Calendar alignment: do not invent monthly observations
Calendar alignment should preserve when a value was measured, what interval it covers and when it became available. A quarterly observation does not become three independent monthly observations because a spreadsheet repeats it across three rows. Choose a mixed-frequency treatment that matches the question and label any carried or interpolated value as a transformation.
Record three different clocks
An economic period, a publication date and a research decision date serve different purposes. Revenue may cover January through March, appear in a filing in May and enter your database in June. A calendar join using only March can place information into an analysis before it was available. A join using only June can conceal the actual operating period.
The FRED series observations API documentation describes frequency aggregation and real-time parameters. Its aggregation choices illustrate why frequency and timing need explicit definitions. They do not justify treating an annual or quarterly figure as newly observed each month, and API defaults are not a substitute for a research design.
A hypothetical quarterly-to-monthly merge
Assume an invented company reports first-quarter revenue of 90 on May 8. You also have monthly shipment counts for January through June. You want to assess whether shipment patterns explain reported sales. Copying 90 into January, February and March creates three copies of a quarter total; summing them would produce 270 and a false monthly pattern.
| Treatment | Meaning | Appropriate limit |
|---|---|---|
| Aggregate monthly shipments to Q1 | Match three-month operating periods | Still check product and geography scope |
| Allocate revenue as 30 per month | An assumed allocation of 90 | Not actual observed monthly revenue |
| Carry Q1 as latest known after May 8 | Information available for later decisions | Keep age and covered period visible |
| Copy Q1 into earlier decision dates | Uses future disclosure | Invalid for a historical information test |
The first treatment is often a sensible diagnostic when the operating relationship is quarterly. The second may support a clearly labeled scenario, but it cannot establish seasonality within the quarter. The third answers an information-availability question and should not be mistaken for contemporaneous monthly activity.
Match aggregation to the variable
Flows such as monthly sales often sum across a quarter. Stocks such as debt outstanding may require an end-period observation. Prices or rates may require an average, endpoint or exposure-weighted calculation depending on the hypothesis. Write that choice beside the series; applying one default aggregation to every column can mix incompatible quantities.
Check calendar coverage before aggregating. A missing month should not silently produce a three-month total from two observations. Count valid subperiods and distinguish partial periods, zero activity and unavailable data. If a series uses business days or a different fiscal calendar, document the mapping and the days included at each boundary.
Control carried values
For a latest-known comparison, use an availability-aware match and retain the original release date. Add an age field and a maximum validity rule justified by the use case. An old value should become visibly stale, not remain permanently current because the join found no newer record. Revisions need their own availability dates if you are recreating historical decisions.
Interpolation introduces a model between observations. It can be useful for a visual aid or a specified estimate, but the new points are not additional evidence. Avoid using an interpolated smooth line to claim precise turning points or a monthly causal link that the source frequency cannot resolve.
Audit before interpreting
For a few rows, trace the raw observation, aggregation method, publication date and output date. Reconcile flow totals to the original quarter and check that no historical decision sees a later release. Finally, describe the resolution the evidence supports. A quarter-level relationship should not become a month-level claim simply because the chart axis uses months.
FAQ
Is forward filling always wrong?
No. It can represent the latest information available after publication. Label it as carried information and disclose its age instead of treating it as a fresh measurement.
Can interpolation recover missing monthly revenue?
It can create an estimate under assumptions. It cannot recreate undisclosed monthly observations or independently validate their pattern.