Given the number of questions we have received about the BEA’s new methodology for the portfolio management fees price deflator, we address the main issues in this short Q&A note. (And yes, likely a note circulated by GS is incorrect)
Q1. How does BEA currently estimate nominal PCE spending on portfolio management and investment advice services?
The precise monthly procedure is not fully disclosed in public documentation. The NIPA Handbook (see page 5-28 or 122 of the PDF here) describes a layered estimation process. Benchmark estimates are constructed through the commodity flow method, primarily using Economic Census data. Nonbenchmark annual estimates use receipts from the Service Annual Survey. QSS receipts are incorporated into quarterly estimates when available. The handbook states that advance and second estimates use CES data on employment, hours and earnings. Therefore, a natural interpretation of the CES procedure described in the handbook is
Nominal spending indicator (roughly) = Employment × Average weekly hours × Average hourly earnings
This is conceptually equivalent to an aggregate weekly payroll indicator, although the handbook does not disclose the precise formula, industry aggregation or adjustments used. The surprising part is that we have been in contact with the BEA and they clarified that nominal spending for the advance and second estimates follows “a judgmental trend informed by market indicators”. When QSS becomes available, BEA incorporates the new quarterly information, while the monthly pattern and the estimates for months beyond the QSS period continue to reflect the market indicator methodology.
Our best reading of the current procedure is therefore that QSS anchors nominal spending at the quarterly frequency when available, while an undisclosed market indicator procedure determines the provisional estimates and monthly pattern. In this sense, the confusion arises because what is stated in NIPA Handbook does not match what the BEA has disclosed. As we discuss below, the actual observed data seem in line with the BEA disclosed procedure (a judgmental trend informed by market indicators), not with the NIPA Handbook. The reference to CES employment, hours and earnings in the NIPA Handbook may describe an earlier procedure, one input within a broader judgmental calculation or documentation that has not yet been updated.
Q2. Will BEA revise nominal PCE spending on portfolio management services?
The announced methodological change does not introduce a new methodology or new source data for nominal spending. BEA has confirmed directly to us that the source data used for the nominal estimate will remain unchanged.
Under the current procedure, BEA uses the portfolio management PPIs to measure prices and derives real spending through deflation. Real spending under the current method = Nominal spending / PPI based price index
Under the new procedure, BEA will construct real spending using a CES based quantity extrapolator and derive the price index residually. Implicit price index under the new method = Nominal spending / CES based quantity index. Nominal spending remains the numerator under both procedures and it will remain methodologically unchanged.
Q3. Goldman Sachs argues that the new monthly implicit deflator will essentially equal average hourly earnings in the portfolio management industry. Is that correct?
This result is mathematically correct only if the provisional nominal numerator is extrapolated using aggregate weekly payrolls and the new quantity denominator is based on aggregate hours. Under those assumptions:
Nominal spending = Employment × Average weekly hours × Average hourly earnings
Quantity = Employment × Average weekly hours
The implicit price index would then be:
Implicit price index = Nominal spending / Quantity = Average hourly earnings
Average hourly earnings would therefore emerge as the residual price index.
The uncertainty/confusion concerns the first assumption. As discussed in Q1, the published NIPA Handbook states that advance and second estimates of nominal spending use CES employment, hours and earnings, which in theory supports Goldman’s interpretation. However, the BEA has told us directly that these estimates follow a judgmental trend informed by market indicators and the data (see Q4) support this second interpretation.
The confusion therefore originates from a discrepancy between the published description in the NIPA Handbook and BEA’s more recent direct clarification. Goldman’s algebra is valid, but its conclusion depends on a premise about the numerator that does not appear consistent with the latest information supplied by BEA nor with the data.
Q4. Is monthly nominal PCE spending more closely correlated with CES data or with the stock market?
Using the currently published historical series, monthly growth in nominal portfolio management PCE has a correlation of approximately minus 0.25 (see here) with monthly growth in aggregate weekly payrolls for the portfolio management and investment advice industry. However, its correlation with monthly returns on the NYSE Composite is approximately 0.57 (see here).
The published monthly pattern therefore looks substantially more consistent with financial market movements than with aggregate weekly payrolls.
A caveat applies: The comparison does not conclusively identify BEA’s real time procedure because the latest historical PCE series already incorporates QSS benchmarks, annual revisions, seasonal adjustment and revisions to the monthly allocation. It may therefore differ materially from the provisional estimates originally published before QSS became available. The evidence nevertheless supports BEA’s statement that market indicators influence the nominal monthly path. For this reason, it is difficult to reconcile the observed historical pattern with a procedure in which aggregate weekly payrolls alone determine nominal spending.
Q5. How will you forecast the new portfolio management PCE price in practice?
We will forecast nominal spending and quantity separately. For nominal spending, our central model will use financial market returns to estimate the monthly pattern between QSS releases. The initial specification will use NYSE Composite returns because they have a materially stronger historical relationship with nominal portfolio management PCE than aggregate weekly payrolls. We will also test broader equity indexes, bond returns, lagged returns and monthly average market values where they improve forecasting performance. When QSS becomes available, we will use it to reanchor the quarterly nominal level. We will allow QSS to surprise the previously estimated level rather than require the monthly market model to predict the eventual survey benchmark exactly.
For quantity, our central specification will use CES employment (or employment multiplied by average weekly hours).
The first releases under the new methodology will allow us to test these competing specifications directly by comparing the official nominal, real and price estimates with aggregate weekly payrolls, aggregate hours, average hourly earnings and financial market returns.