In recent years, economists have embraced empirical research that is yielding ever more convincing, conclusive information about the economy. This has allowed economists to test long-standing theories, some of which have been found wanting.

Sadly, though, applying these methods to macroeconomics is almost impossible. When you’re dealing with the movements of the whole economy, everything is related to everything else, and history only happens once. That makes it very hard to test macroeconomic theories.

But macroeconomists still try! A good example is the so-called real business cycle (RBC) theory, which holds that random changes in productivity — due to variations in government policy or the rate of technological innovation — drive recessions and booms. In 1999, economist Jordi Gali tried to test this idea by measuring changes in the rate of innovation. He found that faster innovation tends to be correlated with higher unemployment — exactly the opposite of what RBC theory predicted.

Of course, Gali’s paper isn’t case closed, since it relied on many assumptions in order to identify the rate of innovation, and those assumptions could be wrong. And even if Gali’s analysis was flawless, a modified version of RBC theory might still be right. In any case, negative results like Gali’s weren’t enough to stop RBC theory from receiving a Nobel prize for its creators in 2004.

In recent years, despite all the challenges, macroeconomists have redoubled their efforts to match theory to data. A recent paper by Michael McLeay and Silvana Tenreyro develop new ways to identify the Phillips curve, a theoretical relationship between inflation and unemployment. In another study, Cristiano Cantore, Filippo Ferroni, and Miguel León-Ledesma try to identify the effect of monetary policy on labor’s share of national income, using methods not that different from Gali’s. These are just two examples among many.

Other researchers examine the pieces that make up the macroeconomy. For example, a recent paper by Princeton University’s Arlene Wong found that much of the effect of interest rates on consumption comes from young people buying houses or refinancing their mortgages. Another paper by Mark Bils, Peter Klenow, and Benjamin Malin finds that prices tend to change over the business cycle more than wages do. Again, there are many, many more examples.

This kind of research is not dramatic, headline-grabbing stuff. It doesn’t give instant answers to the big questions of macroeconomics — how to minimize the risk of recessions, how to fight recessions when they occur, how to avoid inflation, and so on. It doesn’t easily lend itself to investment ideas for asset managers. It can’t easily be adopted in a political platform.

What it is, is science. Not the clean, idealized lab science of physics or chemistry, or the simple, convincing studies of microeconomics. But as messy and limited as it is, empirical macro represents a real, honest, ongoing attempt by dedicated researchers to explore the ins and outs of a hideously complex and hard-to-measure system. Someday, thanks to their modest, diligent efforts, we will probably understand the phenomena of booms and busts much better than we do now.

But in the meantime, policy makers have to make policy, and investors have to invest. Instead of relying on a single grand theory, central bankers tend to use a combination of judgment, real-time data and an array of different models. But elected politicians looking for a bold policy program and asset managers looking for a bold investment thesis are frequently tempted by heterodox economic theories claiming to have simple, sweeping solutions. Two examples are Austrian economics and modern monetary theory.

These movements call themselves theories, but they don’t look much like the mathematical models used by macroeconomists. Austrian economics explicitly rejects math models, while MMT employs them only extremely rarely, preferring to limit its math to accounting definitions. Instead, both schools of thought tend to rely on logical arguments, stated in English.

Writing theories in English makes them hard to test with data. Unsurprisingly, both Austrians and MMTers only rarely engage in quantitative empirical analysis. Some supporters of these movements tend to point to macroeconomic events as vindication of their ideas, such as when self-described Austrian Peter Schiff claimed to have predicted the crash of 2008. But this is not a very good way of testing a theory — often, it takes only a little investigation to reveal more failures than successes.

But the lack of formal models and empirical tests doesn’t stop unorthodox advocates from issuing audacious policy advice. For Austrians, this can include switching to a gold standard or abolishing the central bank, while for MMTers it can mean vastly increased deficit spending and a federal job guarantee.

The boldness and confidence of these predictions — and the comparative lack of math — makes these theories much more attractive to politicians and investors compared to the turgid scribblings of professors. But this is a double standard. The fact that a theory is written in plain English might make it feel more comprehensible, but it doesn’t make the actual economy any less complex. And the fact that decades of painstaking empirical work by humble, careful, smart academics has only modestly advanced their understanding of this complex system makes it seem dubious that a small group of outsiders could swoop in and solve the whole thing with a few feats of logic and artful storytelling.

In other words, both policy makers and politicians should be reluctant to embrace the sweeping claims of unconventional theorists. Instead, a cautious approach, relying on judgment, data and an eclectic mix of theories, seems best.