| Trenzy Odero, Mostafa Zahed |
This study conducts a time series analysis of quarterly air passenger traffic at San Francisco International Airport (SFO)
from 1999 to 2025, emphasizing seasonal patterns, long-term trends, and disruptions caused by the COVID-19
pandemic. Publicly available data from SFO is analyzed using SAS tools including PROC TIMESERIES for exploratory
analysis, PROC ARIMA for SARIMA model estimation and diagnostics, and PROC ESM for comparing exponential
smoothing techniques. To address non-stationarity, both first-order and seasonal differencing were applied. Based on
autocorrelation diagnostics, a SARIMA(0,1,1)(0,1,1)4 model was selected and validated using the Akaike Information
Criterion and residual analysis. Although point forecasts performed well, residuals deviating from the normal distribution
assumption limited the reliability of interval forecasts. Out-of-sample projections through 2027 suggest a steady
post-pandemic recovery in air travel, with consistent seasonal peaks in Q3. Confidence intervals widen over time,
reflecting increased long-range uncertainty. Future improvements may include incorporating exogenous variables and
adopting non-normal modeling frameworks. This analysis highlights the power of SAS for applied forecasting and offers
valuable insights for aviation analytics and operational planning. |