Nowscasting中的参数扩散：问题与途径——在中国实际GDP Nowscasted中的应用

Parameter Proliferation inNowcasting: Issues andApproaches An Application to Nowcasting China’sRealGDP Paul Cashin, Fei Han, Ivy Sabuga, Jing Xie, and Fan Zhang WP/25/217 IMF Working Papersdescribe research inprogress by the author(s) and are published toelicit comments and to encourage debate.The views expressed in IMF Working Papers arethose of the author(s) and do not necessarilyrepresent the views of the IMF, its Executive Board,or IMF management. 2025OCT IMF Working Paper Institute for Capacity Development Parameter Proliferation inNowcasting: Issuesand Approaches—An Application toNowcastingChina’sRealGDPPrepared byPaul Cashin,Fei Han, Ivy Sabuga, Jing Xie,andFan Zhang* Authorized for distribution by Natan EpsteinOctober2025 IMF Working Papersdescribe research in progress by the author(s) and are published to elicitcomments and to encourage debate.The views expressed in IMF Working Papers are those of theauthor(s) and do not necessarily represent the views of the IMF, its Executive Board, or IMF management. ABSTRACT:This paper evaluates three approaches to address parameter proliferation issue in nowcasting:(i) variable selection using adjusted stepwise autoregressive integrated moving average with exogenousvariables (AS-ARIMAX); (ii) regularization in machine learning (ML); and (iii) dimensionality reduction viaprincipal component analysis (PCA). Utilizing 166 variables, we estimate our models from 2007Q2 to 2019Q4using rolling-window regression, while applying these three approaches. We then conduct a pseudo out-of-sample performance comparison of various nowcasting models—including Bridge, MIDAS, U-MIDAS, dynamicfactor model (DFM), and machine learning techniques including Ridge Regression, LASSO, and Elastic Net topredict China's annualized real GDP growth rate from 2020Q1 to 2023Q1. Our findings suggest that theLASSO method outperform all other models, but only when guided by economic judgment and sign restrictionsin variable selection. Notably, simpler models like Bridge with AS-ARIMAX variable selection yield reliableestimates nearly comparable to those from LASSO, underscoring the importance of effective variable selectionin capturing strong signals. Parameter Proliferation inNowcasting:IssuesandApproaches An Application to Nowcasting China’sRealGDP Prepared byPaul Cashin,Fei Han, Ivy Sabuga, Jing Xie,andFan Zhang1 Contents A.Variable Selection..................................................................................................................................7B.Regularization in Machine Learning (ML) Models..............................................................................9C.Dimensionality Reduction using Principal Component Analysis (PCA)........................................10 IV. Application: Nowcasting China’s Real GDP During COVID-19...............................................................12 A.Data Preparation..................................................................................................................................13B.Methodology........................................................................................................................................15C.Results and Evaluation.......................................................................................................................16 V. Conclusion....................................................................................................................................................23 Annex I. Data Description.................................................................................................................................24 References.........................................................................................................................................................27 FIGURESFigure 1. Model Performance: Approach #1 AS-ARIMAX.............................................................................18Figure 2. Model Performance: Approach #2 ML Regularization...................................................................19Figure 3. Model Performance: Approach #3 PCA..........................................................................................20 TABLESTable 1. Forecast Evaluation Statistics...........................................................................................................21Table 2. Selected Variables by Different Approaches...................................................................................22 I.Introduction Monetary policy decisions in real time aretypicallybased on assessments of current and future economicconditions using incomplete data.Since most data, particularlyquarterlymacroeconomic data such asGrossDomestic Products (GDP), are released with a lag and are subsequently revised,assessingthe economicconditionsin the current periodbecomesa challengingtask for central banks.To address this issue,nowcasting techniques have been introduced