Last month, I analyzed three statistics that, according to statistician Nate Silver, best predicted presidential success for incumbent-party nominees and challengers. Donald Trump got the slight edge in that analysis, but economic statistics in recent weeks moved sharply in Clinton’s favor.

The two major metrics that had been pointing to a Trump victory, at the time, were the ISM manufacturing index—or a survey of confidence at American manufacturing firms—and the unemployment rate, which in September was higher than at the beginning of the year. Since World War Two, there hadn’t been an incumbent party candidate that won the election in a year that the unemployment rate was higher on election day than on January first.

But Clinton had plenty of metrics in her favor, as nearly all the indicators other than the unemployment rate showed a quickly improving economy that was filtering down to average Americans. Here are five charts from Torsten Slock, Chief International Economist at Deutsche Bank, that illustrate this point:

Workers are as Comfortable Quitting Their Jobs as Before the Crisis

Layoffs are Lower Than They’ve Been in 16 Years

Young People Are Finally Feeling the Economic Recovery

The Strong Labor Market is Causing the Share of Workers Outside the Labor Force to Fall

Wages are rising

This last chart is particularly important for understanding the strength of the headwinds Trump overcame. Many metrics are suggesting that Americans wages are broadly rising at a pace that hasn’t been seen in nearly a decade. One example is the data above from the Bureau of Labor Statistics, and another was a recent Census Bureau survey showing the fastest wage gains last year since the 1960s. But the increase was recent, and wages had lagged for much of the past eight years, and that likely hurt Clinton.

And given that wages are the most direct interaction most workers have with the economy, it’s not surprising that voters were looking for change.

But what about manufacturing data? Usually the single-best predictor of incumbent-party success is the trend in the ISM index in an election year, according to an analysis by Nate Silver. And if that metric is rising, then Silver found, the candidate of the incumbent party can count on a win. So the ISM readings as of October showing the manufacturing sector was growing seemed to be predicting Clinton would win.

Nonetheless, given the fact that manufacturing employment is so much lower today than it was when Silver started compiling his sample, one could have logically expected that manufacturing relative strength, as slight as it was, would affect the electorate less today than it traditionally has. And with more manufacturing workers on the sidelines than ever, those workers ended up being votes for Trump.

Many Republican observers argued that 2016 was a change election, pointing to statistics that show a large percentage of Americans believing the the country is headed in the wrong direction. It didn’t seem to matter that most Americans, outside those in the manufacturing sector, have fewer economic reasons to believe this.

In the end, when you look at all the charts what you see is the drops were too big and rebounds were too low. And the patience, it appears, for many American had run out.

So while it only became obvious on Tuesday night to most of us a large, and perhaps decisive share of Americans were clamoring for something different at 1600 Pennsylvania Avenue, a deeper look at the economic data should have told us that a change, like Trump, was coming all along.