3.3 Correlation between Intensive and Extensive Product Margins
There is a large literature on whether the size differences across firms are mostly due to extensive (number of products) or intensive margins (output per product).
In order to address this, log number of products is regressed on the log of total sales. The upper panel of Table 6 shows two sets of results. The first two columns show results from a cross-section regression in one year, 2005, and include industry fixed effects, whereas the last two columns are pooled results and include industry and time fixed effects. All the coefficients are positive and significant. The results show that about 16 to 28% of variation in output across firms is due to variations in the extensive margin.
| (1) | (2) | (3) | (4) | |
|---|---|---|---|---|
| Log Number of Products | ||||
| 2005 | MP firms, 2005 | all years | MP firms, all years | |
| Log Sales | 0.215 | 0.165 | 0.280 | 0.231 |
| (16.17)** | (12.02)** | (75.16)** | (61.67)** | |
| Fixed Effects | Industry | Industry | Industry, Year | Industry, Year |
| Observations | 2545 | 1715 | 24075 | 17410 |
| R-squared | 0.02 | 0.02 | 0.05 | 0.04 |
| (5) | (6) | (7) | (8) | |
|---|---|---|---|---|
| Log Average Sales | ||||
| 2005 | MP firms, 2005 | all years | MP firms, all years | |
| Log Number of Products | 0.275 | 0.163 | 0.259 | 0.201 |
| (6.25)** | (2.46)* | (12.91)** | (6.73)** | |
| Fixed Effects | Industry | Industry | Industry, Year | Industry, Year |
| Observations | 2545 | 1715 | 24075 | 17410 |
| R-squared | 0.10 | 0.08 | 0.24 | 0.23 |
Note: Table summarizes OLS regressions of log number of products on log sales in the upper panel and log sales on log number of products in the lower panel. First two columns are run on 2005 data only, while columns 3 and 4 pool across all years. Columns 2 and 4 include only multi-product exporters.
According to the theoretical model in Bernard et al. (2006b), there is a positive correlation between extensive and intensive margins. The ratio of output attributed to a firm's extensive vs. intensive margin is of interest to firms and policymakers since it is important to determine whether a firm is large due to manufacturing of a large number of products or production of more output per product.
To test this theory, the log average sales per product are regressed on the log number of products. The first two columns in the lower panel of Table 6 present cross section results for all firms and multi-product firms, and the last two columns are pooled results for all and multi-product firms. A positive correlation is observed between multi-product firms' extensive and intensive margins in 2005. The pooled results are also positive.
The results show that bigger New Zealand exporters produce on average more products and firms that produce more products on average have larger sales per product. These patterns are similar to Indian and U.S. firms and provide further support to the theoretical prediction of a positive correlation between extensive and intensive margins.
