Example: Credit card debt (2-quarters ahead)

NY Fed HHDC credit-card balances, univariate ARIMA two-quarter-ahead

This model estimates Credit card debt two quarters ahead (six months) with a univariate ARIMA on the New York Fed Household Debt and Credit series. The New York Fed series is quarterly; the six-month horizon is two HHDC quarters. Macro FRED prints are not used.

Our ARIMA model is

CCDEBT_estimated(t) = ARIMA(2,0,1)(0,1,1)[4] two-quarter-ahead from y through t−2q
and yields R^2 = 92.7% as of October 2023 (n = 71 quarters) in-sample on paired actuals through 2023Q4.
On later paired actuals the same frozen coefficients yield R^2 = 88.6% as of April 2026 (n = 10 quarters) out of sample.

The model is a univariate ARIMA fit on the target series from 2004 through 2023Q4. Each print is a two-quarter-ahead forecast from y through t minus that horizon, using those frozen coefficients. On the leftover paired actuals that frozen fit beats predicting the holdout mean. The first R² above is in-sample through 2023Q4 on the same two-quarter-ahead rule; the second is out-of-sample on later paired actuals.