4.7  One-year firms

Another issue which also relates to firm entry and exit is the treatment of “one-year” firms. In the OECD firm-level project, firm entry and exit were defined in terms of three (rather than two) time periods. Firm entries were those observed as (out, in, in) the database at time (t-1, t, t+1), while exits were observed as (in, in, out) at time (t-1, t, t+1) (Bartelsman et al, 2003). Firms that appeared in the database in a given year but not in the two adjacent years (ie, they were observed as (out, in, out) at time (t-1, t, t+1)) were termed “one-year” firms. These firms were excluded from all analyses, including analyses of variables other than entry and exit rates (eg, firm size).

A similar issue to one-year firms is that of non-contiguous firms, which we also term “rebirths”. These are firms that disappear from the database for one or more years and then reappear (ie, they are observed as (in, out, in) the database at time (t-1, t, t+1)). In a sense rebirth firms are the opposite of one-year firms. While the treatment of rebirth firms in the OECD project is not made explicit in Bartelsman’s (2003) paper, we have assumed that they were also excluded from the analyses. This is because rebirths (like one-year firms) do not conform to the definitions of firm entry and exit used in the OECD project. In the remainder of the paper, references to one-year firms should be taken to also include rebirths.[3]

To our knowledge, previous studies of firm dynamics in New Zealand using the BDS data have not tried to exclude one-year firms. We are uncertain of the exact status of one-year firms in the Eurostat, but have assumed that they are included, given that the Eurostat has the most liberal criteria overall for inclusions of firms in the database. However, the Eurostat approach of cleaning the data to remove false entry and exit might have the effect of eliminating some one-year firms. This is because the one-year firm category is assumed to include a high number of measurement errors and false entries and exits (Bartelsman et al, 2003).

4.8  Addressing measurement differences

To address measurement differences in comparisons of firm dynamics across countries, there are effectively two approaches that can be taken. One would be to obtain indicators of firm dynamics from other countries, but using similar measurement criteria to those applied in the New Zealand BDS data. This is often difficult or impossible, given limited access to other countries’ statistical databases.

The other approach – and the one taken here – is to obtain new statistics on New Zealand’s firm dynamics from the BDS source data, by adjusting the measurement criteria to mirror the criteria used in overseas databases (to the extent that this is possible given data limitations).

In practice, this means the following. First, for all of the analyses reported on in this paper, data were obtained from the New Zealand BDS at an enterprise level of analysis, using an employee only and headcount measure of employment. This is because both the Eurostat data and the OECD firm-level project generally use enterprise level data, and an employee and headcount measure of employment.

Secondly, depending on the comparison being made, further adjustments to the New Zealand data were made to reflect differences in the criteria for inclusion of firms in the database, in the treatment of one-year firms, and in sector coverage. When making comparisons with data from the OECD firm-level project, zero-employee firms and one-year firms were generally excluded from the New Zealand data. When making comparisons with the Eurostat data, zero-employee firms and one-year firms were generally included. In addition, for some of the analyses of firm size – in particular, comparisons with publicly available data from the USA, UK, and Australia – the public sector was also excluded from the New Zealand data.

Finally, where possible, the statistics for New Zealand were calculated over the same or similar time periods as those used in the comparator data. However this was not possible in every case. And as discussed above, even where the same time periods are available, there may still be business cycle effects that impact on the comparability of firm dynamics statistics.