Last month, Goldman Sachs economists Jan Hatzius and Kris Dawsey put out a research report arguing US government statistics understate GDP growth because they understate productivity growth. Over the past five years, productivity growth has averaged 0.6% annually vs. 2.6%  over the prior 15 years. Here’s the gist of Goldman’s argument in “Productivity Paradox v2.0″:

— Measured productivity growth has slowed sharply in recent years, and we have reduced our working assumption for the underlying trend to 1½%. This is the same sluggish rate that prevailed from 1973 to 1995 and stands well below the long-term US average of 2¼%. The proximate cause of the slowdown is a slump in the measured contribution from information technology.

— But is the weakness for real? We have our doubts. Profit margins have risen to record levels, inflation has mostly surprised on the downside, overall equity prices have surged, and technology stocks have performed even better than the broader market. None of this feels like a major IT-led productivity slowdown.

— One potential explanation that reconciles these observations is that structural changes in the US economy may have resulted in a statistical understatement of real GDP growth. There are several possible areas of concern, but the rapid growth of software and digital content—where quality-adjusted prices and real output are much harder to measure than in most other sectors—seems particularly important.

— Specifically, we see reasons to believe that the well-known upward biases in the inflation statistics related to quality changes and the introduction of new products are particularly severe for software and digital content. Quantifying the effects is difficult, but it is not unreasonable to think that they could offset a substantial portion of the measured productivity slowdown.

— Our analysis has three practical implications. First, confident pronouncements that the standard of living is growing much more slowly than in the past should be taken with a grain of salt. Second, given the uncertainty around GDP, it is better to focus on other indicators—especially employment—to gauge the cumulative progress of the recovery and the remaining amount of slack. And third, true inflation is probably even lower than measured inflation, reinforcing the case for continued accommodative monetary policy.

Yesterday, JPMorgan economist Michael Feroli and Jesse Edgerton offered a direct counter in the cleverly titled “Do androids dream of electric growth?”:

— Slowing economic growth has prompted speculation that the data aren’t capturing the digital economy

— For this explanation to work one needs to demonstrate that measurement issues are getting worse over time

— There is evidence that there are measurement issues, but little evidence thus far that they are getting worse

— The conjecture that the recent growth slowdown is due to mismeasurement has little empirical support

So two big issues for Goldman. First, calculating the “overall amount of “consumer surplus” for new products is hard. This is an especially troublesome measurement when dealing with “free” digital content. Hatzius and Dawsey: “How important is this issue for GDP measurement? Hard evidence is again hard to come by, but Erik Brynjolfsson and Joo Hee Oh estimated in 2012 that the growth in time spent on free internet sites between 2007 and 2011 created incremental consumer surplus worth ¾% of GDP each year. Although the uncertainty around these types of estimates is obviously very large, as the authors acknowledge, we do think it is plausible that increased new products bias could be holding down the measured real GDP growth rate by a meaningful amount.”

Second, inflation statisticians are better at doing quality-adjustments for computer hardware rather than software. Measured prices of the latter have declined only fraction of the former. [See chart below.] As Hatzius and Dawsey memorably put it, “How much better is Grand Theft Auto V than Grand Theft Auto IV?” Tough to know.

JPMorgan sees things differently. First, while Feroli and Edgerton concede we may be chronically undermeasuring productivity, they are skeptical that bias is getting worse. Cellphones, email, and Internet existed before the Great Recession. The digital economy is not new, but the productivity slowdown is. And do today’s technological advances present “thornier challenges for price measurement than did the introduction of televisions, air conditioning, washer/dryers, electric guitars, dishwashers, microwaves, VCRs, game consoles, and all the rest of the decades-long parade of once-amazing new products that have continually improved living standards”?

Second, why focus on how, say, Uber is an undercounted quality improvement but ignore opposite examples like how airline seats are smaller and delayed departures have increased? Feroli and Edgerton: “Quality-adjusted airfares have likely been increasing much faster than the reported data. Other examples abound. We should be cautious about cherry-picked arguments.

Third, the productivity slowdown is also seen in industries that are easier to measure, such a manufacturing.

So who is right? I think the honest answer is that we aren’t really certain, though I do know that Goldman’s conclusion syncs with the innovation and productivity research of AEI’s Stephen Oliner. More research is needed and, thankfully, is on the way. Policymakers should hope that Goldman is right but act as if JPMorgan is right and pursue needed reforms in areas such as education, immigration, entrepreneurship, and basic research.