used to successfully predict President Obama's victories in 2008 and 2012. There are other election models that use economics to help explain election outcomes, but the Moody's Analytics model is the only one we are aware of that fully incorporates state-level economic activity. This is key since the election is won or lost in the state Electoral College.
To cut to the chase, we predict that the Democratic nominee for president will win the election by the slimmest of margins with precisely 270 electoral votes. The Republican nominee will fall just short with 268 votes.
The Moody's Analytics Presidential Model will be updated monthly.
Economics in the Model
The election model explains the share of the popular vote in each state that the incumbent party will win based on the strength of each state's economy and its politics.
The most important economic variable in the election model is the growth in real personal income per household in the two years leading up to the election. This variable captures the strength of the job market in each state, including job growth, hours worked, wage growth, and the quality of the jobs being created. It also captures how well households are doing on some of their investments, as it includes dividends, interest income and rents. Since it is on a real or after-inflation basis, it also captures the impact of inflation on the purchasing power of households' income.
With regard to the current presidential election cycle, household incomes have been steadily improving in most of the country. Job growth has been robust, hours are back close to previous highs, and jobs across all pay scales are being created. The only missing ingredient is stronger wage growth, which is expected to pick up in coming months as the job market approaches full employment. Inflation is also low. Real household income growth thus favors the incumbent party, the Democrats.
The growth in house prices in the two years preceding the election also influences voting decisions. A two-year percent change is likely measuring, at least in part, expectations regarding future house price growth, as house price expectations appear to form adaptively. That is, homeowners base their thinking about future price growth on recent price gains. House prices also matter only to voting beginning with the 2004 election, which is consistent with the timing of the formation of the house price bubble. Prior to that, house prices mostly always rose in most of the country at close to the pace of household income.
The recent rebound in house prices in the wake of the housing bust again augurs well for the incumbent Democrats. House prices are expected to be rising solidly in most of the country prior to the election.
The final economic variable included in the election model is the percent change in gasoline prices in the two years leading up to the election. Gas prices are a very visible price that matters significantly to the finances and perceptions of many households. Consumer confidence is closely tied to movements in gas prices, as rising prices quickly undermine confidence and falling prices lift spirits.
Oil and gasoline prices have fallen sharply over the past year (almost $1 per gallon), and prospects are good that they will remain low through Election Day. The Saudis continue to increase their production despite the much lower global prices, and it looks increasingly likely that there will be a finalized deal with the Iranians over their nuclear program, which means Iran will soon be pumping more oil. Low gasoline prices also favor the incumbent Democrats.
While real household income, house prices, and gasoline prices are included in the election model, many other variables were tested to see whether they made the model more predictive. Unemployment, stock prices, employment growth, consumer confidence, and many other variables were tested. None worked as well in explaining election outcomes.
The economy's performance strongly favors the Democratic nominee for president.
Politics in the Model
While economics matter a lot on Election Day, so too do politics. Most important, most states tend to vote consistently Democrat or Republican, regardless of whatever else is going on. Massachusetts, for example, is a dark blue state, while Wyoming is dark red. It is hard to imagine what would change voting patterns in these states. This reality is captured in the model by including the share of the vote that went to the incumbent party in the previous election as a variable.
In so-called swing states, the share of the vote that goes to the incumbent party is close to 50%. In these states, economic and other factors can swing the vote to either party. The key swing state is Virginia. Iowa, New Hampshire, Nevada, Pennsylvania and Wisconsin have also been swing states in past elections.
Another political factor that will likely play an important role in the current election is voter fatigue. Historically, the electorate tends to vote the incumbent party out of the presidency if the party has been in office for two consecutive terms. Think Bill Clinton and George W. Bush. The exception of course is George H.W. Bush, who won in 1988 despite following the two Ronald Reagan terms. The fatigue factor weighs heavily against the Democratic nominee in this election. A number of swing states would likely vote Democratic in 2016 if not for voter fatigue.
This is particularly applicable given that the model also includes a variable that down-weights previous voting shares if the incumbent is a Democrat. The intuition is that Democratic voters tend to be less politically ideological than Republicans and are more likely to vote non-Democratic. This intuition has also been borne out in other studies of voting patterns.
The trajectory of the president's approval rating also makes a meaningful difference. If the sitting president's approval rating is improving in the year leading up the election, the incumbent party receives a boost in the election. In most elections, the president's rating has declined in the lead-up to the election, favoring the challenger party. It is assumed in our 2016 prediction that President Obama's approval rating will be the same on Election Day as it is today.
Being a favorite son appears to matter in some elections, but not enough to enter into the model. Being a favorite son mattered most in Ronald Reagan's 1980 win in California, which is a Democratic-leaning state, and Bill Clinton's victory in 1992, when he won Republican-leaning Arkansas.
The national political backdrop strongly favors the presidential bid of the Republican nominee.
The presidential election model has accurately predicted every presidential election since 1980. Over the nine elections since 1980, the model accurately predicted the winning party in 406 of the 459 elections in the 50 states and DC. Nearly all of the misses across states were for very small states, and there were no states that the model missed on a consistent basis.
The model nailed the 2012 election, accurately predicting the election outcome in all of the states and hitting the Electoral College vote on the nose. The model did least well predicting the 1992 election between Bush senior and Bill Clinton. Predicting this election was significantly complicated by Ross Perot's third-party candidacy, in which he won nearly one-fifth of the popular vote and severely hurt Bush's re-election bid. The model also nearly missed the controversial 2000 race between George W. Bush and Al Gore, which was ultimately decided by the Supreme Court.
Democrats Win ... Barely
The next president should be a Democrat (see map). According to the model, the Democrat should win 270 electoral votes to 268 for the Republican. The key swing states that push the election to the Democrats are Colorado, Ohio and Virginia. The Republican candidate will win Florida.
The election appears to hinge critically on Ohio and Virginia. Based on research modeling election results at the county level, the vote in Ohio thus appears to depend critically on the vote in Hamilton County (Cincinnati), which is historically Republican, and Cuyahoga County (Cleveland), which is historically Democrat. The 2016 presidential election could thus come down to voter turnout in Hamilton and Cuyahoga counties -- not much of an endorsement for the Electoral College approach to picking presidents.
Virginia is also on the fence. If President Obama's approval rating falls by any more than 2 percentage points by Election Day, Virginia will swing and the Republicans will win the presidency.
These results are subject to a significant degree of uncertainty and will change as more economic data come in. The current results are based entirely on the Moody's Analytics projections for real household income growth by state and gasoline prices as of the third quarter of 2016. If our forecasts of income growth are overly optimistic, then the Democrats will not do as well as currently expected. And if gas prices, which are expected to remain low, instead rise, that too would boost Republican chances.
The model results are less valid if either political party nominates a non-establishment candidate. Elections since 1980 have been between candidates who are generally thought to be largely in the mainstream of American politics. Some of the current presidential candidates are more on the fringes of the political spectrum.
Of course, it is a bit (a lot) self-centered to think elections revolve simply around financial matters. Surely foreign affairs, social issues and personalities matter also. And at times they matter most of all. But in most times, and 2016 appears to be stacking up as one of those times, it is the economy stupid.
The Moody's Analytics Presidential Model will be updated monthly.
This article is commentary by an independent contributor. At the time of publication, the author held no positions in the stocks mentioned.