Fazzari et al.(1988) put forward a creative way to measure the financial constraints by the sensitivity of investment to cash flow. So far, this approach has a profound impact on the latter stydies, but also is deeply questioned for the lack of solid theory foundations. Against the background, this paper begins with the sensitivity to cash flow as the logic starting point, from six dimensions (investment homogeneity and heterogeneity, the constraints of cost and credit rationing, static model and dynamic model) to analyze systematically the effect of financial constraints on investment and how the relationship between investment and financial variables changing with financial constraints, then carries out a further discussion of the major controversies about the sensitivity of investment to cash flow.The primary conclusion of this paper is that the sensitivity to internal funds, rather than the sensitivity to cash flow, could be an appropriate indicator to measure the strength of financial constraints. On this basis, the paper also derives the following specific conclusions:Firstly, in a one period model, it is indeed that no determined monotonic relationship exists between financial constraints and the sensitivity of investment to internal funds, but this is because the sensitivity to internal funds is the combined effect of finance curves of supply and demand (i.e. investment opportunities), which reflect not only the effect of financial constraints but also the impact of investment opportunities. When Fazzari et al.(1988) initiated to employ a q model that includes cash flow (or internal funds) as the investment equation in empirical research, Their purpose was just to make use of q to remove the effect of investment opportunities, so that the role of financial constraints on investment could be completely captured by the coefficient of cash flow (or internal funds). Therefore, It is completely invalid for the critique of Kaplan and Zingales (1997) which is based on a one-period model to deny the validity of investment-cash flow (or internal funds) sensitivity as a measure of financial constraints. Meanwhile, It is not appropriate for Hubbard (1998) and Hall (2002) and others to apply one-period graphics to illustrate qualitatively the idea of investment-cash flow sensitivity, as the graphics could not provide any theoretical supports for this approach.Secondly, by establishing a dynamic model, solving the equilibrium investment paths, and making simple econometric transformations, this paper confirms that q and internal funds could be the independent invariables to explain adequately the investment provided that the investment is homogeneous, regardless of the firm subject only to the constraints of financing cost, or also simultaneously constrained by credit rationing. In the q equation, the coefficient of internal funds (or its nonlinear function) should be ranged from zero to one, and has a monotonically positive correlation with the strength of financial constraints. Whereas the coefficient of q should have a monotonically negative correlation with the financial constraints. When a firm is subject to credit rationing, the sensitivity of investment to internal funds will be greater than the sensitivity when it is subject only to the constraints of financing cost. If the firm is not constrained by credit rationing, and has a relatively single structure of the external financing, it would be a much more comprehensive and easily operated approach in measuring the financing cost by adding a quadratic function of external financing into the q equation and estimating theirs coefficients than the sensitivity to internal funds. If investment opportunities could be controlled completely, the coefficients of external financing or (and) its squared termâ€™s should be negative, and the higher the firm pay for a premium of external financing, the greater the absolute value of coefficients will be.Third, when a firm that is subject to financial constraints engages simultaneously in R&D and physical capital investment, which are different significantly from each other in earnings, risk, information asymmetry and adjustment costs et al., the two types of investment would compete for the funds. Thus the effect of crowding should be controlled in the q equation of each investment. The tighter the firm being financial constraints, the weaker capacity of each investment to utilize its marginal income, and the stronger the effect of crowding by another investment. After controlling the influence of the marginal income of all investments, the coefficient of the internal funds in R&D and physical investment equation would both monotonically increase with the financial constraints (including financing cost and credit rationing) increasing. And when the firm is subject to credit rationing, the sensitivity of R&D and physical investment to internal funds will both be greater than the sensitivity in case of financing cost constraints. Existing theoretical and empirical studies show that R&D activities rely mainly on firmâ€™s internal funds, and are nearly impossible to get support from debt financing. Thus it is probably the constraints suffered by physical or total investments, rather than R&D investment, that is captured by the coefficient of internal funds in the R&D investment equation, and the discrepancies of sensitivity to internal funds of R&D and physical investment result completely from the differences of the adjustment costs. The greater adjustment cost of R&D investment determines its smaller sensitivity to internal funds. When the firm only subjects to financing cost constraints, for. any type of the investment, the shadow value of its capital stocks and quadratic functions on external financing could be used as independent variables of investment equation, without considering the relationship of competing for funds among different investments. The coefficients of external financing or (and) its squared termâ€™s should be negative, and the absolute value of coefficients would be positively (negatively) correlated with financing cost (marginal adjustment cost). The coefficient of marginal q is positive, and negatively determined only by parameters of marginal adjustment cost. The higher adjustment cost determines that the coefficients of q and external financing (absolute value) in R&D investment equation should be less than those of physical investment.Finally, the reason why Gomes (2001) and Alti (2003) and others found firms not subjecting to financial constraints also display a significant investment-cash flow sensitivity, is that the cash flow in their models contained the information of income shocks. If they use the sensitivity to internal funds as a measure, the problems of the significance may be alleviated to a large extent. More importantly, even if internal funds involve some income shocks, the differences of sensitivity to internal funds among groups of prior different constraints will be underestimated rather than overestimated. The classification criteria used by Fazzari et al.(1988) and Kaplan and Zingales (1997) to identify financial constraints are of internal consistency, and not the primary explanation for the conflicting empirical conclusions. When there are bias errors with the classification criteria, or the Econometric models are specified with a cross-term of cash flow and dummy variables that characterize the extent of financial constraints, the difference of sensitivities to internal funds between the constrained and unconstrained groups will be weakened. It is very serious mistakes for some studies to put the variables of cash flow together with debt financing as well as equity financing into the q equation or other investment models, which not only canâ€™t mitigate the estimation bias, but also make the coefficients of independent variables no longer having clear interpretations. When the Accelerator Model and (or) Error Correction Model that derived under perfect market are adopted, investment opportunities will be underestimated, and the more financing premiums the firm have to pay in future, the greater the underestimation bias will be. If the sensitivity to internal funds (but not cash flow) is used to measure the financial constraints, the sample firms that fall into financial distress (whose internal funds is very low but cash flow is positive) or newly startup (whose internal funds is relatively high but cash flow is negative) probably need not to be censored. |