An Application of the Forecast-Based Interrupted Time Series Methodology to Evaluate Policy Impact Using Daily Data : The Impact of Covid-19 Cash Handouts on Daily Spending

Slow-moving traditional macroeconomic data cannot adequately capture the dynamics of shocks in volatile environments, hence the increasing use of high frequency data, such as daily transactions, and mobility since the COVID-19 pandemic. While difference-in-difference (DiD) estimators are the standard of impact evaluation in non-trial settings, economic policies were typically implemented at the national-level, resulting in a lack of observed control groups. Fortunately, the interrupted time series (ITS) methodology, prevalent in the adjacent field of epidemiology, addresses this challenge using aggregated, high-frequency, time series data. This study applies the ITS methodology to evaluate the impact of three episodes of cash handouts in Malaysia (May 2020, July 2020, and September 2021) on daily spending. This forecast-based ITS setup constructs a post-intervention 'counterfactual' using forecast of daily transactions from an autoregression, with daily mobility as exogenous predictors (AR(n)-X) to account for changes in non-pharmaceutical interventions (NPIs). In all three episodes, the impact on spending is positive, precisely estimated, and dissipates within a month. The spending multipliers were lower when economic conditions were known to be more accommodative. Further research may help policymakers consider the interaction between timing, quantum, and precision (targeted versus broad) of both direct, and indirect stimulus measures on spending