Fuqua Insights Podcast: Can Public Companies See What the Government Misses?

Jenny Laurence  00:00
In recent years, there’s been growing concern about how much we can trust public macroeconomic data, like reports on economic growth, inflation, or employment, some economists have even labeled it a macro data crisis, with the potential to undermine policymaking financial markets and public trust. That concern feels especially relevant now, as the ongoing government shutdown has disrupted the release of several key economic reports. I'm Jenny Laurence, an MBA student at Fuqua. Today, I'm joined by Professor Bill Mayew, an expert on how information shapes financial markets and corporate decision making. His research on errors in government economic data examines the impact of firm responses to these flawed signals. Professor, thanks for joining me.

Bill Mayew 00:38
Thank you for having me happy to be here.

Jenny Laurence  00:40
So, whether it's a delay in data releases, as we're experiencing right now with the government shutdown, or a major revision after the fact, moments like these remind us how dependent we are on a steady flow of economic information. So let's start with the key issue being raised. Do you think we're in a macro data crisis?

Bill Mayew  00:56
It is certainly a reasonable question to ask, given what you see happening in the world. For example, in March of 2025, there were articles about the UK's Office for National Statistics’ failure to maintain reliable economic data. The issue there pertained to errors when creating the Producer Price Index and the Service Producer Price index, which both flow into estimates of GDP, or gross domestic product in the United States. In August of 2025, you may have heard about the Bureau of Labor Statistics that was in the news for revising a previously issued jobs number downward. A revision in macroeconomic data means that the originally provided data had an error in it, and essentially that needed to be fixed, right? So it's good practice to monitor the data that we receive from the government agencies that provide them to us and scrutinize any errors. Regarding whether the errors we see suggest we're in some sort of crisis or not, it really boils down to the level and the nature of the errors, not that errors exist in the first place.

Jenny Laurence  01:59
Okay, so you're saying that you will always sort of expect errors to exist in macroeconomic data in the first place.

Bill Mayew  02:05
Absolutely, and to understand why, you have to understand that many of the performance signals you see in the economy are at the most basic level, representing a tradeoff between timeliness and accuracy. And to get some intuition for this, let's go down to the firm level instead of the macroeconomic level. Think about reporting earnings. You read about those in public company financial filings, and one of the first concepts taught to MBA students in introductory financial accounting is that earnings numbers in these reports may not fully represent economic truth. Earnings likely have some errors in it. And why is that? Because as investors, we would rather have a signal of performance that is not perfect but available now, rather than wait a long time to get a perfectly accurate signal. So, here's the intuition. Suppose a firm sells $100 million worth of product on credit, which means the customers get the goods but haven't paid yet. And then suppose also, it turns out that the customers only end up paying $95 million. We have two choices: report earnings now, of 100 million in revenue, and record an educated guess about the $5 million non-payment, knowing it can be wrong; or wait, report earnings of $95 million in cash when it comes in. Unless we can perfectly guess about the $5 million non-payment, which is really difficult, reported earnings today will be more timely, but potentially less accurate than waiting until the future to report the ultimate economic truth of $95 million right? The accounting system is built on the notion that we are willing to live with a less than perfect earnings number because it's provided on a more timely basis, and the same idea holds with data at the macroeconomic level

Jenny Laurence  04:01
Okay, well, that makes a lot of sense. So, if errors are reasonably expected to exist with timely data, as you say, why might some economists characterize this as a crisis?

Bill Mayew  04:13
I think the concept of a crisis here boils down to the nature of the error in terms of both its size and the reason the error occurs. Obviously, the larger the error, the more inaccurate it is. And one might have some threshold in mind about what too big is. But of course, the notion of bigness is a subjective issue. If errors are increasing in magnitude, one would like to know the reason why and how to possibly shrink the errors. The other issue is the direction of the error, and that's typically referred to as bias. So for example, in accounting, when earnings are directionally misestimated on purpose, we call that inaccuracy ‘earnings management.’ In macroeconomic settings, there is some research that suggests purposeful misestimation of macroeconomic data can happen. It typically happens in more autocratic regimes, relative to democratic regimes. But as you've seen recently, in the US, there are some suggestions of political bias possibly playing some role in macroeconomic data releases.

Jenny Laurence  05:15
Okay, so for those that think we are in a data crisis, what is the solution?

Bill Mayew  05:22
If there is one crisis or not, I think it is reasonable and smart to be thinking about how we can make data both more accurate and more timely. When you review potential reasons for errors creeping into macroeconomic signals, one reason is that lots of macro data is based on surveys of constituents in the economy, and response rates have empirically been declining over time, so you're not getting the same amount of data anymore, and therefore it's more likely that errors will creep in. Fixing this problem is hard, of course, because you can't force someone to respond to a survey. So then the question moves to what can be done to replace that missing data? Some have suggested turning to private sector data. So for example, if employment data by the Bureau of Labor Statistics is poor, maybe payroll data from ADP, for example, could be useful to complement or even replace the government data.

