It’s California that really grates on President Trump.

Early in the campaign, Trump’s “we’ll upset the expectations” rhetoric included bold predictions that he could win not only his home state of New York but the even-bluer Golden State.

“I think we can win the state of California and win it pretty substantially,” Trump said in San Jose in June. “Now, I’ve been told by all these geniuses, all these brilliant guys — they all say you can’t win the state of California. I think we can.”

The geniuses were right. Trump lost California by a historic margin — nearly 4.3 million votes, a million more than the next-biggest blowout in the state’s history (which came in 2008). Republican frustration with the size of that loss led first to the common refrain that Trump would have won the popular vote if you excluded the country’s most populous state, which is true. (It’s also true that flipping the electoral votes in the country’s second-largest state, Hillary Clinton would have won.)

Then another argument arose: California only went so heavily for Clinton because noncitizens voted illegally. Trump himself raised this idea to aides, The Post reported Wednesday.

As with so many charges of illegal voting, though, there’s no evidence that this is the case.

I reached out to Paul Mitchell of Political Data, a California-based voter data firm. He was generous enough to provide me with breakdowns of the vote in California’s counties and congressional districts, including the number of voters in each district who were born in Latin America and who voted for the first time in 2016.

Because that’s really the argument, and it’s one that Trump himself has made in the past. In October, Trump highlighted a claim from a border patrol agent who said that noncitizens were pouring in to register to vote. The agent later clarified that he had misspoken.

The Post’s Michelle Ye Hee Lee explains why White House press secretary Sean Spicer’s claims about voter fraud in the presidential election don’t add up. (Bastien Inzaurralde/The Washington Post)

According to Mitchell’s data, it’s not the case that immigrants were hustling across the southern border and being handed voter registration forms by ne’er-do-wells. Some 14.3 million people voted in the state last November, up and down the ticket. Nearly 5 million of those voters were Latino — but only 1 million were born in Latin America. Of that group, only 148,000 were newly registered to vote.

That’s about 3 percent of Clinton’s margin of victory in the state. And, of course, there’s no evidence that these voters are anything other than American citizens, casting perfectly legal ballots.

There’s also no real correlation between the density of the vote from first-time, Latin-American-born voters and the presidential election results. Colusa County, which is relatively small, has a high density of such first-time voters. It went for Trump. Imperial County also has a high density of those votes; it went for Clinton. Kern County had a higher density of such voters than neighboring San Luis Obispo County, but Trump won the former and lost the latter.

There’s a statistical measure of correlation that can be applied. The closer an r-squared correlation is to 1, the stronger the correlation. The r-squared value for the relationship between the margin between the presidential candidates in a county and the density of the vote from those born in Latin American countries is 0.09.

Interestingly, the correlation is much stronger between the density of Latin-American-born voters and the margin in congressional elections. The correlation between those two is 0.22 — still low, but higher.

(Worth noting: A national look at noncitizen populations from researchers at Dartmouth found no evidence of illegal voting either.)

There’s strong evidence that the real factor in Trump’s loss in the state was Trump. In seven of California’s 53 congressional districts, Clinton won more presidential votes but a Republican won the House seat. (Presidential results by congressional district are from Daily Kos’ tally.)

The widest swing was in the 21st district in the state’s Central Valley. There, Clinton won by 15.5 points but the Republican House candidate won by 13.4. So did a number of noncitizen voters make the difference, voting only for Clinton? No; 98.6 percent of those who voted in the district cast a ballot in the congressional race. In other words, there was no significant drop-off that would explain the difference — save more affection for the representative than the Republican nominee. That four congressional districts in the Orange County area — a traditional Republican stronghold — went for Clinton and a Republican member of Congress is revealing.

Of course when we talk about margins in California, we’re mostly talking about Los Angeles County. Clinton blew Trump out in the county, and a large percentage of voters there were born in Latin America. But only 1.4 percent of all voters in the county were born in Latin America and voting for the first time there. That would be about 46,000 votes.

This forces those looking for a conspiracy to throw the results to come up with other theories. Maybe those voters aren’t new voters (contrary to the picture painted by Trump). Maybe they’re lying about their nativity. Maybe The Washington Post is doing fake news.

This is why logicians like to point out that it’s incumbent on the person making the unverified claim to offer proof. Trump has promised an investigation into voter fraud that he purports will bolster his claims. Our research, and data from the country’s largest state, suggests that he has his work cut out for him.