Investment dynamics in Japan

Abstract

This thesis contributes to the literature by providing evidence that the drivers and impacts of Japanese private fixed investment change over time. There is substantial research into aggregate investment, and it remains an active area of investigation. This is particularly the case in Japan, where policy makers and macro-economists alike are encountering novel challenges in the aftermath of the ’Lost Decade’. This is partly because theoretical and empirical aspects of investment are both individually complex and jointly difficult to reconcile. This thesis considers the suggestion such intractability may be due to different models for private fixed investment applying at different points in time. Chapter 2 evaluates this suggestion in Japan by comparing optimal forecasting models produced by Dynamic Model Averaging. From 1965 to 2014, the optimal forecasting model for investment exhibits substantial composition variation. This variation appears to be linked with known Japanese business cycle and financial crisis events. Chapter 3 explores the Chapter 2 finding that the forecasting utility of uncertainty displays little time variation. This appears true even across the ‘1980s bubble’ and its collapse as well as the Asian and Global Financial Crises. The chapter discusses the potential for ‘real options’ theory to explain the apparently linear relationship between uncertainty and investment. Using Time-Varying Parameter VAR modelling, this chapter compares realised investment responses to various uncertainty shocks with the canonical response shapes from real options theory. Results differ depending on measure, but domestic, forward-looking shocks appear consistent with the real options theory. Chapter 4 explores the Chapter 2 indication of substantial time-variation in the forecasting utility of investment factors, focusing on one particular component with substantial policy relevance, the yen. Time-Varying Parameter VAR estimates demonstrate a priori unanticipated responses of both the real exchange rate and output to investment shocks. To explain this variation, Chapter 4 derives anticipated sectoral sensitivities from an optimising framework and re-considers the decomposed shock components in a simple non-linear regression format. Results indicate a structural change in the industrial composition of output growth in Japan. These three empirical analyses are consistent in demonstrating the existence of time-variation in models of Japanese private fixed investment. To re-phrase, this thesis presents evidence that the underlying, data-generating processes behind Japanese private fixed investment changes over time.