A client kindly pointed out to me a paper published this past week at the National Bureau of Economic Research on Business Applications as a Leading Indicator.
The paper is quite involved, and its ultimate objective is to show that applications to register businesses that have a “high propensity” to hire (as opposed to just provide self-employment) add significant explanatory power to a broader model using standard monthly business cycle indicators to predict nonfarm employment growth. I will not discuss that model here because its estimates are out of date and because it would be a great chore to replicate it. But in this note I offer brief and very preliminary thoughts on the paper and some simple pictures of the updated data.
Source: Federal Reserve Bank of St. Louis (FRED), FH calculations
Data are actual to October and seasonally adjusted.
Away from the development and estimation of the model, the paper makes two descriptive points that seem particularly striking:
- Business applications are highly predictive of actual business formations, which is a major advantage because business formations play a major role in the business cycle but are reported with an extensive lag. See the discussion beginning at the bottom of page 7 of the linked report for documentation of how applications reliably predict formations.
- Considered in isolation, so-called high-propensity business applications (HBA) are neither the most tightly correlated with subsequent employment growth nor the longest leading indicator of that. Retail sales are substantially better correlated, although with a much shorter lead time. And single-family new home sales are both better correlated and lead by just slightly more. But HBA provides the second longest lead, among the high-profile indicators tested, and has a reasonably high correlation. Table 5 on page 29 of the report documents this and is largely self-explanatory.
The paper finds that the correlation between the 12-month change of the HBA and the 12-month change of employment growth peaks when the former is lagged 11 months. And the chart below shows a history of the two series updated to the current period, although with no lag imposed and with much of the Covid period censored to avoid distraction from the noise there.
Source: Census, Bloomberg, FH calculations
Applications data are actual to October. Employment data are actual to November. Note that much of the Covid period is censored to avoid distraction created by noise there.
The choice of dual vertical scale is subjective. But even taking that on board, it is obvious that the 12-month change of the HBA looks somewhat low relative to its own history. This might be taken as evidence that the HBA – taken in isolation – predicts a further deceleration of employment growth. But this raises what would be my main criticism of the paper, although admittedly a tentative one.
It is not obvious to me that the flow of new business formations, as predicted by applications, should be taken as a rate of change when being related to employment growth. To some extent new business formations create employment growth. Or perhaps new business formations expressed as a deviation from some sort of mean create employment growth. For example, note in the chart below that the level of the HBA is related to the pace of gross hiring from the JOLTS report.
Applications data are actual to October. Employment data are actual to November. Note that applications are here shown as a 3-month moving average and that the Covid period is largely censored.
To stake a strong claim about this would require a lot more work. But I would not take it as given that the HBA being lower than its quite high level 12 months ago is a long leading indicator of a downturn of employment growth. My tentative conclusion would actually run the other way, from a recognition that the HBA looks fairly high – and would continue to do so even if we normalized it to the level of employment, although the charts do not depict that. I will continue to monitor this series and may express a stronger view when I have one. For now, I would take it as marginally pro-expansion.
