What determines the Japanese firm investments: real or financial?

Abstract

In this article, I test both the real and the financial frictions in the Japanese firm investments using structural approach. The real represents the nonconvex capital adjustment costs, and the financial means the financing constraints. These two factors have been studied for a long time, but rarely analysed simultaneously. Through this analysis, I find the following results. First, both these two factors affect Japanese firm investments. Second, convex adjustment costs parameter is estimated at a very small level. Third, this small convex adjustment costs lead to the model introduced in Wang and Wen (2012), which insists that the firm investments are determined by the upper bound of the borrowing constraints.

Keywords:

• firm investments
• nonconvex adjustment costs
• financing constraints
• structural estimation

JEL Classification:

• D21
• G31

Acknowledgements

The author is very grateful to Hiroshi Teruyama and an anonymous referee for their helpful comments and suggestions.

Notes

¹ Empirical research using the reduced form should be interpreted carefully because the statistical significance of cash flow in a standard Q investment function does not necessarily reflect the importance of financing constraints. Erickson and Whited (2000) point out that cash flow may appear to have a significant effect due to error in measuring average Q. Cooper and Ejarque (2003) insist that this result is due to monopolistic power in the goods market.

² This type of nonconvex adjustment cost is assumed in Caballero and Engle (1999) and Cooper and Haltiwanger (2006) as well. Cooper and Haltiwanger (2006) also consider fixed costs proportional to the plant-specific capital stock () and conclude that either of these nonconvex costs is necessary to explain firm level investment behaviour.

³ To derive this form of profit function, two assumptions are made. One is Cobb-Douglas production function and the other is isoelastic demand curve.

⁴ Bayraktar et al. (2005) formulate this financing constraint through a premium on the cost of new debt, which depends on the firm’s leverage ratio. However, in Japan, financing constraints seem to operate following Equation 6, so I consider this type of financing constraint here.

⁵ It is also possible to estimate these two functions simultaneously. I avoid this approach for two reasons. First, the error terms of these two functions ( and ) are orthogonal, which means that estimating them separately does not reduce efficiency. Second, if either of these two functions is misspecified, then bias due to this misspecification contaminates the rest of the system. (see Hayashi (2000)).

⁶ It is common to formulate the variable as a function containing a constant term. However, in my data set, the objective function is not concave if I use a constant term, so I specify the variable without a constant term here.

⁷ They term the spell complete if a firm adjusts its capital stock in some period and adjusts it again in period , and they term the spell incomplete if a firm invests in period and does not adjust again. Then, they estimate the investment Euler equation using samples with a complete spell but also the samples with an incomplete spell. They show that one cannot estimate the structural parameters consistently without the incomplete spell samples.

⁸ I use this database from the online service ‘Needs-financial QUEST.’

⁹ Some values are missing. The number of observations is as follows. In the profit function estimate, machinery has 1682 observations, electrical machinery has 1698, motor and transport equipment has 866, chemicals has 1900 and steel has 500. In the investment Euler equation estimate, machinery has 1628 observations, electrical machinery has 1708, motor and transport equipment has 671, chemicals has 1629 and steel has 560.

¹⁰ I define the inaction rate as the proportion of investment that is less than 1% in absolute value. Further, I define the spike rate as the proportion of investment that is more than 20% in absolute value. These criteria follow Cooper and Haltiwanger (2006).

¹¹ The data also show the irreversibility of investment, because the rate of negative investment is much lower than that of positive investment. I should therefore consider the asymmetry of adjustment costs in setting the models. However, there are too few observations with negative investments to estimate the structural parameters stably, and furthermore, it is difficult to identify capital goods purchases from their disposal. Instead, I focus on investigating nonconvex adjustment costs and credit market imperfections.

¹² Let be the demand curve, where S is a demand shock and the inverse of the demand elasticity, and let be the production function. Then, maximizing profit over the flexible factor, l, leads to a reduced form profit function . The exponent on capital is (see Cooper and Ejarque (2003)). In this model, the mark-up is represented as .