But history suggests that running mates don’t actually move the needle much. Take the example of North Carolina between 2000 and 2004. Democrats, under Al Gore, lost the state in 2000 by a margin of 56.03% to 43.02%. When John Kerry added popular North Carolina Senator John Edwards to the ticket four years later, Kerry lost the state to Bush 56.02% to 43.58%.

In other words, for all the attention a VP pick receives, it may not have much of an impact on election outcomes.

On the other hand, there’s much better evidence that economic conditions in the months leading up to an election have a major influence on election outcomes. And while the smart money is betting on Hillary Clinton—based on Donald Trump’s very high unfavorable ratings and the tough time Republicans have had in the electoral college over the past 25 years—other pundits have been arguing that the economy will be in Trump’s favor this fall.

The Guardian’s economics editor Larry Elliott argues as much in a recent article, writing “The state of the U.S. economy suggests that [Donald Trump] has a fighting chance.” He continues:

The first point to recognise is that the economy is struggling. Growth slowed to an annualised 0.5% in the first three months of 2016, which means that it was barely expanding at all. Unlike in previous years, this could not be blamed on inclement weather.

Jobs are still being created but at a slower pace. Last week’s figures for non-farm payrolls – the benchmark for the state of the labour market – showed an increase of 160,000 in employment last month, well below the 200,000 Wall Street had been expecting.

Elliott points out that productivity and wage growth have been nearly non-existent in recent years causing “millions of Americans [to] think they have been left behind.”

While this is a good summary of the American economy’s woes of late, it’s not a slam-dunk case that Trump has an advantage, at least on economic grounds, come November. Back in 2011, blogger and statistician Nate Silver published a study on which economic indicators correlate strongly with presidential election outcomes. He found that GDP growth is a much weaker predictor than metrics like total job growth and the ISM manufacturing index. Those two indicators look like they will help boost the incumbent party candidate, Hillary Clinton. While the global manufacturing sector began to struggle last year, it has recovered in recent months, gaining 2.6 points since January and showing that the sector has been expanding in March and April.

Job growth did slow slightly in April, when the month’s jobs report showed that the U.S. economy added 160,000 new jobs, fewer than economists expected. But even at this level, the economy is adding roughly twice the number of jobs needed to keep up with the growth of the labor market. Furthermore, job growth in the first quarter of 2016 came in at 667,000 new jobs, faster than three of the four quarters in 2012, when President Obama won a second term.

Elliott is right to point out that wage and productivity growth has been much slower than Americans are used to. But if productivity growth were stronger, employers would be under less pressure to hire more workers, because their existing employees would be getting more done.

At the same time, we should hesitate to put too much faith even in the metrics that have a good track record of predicting election results. As Silver wrote at the time, none of the individual indicators correlates strongly enough with election results to explain more than 46% of presidential election results, which means they will give you incorrect predictions more than half of the time.

So, it doesn’t make sense to only look at economic data when predicting elections. Other factors, like individual candidates’ approval ratings and positions on issues matter a great deal too. But at this point, the most accurate economic indicators of election outcomes suggest that Donald Trump cannot rely on a weak economy to propel him to the White House.