I’m pretty down on economists these days. Actually, it’s “authoritative” public-policy macroeconomists that most invoke my ire. I’m not as bad as Nassim Taleb though. Among Taleb’s many econo-bombs, one from “Fooled by Randomness,” the first of his best-selling trilogy, is a favorite: “Economists are evaluated on how intelligent they sound, not on a scientific measure of their knowledge of reality.”

Relatedly, I have two beefs with policy-promoting macroeconomists. The first is that many consider their discipline a hard science, when it’s in fact anything but. Physical scientists often condescend that economists suffer from physics envy. And, as if to prove that, macroeconomics has evolved to very mathematically-sophisticated models, most of which prove quite naïve – as was amply demonstrated four and a half years ago. It’s a damn shame reality gets in the way of elegant models. Indeed, if you wish to study economics at the graduate level from a top school now, you better have the equivalent of an undergraduate math degree to even apply.

At the same time, what many economists don’t acknowledge is that they’re very much political creatures whose views slant their analyses. Tell me an economist is politically conservative, and I’ll tell you she’s a neo-classicist, with an abiding faith in rationality and unfettered markets. Tell me she’s liberal, on the other hand, and I’ll assuredly bet she’s a Keynesian, conceding imperfect markets and the need for government intervention to fight recessions. Knowing the “political economics” tells you a lot of what the practitioners think about human nature as well.

Over the last few years, the “neos” have promoted a belt-tightening, austerity resolution to the economic malaise, while stimulus-championing Keynesians have complained that the dole four years ago was woefully inadequate. The U.S. and Europe seem stuck in a high-unemployment, low-growth quagmire, neither the stimulus of the U.S. nor the spending cuts of Europe producing significant results. Yet both sides seem resolute that they’re right and their opponents wrong. Even as a life-long Keynesian, I can only say bunk – to both sides. Maybe the “science” of macroeconomics isn’t so scientific after all. And maybe some of what we’re experiencing is in fact “just one damned thing after another.”

As I was about at wits end with economic frustration, I stumbled across a public policy brief from the Federal Reserve Bank of Boston that gives me hope for the discipline – and at the same time provides guidance on how to conduct analytics in business.

The two Northeastern University scholars, Rand Ghayad and William Dickens, set out to evaluate the efficiency of the labor market, assembling data on the relationship between monthly percentage unemployment and job vacancy rates. One would obviously expect an inverse relationship in this so-called Beveridge curve: as job vacancies rise, unemployment falls.

What the authors found in the monthly data from January 2001 to June 2012 was that the relationship was quite well behaved – inverse curvilinear – up to September 2009, the height of the recession. At that point, there was a shift “up” in the relationship – higher unemployment associated with higher vacancy rates. Though such an upward shift is not unusual at the start of a recession, its persistence over time remains puzzling.

Somewhat surprising were the findings when the data were decomposed or dimensionalized by length of unemployment. “The relationship between short-term unemployment and vacancies is unchanged. Thus, all of the increase in vacancies relative to unemployment has taken place among the long-term unemployed.” In other words, the culprit for the current aberrant Beveridge relationship appears to be those unemployed for longer than 26 weeks. The findings persist across potentially confounding factors such as age, education, blue-collar/white collar and industry, though it would have been nice to see the curves split by the combination of short-term/long-term unemployment and those dimensions.

To further elucidate the research, Ghayad conducted a randomized experiment in which he sent 4,800 fictitious applications for 600 real job openings. The independent factors included length of unemployment, job history and experience. What he found was the “resumes” of the long-term unemployed received callbacks at significantly lower rates than the short-term unemployed – even when they presented more pertinent job experience. “…there are good reasons to suspect that the loss of skills and connection to the job market provides a significant and lasting impediment to long-term unemployed reentering the workforce … puts the spotlight more on structural and fiscal policies, including measures aimed at helping those who are unemployed to find jobs.”

In BI parlance, the economists in this intriguing study identified a relationship between unemployment and vacancy rates. They then decomposed that relationship, dimensioning or “disaggregating” by potentially confounding factors. The one factor that provocatively stood out was length of unemployment – the well-behaved Beveridge relationship queered by the chronically unemployed. The randomized experiment to shed light on the “causal” conjecture behind chronic unemployment was icing on the methodological cake. BI and analytics professionals could do a lot worse than to consider this analysis of decomposition and experimentation as a model for their craft.