Dynamic Models of Portfolio Behavior: More on Pitfalls in Financial Model Building
指出Brainard-Tobin的财富约束模型在存在调整成本时，将储蓄与投资组合决策分离的做法不成立，应整合两者以确定资产和消费的最优时间路径。
In an important article in this Review, William Brainard and James Tobin have emphasized the role played by the wealth constraint in systems of asset demand equations. The wealth constraint gives rise to consistency conditions which must be satisfied by the demand functions when such a system is specified and estimated. As Brainard and Tobin caution, care must be taken to ensure that unrealistic coefficients are not inadvertently imposed on omitted equations by failure to recognize the consistency conditions.' Noting that the wealth constraint applies out of, as well as in, portfolio equilibrium, Brainard and Tobin focus attention on systems in which actual and desired stocks of assets differ. They specify a multivariate stock adjustment model wherein the desired change in holdings of any asset depends in general upon all asset stock disequilibria; the existence of such stock disequilibria can be implicitly rationalized on the basis of costs of adjustment which impinge on the rate of change of at least some assets. In this framework they show that the stock adjustment coefficients must also satisfy certain consistency conditions to ensure that the wealth constraint is satisfied. An important feature of their analysis is that the total change in wealth (savings plus capital gains) is treated as exogenous to the financial sector, and the asset flow demands described above are conditional upon the exogenously given change in wealth. This strategy of separating the portfolio balance decision from the consumption-saving decision is one that Tobin has explicitly used and justified in his 1969 article (especially pp. 15-16), and is one that has been widely and effectively used in modern macroeconometric models. The central argument of the present paper is that this of flow-allocation and stock-allocation decisions is not legitimate in the presence of adjustment costs attached to changing the level of individual asset holdings. The existence of adjustment costs means that there is no portfolio balance problem per se (in the sense of allocation of a given level of wealth), but rather a (longer run) problem of determining an optimal time path for each asset and for the level of consumption. Thus a natural extension of the Brainard-Tobin model is to treat saving and portfolio decisions in an integrated fashion.2 Note that the Brainard-Tobin model is perfectly consistent with any model of savings behavior and hence no logical con*Queen's University, and Cowles Foundation for Research in Economics, Yale University. I am grateful to Adrian Pagan, Gordon Sparks, and James Tobin for helpful discussions, and especially to Gary Smith who, as well as patiently discussing many of the issues, provided detailed comments on earlier drafts of this paper. This research was partially supported by a National Science Foundation grant to the Cowles Foundation and by a Canada Council grant to the author. Remaining mistakes and opinions are my own. I This also has implications for the common practice in macro-economic models of leaving the bond market as implicit. Care must be taken to ensure that silly behavior is not inadvertently attributed to bondholders. William Silber, Tobin, and Alan Blinder and Robert Solow have initiated research which reintroduces the bond market into macroeconomic models. 21t appears to be a fairly general result that the existence of adjustment costs leads to integrated behavior. M. Ishaq Nadiri and Sherwin Rosen have established a similar result for the theory of the firm, and Robin Mukherjee and Edward Zabel have recently shown that the separation theorem prominent in the finance literature on the mean-variance approach to optimal consumption-portfolio behavior fails to hold when transactions costs are introduced. In my 1975 paper (Appendix), I have argued that the integration of saving and portfolio balance decisions also applies in continuous-time models, even though such models are characterized by separate stock and flow budget constraints.