The Use of Real Time Spending Sources for Economic Measurement: An Analysis Using Multiple Data Sources

The rise of big data in the past 20 years has opened up the possibility of using new types of experimental data to measure consumer spending. Despite the promise of card transaction data, concerns remain about their representativeness and stability. This study makes a novel contribution by integrating transaction data from multiple sources to measure consumer spending, rather than relying on a single vendor. By blending several card datasets, the analysis mitigates the noise or biases specific to any one source. Three core applications help assess the robustness of these new data sources: (1) evaluating their correlation with established national benchmarks, (2) testing their forecasting ability for official measures, and (3) comparing spending trends at finer geographic levels. In all three applications, I find evidence of the potential usefulness of these sources to provide meaningful economic signals. Additionally, I find evidence that these data sources are complementary, finding a stronger signal when multiple data sources are combined, relative to when a single source is applied. However, considering any single source at the national level, I find several of the data sources provide a similar economic signal, demonstrating that the data sources are potential substitutes.