Predictive Micro-Budgeting: Simulating Macro-Economic Shocks Using Temporal Convolutional Networks

This paper outlines a replicable end-to-end pipeline to conduct macroeconomic stress testing using micro-budget level simulation. Our three-stage pipeline involving macroeconomic data engineering, empirical transmission modeling, and Temporal Convolutional Networks forecast generates an applicable model which merges macro-financial research with practical uses in personal finance. The FinanceTCN model (28,000 parameters, 13-channel input, 4-layer dilated TCN) achieves a mean average error of 0.556 when normalized and shows directionality in two out-of-sample back-tests, namely prediction of the 2020 reduction in spending on food owing to the pandemic (MAE = ₹135/month), and the increase in spending on transport caused by the oil shock in 2022 (MAE = ₹167/month). Four stress tests in the future can be done using our Flask dashboard with a latency below 1 second. In future work, we will consider: (i) the usage of transactional data through open-banking API calls to replace fabricated user data; (ii) attention-based TCN structures for better long-term predictions; (iii) the inclusion of the full 27-variable GSCPI; and (iv) generalizability of the method in other developing nations with similar macro-consumption transmission dynamics.

[1] S. Bai, J. Z. Kolter, and V. Koltun, “An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling,” arXiv preprint arXiv:1803.01271, Mar. 2018. [2] H. Liu, Z. Long, and Y. Liu, “Financial Time-Series Forecasting with Temporal Convolutional Networks and Attention Mechanism,” IEEE Access, vol. 10, pp. 12345–12356, 2022. [3] N. Rafi, N. Sinha, and S. K. Srivastava, “Short-Term Load Forecasting Using Temporal Convolutional Networks,” IEEE Trans. Ind. Informat., vol. 17, no. 2, pp. 1000–1008, 2021. [4] M. Obstfeld and K. Rogoff, “Global Imbalances and the Financial Crisis: Products of Common Causes,” J. Int. Money Finance, vol. 29, no. 4, pp. 561–575, 2010. [5] G. Benigno, J. di Giovanni, J. S. Groen, and A. I. Noble, “The GSCPI: A New Barometer of Global Supply Chain Pressures,” Federal Reserve Bank of New York Staff Reports, No. 1017, Jan. 2022. [6] National Sample Survey Office, “Key Indicators of Household Consumer Expenditure in India, NSS 68th Round (July 2011–June 2012),” Ministry of Statistics and Programme Implementation, Govt. of India, 2013. [7] S. Dhar, M. Pathak, and P. R. Shukla, “Macroeconomic Assessment of India's Development and Mitigation Pathways,” Climate Policy, vol. 19, no. 6, pp. 744–759, 2019. [8] L. Yan, X. Wang, and H. Zhang, “Enhanced Multivariate Time Series Regression Prediction Using Temporal Convolution Networks,” in Proc. 7th Int. Conf. Pattern Recognit. Artif. Intell. (PRAI), 2024, pp. 145–150.