There’s one question on Washington’s collective mind that hovers above the never-ending swirl of Trump news: Who will control Congress after the midterms?

As with everything in politics these days, there are lots of predictions – and those who manage to be right will be celebrated. But here’s the problem: It’s almost impossible to figure out if a blue wave will happen – and how big it might be – until it happens.

A quick look at House turnover in midterms since 1960 reveals one thing: uncertainty. The average midterm House loss for the sitting president’s party is 22 seats. Yet the actual numbers have been all over the map, ranging from +8 in 2002 for George W. Bush to -63 in 2010 for Barack Obama.

But can’t we look at some key indicators today to figure out what’s going to happen? Sadly, no. Here are just a few “leading indicators” that indicate… almost nothing.

GDP growth was at 4.2 percent in the second quarter – the highest since 2014 – which should help Republicans, right?  Yes, that’s a strong number. But looking at second-quarter GDP over the past 58 years shows no correlation to House gains or losses.  Indeed, Obama’s large loss in 2010 came at a time when second-quarter GDP was 3.7 percent.  And Bill Clinton’s 54-seat loss happened with GDP sitting at 5.5 percent.  By contrast, George W. Bush gained eight seats in 2002 with GDP growth at 2.4 percent.

While third-quarter GDP is slightly more (but not always) related to midterm turnover, we won’t know that number until after the election.

Trump’s approval rating is in the low 40s – which means a blue wave is coming. Possibly – but not necessarily. When the president is sitting above 50 percent approval, his party tends to perform better — with a range of -12 to +8. None of those numbers would flip the House.

So what does 41 percent approval (using the ) mean at this point? Obama lost 63 Democratic members when he was sitting at 44 percent approval at this stage in the cycle. But Jimmy Carter lost just 15 seats with a 42 percent approval in mid-September 1978 – which, if repeated, wouldn’t be enough for Democrats to flip the House. This means…that there’s precedent for either outcome.

Unemployment is at 3.9 percent – so voters will reward Trump and Republicans at the ballot box. Again, not necessarily. The unemployment rate has declined steadily dating back to the early years of Obama’s presidency – and we now have the lowest unemployment rate approaching a midterm since 1966. What happened that year? Lyndon Johnson lost 47 seats in the House.  And there have been years of higher unemployment with minimal incumbent party losses: 1986 (7 percent unemployment, and a loss of just five seats) and 1990 (5.9 percent unemployment, and a loss of just eight seats).

So why do analysts focus on GDP growth, unemployment, and approval ratings? Because they matter – for presidential re-election.  In 12 to 18 months, when it’s time to handicap the 2020 campaign, these will be critical places to start.  But midterm performance has been so volatile that it frequently defies prediction.

Sure, in retrospect, it seemed that 2010 was going to be tough for President Obama – but few thought it would be that bad. And just four years earlier, it seemed impossible for Democrats to take back both chambers – but they did.

We will know more as Election Day grows closer and we understand individual races. But to know the magnitude (or existence) of a blue wave – we’ll only know that after Election Day. And perhaps the most important thing to note: Midterm performance is not correlated with presidential re-election success.

Doug Usher is managing partner for research at Purple Strategies.