5  Why do Firms Alter Their Product Mix?

The previous section has shown that changes in product mix are frequent, span a variety of industries and make up a significant portion of a firm's exports. Given these results, we consider different explanations for this behaviour that is observed among New Zealand firms.

First, product-specific factors that are common across firms are considered. For example, demand shocks can lead to firms adding ‘hot' products for which demand is rising and drop ‘cold' ones for which it is falling. Likewise, supply shocks due to technological progress or international trade might lead to dropping of uncompetitive products in line with the theory of comparative advantage. If this indeed is the main cause of the product switching behaviour, we should observe a negative correlation between add and drop rates.

Figure 7 shows the annual mean rate at which product are added and dropped by New Zealand exporters. The add rate in year t is the number of firms that added the product between years t-5 and t divided by the average number of firms producing the product in both years. Drop rates are calculated similarly. Our results indicate that there is a slight positive correlation between adding and dropping rates. This suggests that the patterns of product mix change are not solely a result of reallocation of resources from one group of products to another due to say a demand shock, which would have implied a negative relationship between add and drop rates.

Figure 7 – Average Product Add and Drop Rates

Note: Figure shows average rates at which products are added and dropped by year.

Second potential explanation focuses on factors that specific to firms but common to all products. For example, a decrease in trade costs might increase the profitability of all the products and lead the firm to add a marginal product that it did not previously export. However, the evidence presented in Section 4 that shows that firms simultaneously add and drop products does not support this explanation. This suggests that firm specific shocks may affect products differently.

Third potential explanation considered is firm-product factors that focus on product sunk costs of entry and producer heterogeneity within and across product markets. This can be thought of an extension of the existing models of industry dynamics (Jovanovic 1982, Hopenhayn 1992, Ericson and Pakes 1995 and Melitz 2003) that are supported by empirical studies such as Baily et al (1992) and Foster et al (2001, 2006) from firm entry and exit to extensive margin of firms, i.e., product entry and exit. Continuing firms face ex ante uncertainty about their productivity in new product markets. For each new product, the firm incurs a product-specific sunk cost. Theoretically, in equilibrium, the flow of firms that add a product each period must equal the flow of firms that drop the product such that equilibrium add and drop rates are positively correlated. Add and drop rates depend on the costs of entry of products: products with low sunk costs of entry exhibit high entry and exit rates and vice versa.

A plausible interpretation then is that firm-product characteristics are important in explaining product switching behaviour of firms. To analyze this, we go back to theories of sunk costs of entry models that suggest that exiting firms should have a relatively low output and should have produced for a short period of time compared to other firms. Extending this to product entry and exit analysis implies that firms which are dropping a product should have relatively low market share of that product and should have exported the product for a short time compared with other firms which continue to export it.

Table 12 reports OLS regressions of a dummy variable indicating that a firm drops a product between 1999Q3 and 2004Q4 on firms' 1999Q3 product size and age (relative to all the other firms exporting the same product) as well as firm size and age (relative to the average size and age of other firms) and the number of products the firm produces,

where j and i index firms and products, respectively.^[14]

Several specifications are considered. The first column reports industry fixed effects, second column has firm fixed effects (these are firm characteristics that are common across products, such as total exports of the firm, age of the firm, whether it is a multiple product exporter) and the third column reports product fixed effects (these are product characteristics that are common to all firms, such as aggregate demand supply for that product). These fixed effects control for unobserved firm and product specific effects that influence the probability of a product being dropped. Note that firm specific variables are dropped when firm fixed effects are included. The results are robust to inclusion of different fixed effects.

Table 12 shows that firm-product relative size and age are negatively correlated with product dropping. The negative coefficients on relative firm-product size and age are in line with the predictions of sunk cost models and suggest that firm-based explanations by themselves are incomplete. Furthermore, these variables are constructed relative to the average of all firms in that product market, making product-based explanations incomplete as well. The significance of product variables, after controlling for the characteristics of the firm, provides support to the third potential explanation discussed above.

Note: Table reports OLS regression results of a dummy variable indicating product dropping on firm-product and firm attributes. Attributes are relative to the whole sample of firms. Standard errors are clustered at the product level. Industry, firm and product fixed effects are used in different specifications. Number of observations rounded to nearest five for confidentiality reasons.

We have shown in the previous section that firms' product-mix alterations are an important part of aggregate exports. Here, it is shown that firm-product drops exhibit lower relative product size and tenure. If we assume that relative product size and age are positively correlated with firm-product productivity, we can potentially say that systemic reallocation of resources towards high productivity sectors occurs across products within firms as well as across firms.

Likewise, firm relative size and age are negatively correlated with product dropping, suggesting that large and old firms are less likely to alter their product mix, so it is the new and small firms who engage in such behaviour. This is consistent with the literature on firm entry and exit that finds these variables to be negatively correlated with firm exit. This is in line with our theory stated at the beginning of the paper that changes to the product mix might be used as a hedging mechanism for firms who are more prone to fluctuations in export markets.