When you hear “They say the economy is on the way back up,” who is “they” and why do they say that? There are so many components from so many industries that produce so much data that it can be hard to know the actual state of the economy. One measure is the gross domestic product (GDP), which is the total value of all goods and services traded in the country.¹ Economists who try to make sense of uncertainty and make informed analyses about the state of the economy are called nowcasters: formally known as macro-econometricians, casually known as magicians.
While calculating all the components of the GDP might sound like guesswork, I’m willing to bet you already rely on similar “guesstimates” without even thinking about it. If you were attending a wedding at a venue you’ve never visited, you’d probably pull up Google Maps the day before to see the time needed to make the drive. Google Maps looks at the baseline time for your route’s mileage then looks how factors such as weather and traffic have affected trips in the past to make an educated guess about how they’ll affect your trip in the present.
Nowcasters do the same thing. We use data about trade, consumption, and productivity to make educated guesses about how current conditions are affecting the current GDP.² The economy happens in real time, but the GDP is not just a simple accounting sheet. It’s literally unknowable.
To arrive in time to hear “I do”, you’d make plans based on Maps’ estimated travel time, perhaps topping up on gas before your trip and planning when to eat your last meal.³ Our country’s monetary policy makers aren’t rushing to a wedding, but they do need to know the current state of the economy and what’s soon to come in order to make the right policy choices. Rather than turning to the sentiment of their peers, or, heaven forbid, news pundits, they turn to the Google Maps of the economy: GDP nowcasts. Understanding the processes, strengths, and limitations of Google Maps can give us a better understanding of the inner workings of macroeconomic nowcasting.
- Google Maps updates its estimates as you travel. Current conditions, by nature, will change as your trip progresses. As that traffic jam hundreds of miles away clears up, you’ll shave some minutes off your ETA, but a flipped semi just a few miles up the road will add them right back. A nowcast of GDP updates as new data becomes available, of which much is released on fixed schedules from government bureaus.
- Google Maps’ algorithm can assign “weights” to its data inputs to account for differing importance. Weather has less impact on getting you to the venue than does gridlock traffic leaving town on a Friday afternoon. GDP nowcasts also assign weights because components differ in impact. Net exports and industrial production are far more important in determining the GDP than, say, the amount of weddings rings sold. These weights allow us to decompose the overall estimate to evaluate the influence of each component.
- Google Maps’ accuracy increases as you near departure. The calculated travel time will be more accurate if you leave right now than the travel time would be in the future. It can’t provide a pinpoint ETA for that wedding next summer, even for a known route, because there will be uncertainty in weather, construction, or traffic.⁴ Likewise, nowcasts of the GDP decrease in accuracy the further out you look. A model is only as good as the data on which it’s built, and we have a better idea of current conditions than the future. Observation turns into speculation and a nowcast turns into a forecast when quarters turn into years.
- Google Maps cannot predict future travel times because processes change. Change will come not from current conditions but also to on a fundamental, traffic-generating level. Autonomous vehicles have already begun to disrupt driving habits and flying cars would challenge the notion of traffic as we know it. So too do the factors of economic production change, making long term predictions unreliable. “Fintech” and the “gig economy” have shifted how money is made and moved in the past decade. The further out we look, the less sure of the generating structure we can be.
- Google Maps could be more useful by providing a range of expected arrival times rather than a single number. Communicating your arrival time in a 15-minute window would relieve your host’s anxiety when it’s 10 minutes after you said you’d arrive. A GDP nowcast includes a range of the possible actual values, providing a better picture than just a single number.
- Google Maps’ ETA wouldn’t necessarily be in the center of this range, though, since the distribution of travel times is likely skewed. Shaving minutes off your trip is quite hard; adding, easy. Otherwise we’d find ourselves just as likely to be early as late!⁵ The GDP nowcast’s range, known technically as a confidence interval, usually isn’t uniform — though it’s not likely to be as skewed as travel times. Sometimes it may be nearly symmetric, but sometimes the most likely outcome is closer to one end of the range than the other.
Macro-econometricians have an even harder job than Google Maps, though. An ETA is composed of factors that change with time but are often frequent, easily observable, or both. GDP, on the other hand, is affected by data that are either released well after occurrence (median home sale prices), unobservable (the natural rate of unemployment), difficult to measure (total inventory count in every factory and store in America), or some combination.
While you can instantly calculate the duration of your road trip after arrival, you cannot instantly know the GDP at the end of a quarter. Instead, the Bureau of Economic Analysis rolls out estimates of what the past quarter’s GDP might have been: an “advance” a month after the quarter ends, a revision the next month, and a “final” version the month after. Throw in an update every summer of the previous year’s estimates for good measure. Oh, and they can amend previous years’ GDP figures during the summer, too. Unlike the number of unemployment claims or the S&P 500 index, GDP is neither in stone cast nor directly observable in real time.
This is where the magic of macro-econometrics is revealed. We do our best to estimate what the Bureau of Economic Analysis will estimate the GDP to be for the current quarter, knowing that this value is neither known nor knowable. The goal is to understand the current state for the sake of appropriate monetary policy; the far future is not of tremendous concern. Nowcasters seek to create useful, not necessarily perfect, estimates. 20th century British statistician George Box reminds anyone who has used Google Maps to get to the right place but still been late that “all models are wrong, but some are useful,”⁶ and John Kenneth Galbraith keeps nowcasters themselves humble: “economic forecasting exists to make astrology look respectable.”⁷
 This is a common aggregate measure but cannot measure the well-being of individuals. Back to paragraph.
 These educated guesses may come in the form of state-space dynamic factor models, but the term “guesstimate” works too. Back to paragraph.
 Or planning to leave a few minutes early, knowing that at least one person in your party will be making some last-minute fashion decisions. Back to paragraph.
 Whether or not you’d look at travel times that far in advance comes down to personality, along with the happy couples’ choice to send invites a year out. Back to paragraph.
 If you think you are an outlier due to immaculate driving skills, consider recording “early” or “late” relative to the initial Google Maps ETA for your next ten trips. The author does not condone reckless driving in the name of data collection. Back to paragraph.
 Dr. Box wedded a daughter of statistician R.A. Fisher, and was a Fellow of the Royal Society and a member of the American Academy of the Arts and Sciences. Back to paragraph.
 Dr. Galbraith advised four presidential administrations, spent nearly 6 decades as Harvard faculty, and received both the World War II Medal of Freedom and the Presidential Medal of Freedom. Back to paragraph.
Michael Boerman is a Research Assistant for the Board of Governors of the Federal Reserve System. The opinions and views expressed in this article are solely those of the author. They do not purport to reflect the opinions, views or policy of the Federal Reserve System or its members.
Many thanks to Eric Boerman and Cameron McWilliams for their edits and insight. Any remaining errors are are those of the author alone. The original idea for likening Google Maps to economic forecasting comes from John Kay and Mervyn King in Radical Uncertainty.