This paper argues that interest rates affect output through their impact on the cost of variable inputs such as materials and labor. The cost of variable inputs increases in the interest rate because of time lags in the production process. When a firm sets out to produce a good, first it has to acquire raw materials and put workers to work, and only after a period of time it is able to sell the finished good. In the interim, the firm foregoes interest income or incurs interest costs. If the interest rate goes up, the production process becomes more costly so that the firm optimally chooses to reduce its sales.
I exploit the following cross-sectional implication of the theory: industries that usually hold more inventories relative to their variable costs should react more strongly to changes in the cost of capital. The reason is that inventories account for the value of all variable inputs accumulated between the time of their acquisition and the sale of the corresponding output. As a consequence, the ratio between total inventories and variable costs captures the importance of the foregone interest rate income associated with the purchase of variable inputs.
The empirical analysis focuses on episodes in which the link between capital flows and output was most likely to be visible. These are the sudden stop episodes identified by Calvo et al. (2006). The discrete jump in current account surplus observed in these episodes suggest that they were largely unforeseen shocks. The large and persistent drop in investment/output ratio that follows suggests a large and persistent increase in the cost of capital. Together these suggest that these episodes provide interesting natural experiments on the effects of shocks to the cost of capital on output.
I find that in the course of these episodes there is a large and robust reallocation towards sectors with low inventory-to-cost ratios. Figure 1 shows the reallocation pattern in the data. Each point refers to a manufacturing industry. The vertical axis depicts the difference in the log of value added in the year when GDP was at its lowest and the value three years before, averaged over 21 sudden stop episodes. The horizontal axis represents a proxy for the inventory-to-cost ratio calculated for each industry. As I discuss in section 3, the addition of controls reduces the slope of the regression coefficient, but the correlation is remarkably robust.
The reduced form estimates establish that the reallocation goes in the direction predicted by the theory. To assess whether time lags in production can plausibly account for the magnitude of the estimated effects, I set up a multi-sector general equilibrium model with delays between the use of inputs and the completion of the output. I calibrate the production lags so that they are compatible with the inventory-to-cost ratios in the data. I then simulate a crisis in the model as a combination of shocks to foreign interest rate and total factor productivity, chosen to match the magnitude and persistence of changes in GDP and fixed investment. The model is able to generate a sectoral reallocation in response to a crisis-like event in line with the estimates from the data.
Lastly, I can also use the model to assess the importance of the production time channel for the impact of interest rate shocks on aggregate output. The model implies that the shock to the foreign interest rate accounts for about one quarter of the initial drop in output. In particular, because of the interest rate shock alone, GDP drops about 4% over the first two quarters after the shocks.
Download
PDF Ebook Time to Produce and Emerging Market Crises
