Alternative data is growing in importance for the financial community. How can alternative data be leveraged from a macroeconomic perspective?
- Alternative data's use in finance is gaining traction, particularly since the pandemic, and even by major players such as central banks.
- Done methodically and cautiously, using alternative data as an input into the decision-making process can increase precision and speed.
- Alternative data can even lend a hand to environmental, social, and governance theses.
"I am informed by Commissary White that you have done well by the early information which you had of the Victory gained at Waterloo.” —John Roworth, 19th century courier, in a letter to the Rothschild family
Though the Rothschilds were already one of 19th century Europe’s richest families, their fortune increased greatly following the Battle of Waterloo. On June 19, 1815, a courier brought news of the British victory back to the family two days ahead of the British government finding out the news.1 Though details remain contested, what is clear is that the Rothschilds leveraged this single yet crucial data point, buying assets in the British markets and further increasing their fortune.
For any investor, the ability to procure, analyze, and act on data can make or break an investment thesis. Properly analyzed, data is transformed into information, and information is paramount to the investment decision-making process.
Data, however, is constantly evolving; while some series have existed for centuries and continue to be relevant, our increasing reliance on technology has brought about new ones that initially seem somewhat unorthodox. These alternative data sets have often been embraced by both buy-side and sell-side analysts in their constant search for alpha, but have become particularly important since the beginning of the COVID-19 pandemic.
How has the use of alternative data evolved during the pandemic? How can it be leveraged for economic outlooks, and what are the implications for investors?
What is alternative data?
In the financial world, alternative data abounds, particularly in security analysis—most famously, the use of satellite imagery to predict stock prices in the retail sector gained traction among hedge funds and then by sell-side analysts and buy-side fund managers. Today, geolocation, social media, mobile app usage, credit card payments, and even the weather can be used as inputs in the development of financial models.
Traditionally, economists have used a host of publicly available and widely accepted data series such as GDP, the Consumer Price Index (CPI), and the unemployment rate to produce forecasts and outlooks. These series, usually provided by governments and statistical agencies, are tried and true, and have been mainstays of economic decisions for decades
But at a time when information is more readily available than ever before, more traditional economic data sets, which are typically released on a monthly or quarterly basis, are beginning to show their age. As a result, although they’re still the main input into most economists’ decisions, their usefulness as a tool for economists leaves room for improvement.
Enter some fairly new data series, many of which are provided by the private sector: Google mobility trends, the number of people passing through TSA checkpoints or New York City subway turnstiles, and OpenTable restaurant reservations. These are just some of the sets of information that, just a few years ago, weren’t being incorporated into robust economic modeling. Recently, however, they’ve become more and more important to investment analysis and decisions. These new data series provide information that is not only different from traditional data sets, but also more timely, often released on a weekly or daily basis.
|CPI||TSA passenger checkpoints|
|Nonfarm payrolls||New York City subway turnstiles|
|Jobless claims||OpenTable restaurant reservations|
|Housing starts||Traffic indexes|
|Retail sales||Private sentiment surveys|
|Industrial production||Satellite images of activity|
Alternative data can provide us with better precision
With the repercussions of the pandemic growing by the day early in 2020, monthly and quarterly data proved too lagged or too imprecise to help pinpoint true inflection points of the economy. Our team therefore began incorporating some of the alternative types of data to help guide our decisions, most notably, mobility data produced by Google and Apple, which helped demonstrate when workers returned to offices, as well as retail activity and public transportation.
With more and more financial market participants using higher frequency data, an interesting consequence has been the reduced value of more traditional but lagged economic data. For example, the shift toward higher precision survey data about job growth can mute the value of traditional monthly jobs reports, such as the U.S. Bureau of Economic Analysis’s Nonfarm Payrolls report. While these data points remain highly relevant to markets, an expanded focus on alternative data can increase our ability to predict and front-run these data prints.
In addition, the COVID-19 recession and immediate aftermath have muddied our ability to get precision reads from traditional data, which has been heavily influenced by base effects, pent-up demand, and stimulus measures. This is why large beats and misses on economic data prints (i.e., how data comes in relative to expectations) have become less relevant to markets than they would have in the past. While we believe this phenomenon will normalize by 2022/2023, the value of higher precision and more timely alternative data is amplified.
Policymakers broaden their data sets
Economists produce a range of forecasts but often place a great deal of emphasis on their expectations for GDP because it has been historically considered one of the better proxies for total growth in an economy. To do so, economists break down GDP into a range of subcomponents, including consumption, investment, government expenditure, and net exports.
And yet, the COVID-19 recession demonstrated that our long-standing view of GDP as representative of the total economy was perhaps misguided. In 2020/2021, GDP roared back to life even as unemployment rates remained extraordinarily elevated. This was in part because of historically large levels of stimulus, but also because GDP was disproportionally representative of wealthier households and larger firms. Meanwhile, inflation measures such as the CPI suggest there is limited inflation in the system even as home prices, commodities, and food prices rocket higher, largely because of how this index is composed. That doesn’t make the calculation incorrect, but it does suggest an important policy question: Are GDP and CPI the best measure of the general economic health of the total population?
