It is not often that data revisions make national news. The latest employment report showed the economy added 73,000 jobs in July, below the consensus estimate of 104,000. But more significant were the downward revisions to prior months, with the May and June jobs notched lower by a combined 258,000. Revisions are routine, but that kind of adjustment can reshape the underlying picture of labor market momentum.

The White House responded to the disappointing employment data by firing the Commissioner of the Bureau of Labor Statistics (BLS). This move was seen by some as challenging the norms on the government provision of data, leading to potential questions about data integrity going forward, which might concern some investors. 

To understand the concerns, it helps to clarify how economic data is constructed. The data that statistical agencies provide sort into two groups.

1. Early measures of the universe of certain economic activities that are refined over time as data on the universe fill in.
2. Samples of certain economic activities that reflect only a portion of the economy.

The monthly Employment Situation report includes both. Businesses submit payroll data, and households respond to surveys about their job status. These inputs aren’t perfect—some arrive late and others require correction. That’s why revisions to prior months are routine.

In recent years there is a pattern of downside revisions—one possible explanation being that companies who are late to report could be more likely to be suffering business stress than others. But over time, administrative records like tax filings help fill in the gaps. Not all survey questions, however, can be refined over time. For example, reasons for leaving a job or expectations about future work are inherently subjective.

In essence, statistical agencies sketch a picture of economic behavior. Sometimes that sketch is refined as more data becomes available, like the payroll data. Other times, it remains the only version. If the sketch is later corrected, any distortions – regardless of root cause – will eventually be revealed. But if the sketch is final, it becomes the historical record. Politicians may care more about the first draft because it drives headlines, but investors care about the final, hard numbers.

This brings us to another key data point: the Consumer Price Index (CPI). Unlike the employment data, the CPI relies heavily on judgmental imputations and is not revised to align with administrative records. It’s a sketch that never gets redrawn. That is potentially problematic, especially since the CPI serves as the benchmark for the over 2 trillion-dollar Treasury inflation-protected securities (TIPS) market.¹

Emerging-market investors know this story well. There are usually fewer resources devoted to data collection, leading to larger margins of error and/or more frequent revisions. When that happens, the private sector usually provides an alternative. In Argentina, for example, a think tank’s inflation report eventually carried more weight than the government’s. But building trust in alternative data takes time. In the meantime, risk premiums rise.

So how big is the risk premium? One way to gauge it is through measures of political uncertainty, which track media mentions of economic policy terms. Concerns about data integrity add to that uncertainty. Historically, these measures correlate with equity volatility and credit spreads—as uncertainty goes up, volatility tends to increase, and spreads widen. Yet over the past nine months, those correlations have weakened even though measures of policy uncertainty have increased. Investors may have become receptive to risk taking—perhaps fatigued by the constant churn of headlines. 

Official data and its accuracy are foundational to global financial markets. If investors were to begin to question the numbers, it is not just a headline risk—it’s a structural one. And structural risks have a way of resurfacing when least expected.