Jenny Laurence  06:15
I feel like surveys have always been a tricky thing to encourage people to do.To go back to your research, you focus on GDP, not employment data like the BLS, what made you want to study how firms respond to errors in the government's GDP reports?

Bill Mayew  06:41
GDP is a flagship metric in the economy, and it's been extensively studied in the economics literature. And as an accounting professor, I primarily studied firm communication during earnings conference calls, and I noticed at times, large firms in the economy would sometimes speak about a GDP release and comment on whether their own performance was different from what the macroeconomic data suggested. So as a hypothetical example, if the BEA (Bureau of Economic Analysis) says GDP dropped by 1%, a firm might say, we are seeing growth, not contraction. What that implies potentially is that perhaps big public firms have a unique window into the true state of the macro economy, and if so, maybe big firms can see through the error that is inherent in GDP. If that's true, then maybe already available company data, which is easily available in SEC filings, could be a complementary source of macroeconomic data. But to even consider that idea, though, we have to first know if, on average, firms can actually see through the errors in GDP data. We can get a sense of that by seeing how publicly traded firms respond when GDP signals come out.

Jenny Laurence  07:53
Okay, so for listeners who don't follow data releases closely, can you just walk us through how the government's GDP estimates are created in the first place, and maybe why the early numbers can differ so much from the final ones?

Bill Mayew  08:05
So Gross Domestic Product, or GDP, that's a measure of the market value of all the final goods and services produced within United States in a given period of time. This is a quarterly measure. So if we want to know what happened in, say, quarter one of the year, 2020, about three weeks after that quarter end, the GDP will be released as what they call an initial estimate. Less than half of the information in this estimate comes from survey data during those three months. The rest of it comes from extrapolations and trend data. But then the BEA continues to collect data and receive survey responses over time. With the new data they receive, they revise the initial estimate. The BEA revisits this initial estimate at the one year mark and every five years. What this means is that, as a researcher, we can use the final vintage of a given quarter's GDP estimate as the truth, and compare it to the initial or the first GDP estimate that came out. The difference between those two, you could label the error in the original initial GDP estimate. So as researchers, we can decompose the initial GDP estimate into the true GDP component and the error component. Now, of course, this was not possible to do in real time for firms at the time the initial GDP release came out. But if firms can see through the initial GDP estimates to filter out the error, then we should see their behavior correlates differently with the true component of the initial GDP estimate versus the error component

Jenny Laurence  09:37
Right, which again shows the value of more time to kind of refine the accuracy of the data. So your findings show that firms don't filter out that noise particularly and if we look at what's going on at the employee level, how are managers typically reacting when a report looks strong, and what can some of the consequences of that reaction be?

Bill Mayew  09:58
Well, what we find in the data is that firm earnings, capital expenditures, by firms, production, inventory in a given quarter, all of those are positively associated with the GDP signal from the prior quarter. That means, if the GDP signal is favorable today, next quarter, firm earnings are higher. They invest more in capital, they produce more inventory, and they have more inventory on the shelf. If the signal is negative, we see reductions in all of these items. So importantly, this association is statistically the same for both the true component of the GDP signal and the error component. So if firms were able to filter out the error, we would have expected a smaller effect on the error portion of the initial GDP estimate relative to the true portion.

Jenny Laurence  10:47
Right, thank you. So something I found really interesting about your research is that firm level overreactions can really amplify into larger economic swings. Can you just walk us through how that happens?

Bill Mayew  11:00
Our findings for firm reactions next quarter are similar to what economists have documented at the macro economy level. That is at the macro economy level, when GDP is favorable today, true GDP next quarter increases. However, this is typically where macro economists stop. But we extended our analysis at the firm level to eight subsequent quarters to see what happens. What we found was that for the four quarters after the initial GDP announcement, as the GDP signal increased, firms increased capital expenditures, productions and inventory levels. This stopped at five quarters after the announcement, but in the fifth through eighth quarters, we saw earnings fell. What this implies is that firms initially overreacted to the favorable initial GDP signal. Now why they overreacted is unclear, and it's an empirical phenomenon. Others have documented. More work is needed to understand exactly why this happens.

Jenny Laurence  12:01
Okay, so final question, what would you say is the biggest takeaway for business leaders and policy makers about these errors that may exist in government data like GDP estimates?

Bill Mayew  12:11
I think all we can say from our research at this point is that there's no evidence that firms, on average, can see through the errors inherent in initial GDP signals. So all that means is that we should pause before believing that data from the private sector, such as what we can obtain from public company financial reports, can substitute for what the BEA does. So we need to think of other ways to improve government data, which may be increasingly possible as new and creative ways of measuring economic activity occurs.

Jenny Laurence  12:41
Incredible. Thank you, Professor, kind of sounds like there are some interesting business opportunities out there for the innovators and connectors listening to this podcast, Professor, thank you so much for your time. Really appreciate hearing more about your research and have a good day.

Bill Mayew  Thank you.