Central bankers and national statistics providers might be wondering the same thing and continue to develop their focus on measuring the economy differently:
- In the minutes of the July 28, 2020 FOMC meeting, U.S. Federal (Fed) Reserve Chair Jerome Powell cited “restaurant dining, hotel accommodations, and air travel” and “high-frequency indicators (such as credit and debit card transactions and mobility indicators based on cellphone location tracking)” as having been used to help gauge the state of the economy.2
- In May 2020, the Dallas Fed created its Mobility and Engagement Index using mobile geolocation data in order to provide “real-time insight into the economic impact [of the pandemic],” calling the index “a key metric in forming our assessment of economic conditions and the outlook for future activity.”3
The Dallas Fed's alternative data index tracked economic activity well at the height of the pandemic
New Mobility and Engagement Index vs. traditional Weekly Economic Index
Source: Federal Reserve Bank of Dallas, as of May 13, 2020.
- The St. Louis Fed developed a model estimating employment changes using daily employment data from a free, private sector scheduling and time tracking tool.
- According to Statistics Canada, “By March 2020, approximately half of all prices used in the calculation of the CPI were collected through some form of alternative data source.”4
- Since July 2020, Statistics Canada has provided an adjusted price index—albeit an unofficial one—with a notably increased weight on food and household operations and furnishings, at the expense of transportation and recreation (changes in spending patterns which we can all relate to). As of February 2021, that index has shown inflation to be 0.4% higher than the official data suggests.5
Canada has adapted to the new data reality
Canadian official CPI vs. adjusted price index
Source: Statistics Canada, as of April 12, 2021.
To be clear, we still think GDP is by far the best gauge of an economy and the most important data point we have as economists. But if central bankers and policymakers are looking at traditional data differently and are considering the use of alternative sets as outputs, then it behooves economists (and other financial market participants) to change the way we analyze traditional data and incorporate the use of alternative data in our estimates as well.
Trust but verify your alternative data
The caveat to the use of any data, whether alternative or otherwise, is that the output of your model (for example, an economic forecast) can’t be any better than the underlying data itself.
With so many different data sets,6 the danger lies in finding seemingly predictive patterns in them, and then blindly using those signals instead of analyzing them to ensure they’re true signals rather than just noise. In 2011, a hedge fund and then two years later a trading platform launched with a strategy based on research demonstrating an ability to use Twitter sentiment to predict movements in the Dow Jones Industrial Average with 87% accuracy7; the hedge fund closed a month after opening, and the trading platform is now defunct.
By definition, alternative data sets are new and unproven, and without a long history, we often can’t accurately back test their usefulness as an economic indicator. This makes their use even more perilous, so while we think alternative data sets are the wave of the future, we can’t give them blanket reliability; they require an additional level of scrutiny and analysis when developing them for an economic use case. Still, we do believe that, in many cases, these high-frequency metrics get us closer to the economic reality of the present than the standard fare of more traditional data sets.
ESG and alternative data go hand in hand
Our increased integration of environmental, social, and governance (ESG) aspects into our process also requires thinking outside the traditional data box.
For example, as we develop our long-term GDP forecasts, our team now incorporates a one-degree increase in global temperature and models a variety of climate-related scenarios. We also need to incorporate the effects of more severe weather at a higher frequency, which means we have to keep an eye on climate data.
Even data that would generally be considered traditional like the labor force participation rate requires a different lens in this context; for example, an increased desire to support early learning and childcare among several developed nations such as Canada and the United States implies we must monitor for the rise in female labor force participation rates, which can boost GDP in a statistically significant way.
While these are just a few examples, we view ESG data as an alternative and yet equally crucial input into our decision-making process, and expect that others in our field would agree.
From alternative to essential
Beyond their courier bringing them news from Waterloo, the Rothschild family’s avant-garde use of carrier pigeons to communicate market opportunities and prices was key to their financial growth. The Rothschilds likely weren’t the first investors—and certainly won't be the last—to use nonfinancial data to their financial advantage.
Though alternative data’s prominence was already on the rise, the global pandemic accelerated that shift drastically. Now, it’s no longer ancillary to economic decisions—properly analyzed and scrutinized, it has become essential to our processes and, we believe, to other economists.
1 The Rothschild Archive, 2021. 2 Federal Reserve July 29, 2020. 3 Dallas Fed, May 21, 2020. 4 Statistics Canada, February 17, 2021. 5 Statistics Canada, April 12, 2021. 6 For example, one NASDAQ-owned alternative data provider hosts over 25 million data sets. 7 “Twitter mood predicts the stock market,” Johan Bollen, Huina Mao, Xiaojun Zeng, Journal of Computational Science, March 2011.