1.  The study was undertaken in 11 plants, the subsidiaries of four multinationals. Four plants were based in the UK and each had a counterpart belonging to the same multinational and producing a similar product range, in a continental European country, and in the US. The companies selected for the study agreed to make available production worker personnel information, and all have retained long runs of personnel records, so that data on hirings was available for the period 1975 to 1994 in most cases. The plants and their products are given in Table 1, discussed below.

  2.  Our 11 plant, five country, two decade study involves data on about 6,000 individual recruits taken onto "permanent" contracts. (We define permanent to include workers hired initially on a temporary basis, so long as they become permanent after one year with the firm.) The observations for the hiring standards analysis are formed by averaging information from individuals recruited to work in each plant for every year. This plant reduces the data on individuals to 150 to 180 plant-time datapoints, depending upon missing values. The datapoints thus represent average behaviour for each plant in each year.

  3.  The use of matched plants allows us to standardise for the product and the technology used within each triplet. Since the product is held constant we can sensibly make comparisons across countries, for example, regarding labour costs or recruitment variables such as recruits' previous experience. Country groupings enable us to investigate whether plants in "unregulated" UK labour markets behave differently from those in "regulated" continental Europe. In practice, however, after allowing for plant fixed effects, we generally find that the same plant hiring standards equation can be applied to all plants irrespective of country or product.

  4.  To give a flavour of the analysis, Table 1 shows some data from the plants. The labour cost column shows wide differences between countries, given the product. The UK plants tend to have the lowest costs. Furthermore, contrary to what is usually thought, the US plants are high cost in two cases out of three, due to factors such as high company pensions and health insurance. High US labour costs demonstrate that the US is not as "unregulated" as it sometimes thought to be. Labour costs are important, because a high cost plant has to be a high productivity plant, which then cannot take a chance on unqualified workers, for example, inexperienced workers (including the unemployed). The second column gives the position regarding recruits' previous experience in each of the plants. The simple correlation between the two columns is -0.54 (p <0.1), which illustrates our point that the high cost plant will recruit fewer inexperienced workers.

  5.  The two end columns of Table 1 show recruits' education and its changes over time. Most plants have experienced a rise in the education of recruits, as noted above. Apart from this there might seem to be little pattern. In fact, however, the simple correlation between education changes and previous experience (or starting age) changes is quite high, -0.53 (p <0.1), illustrating the substitutability of education and experience to which we have already drawn attention. This substitutability emerges even more strongly when we hold other variables constant in a multiple regression.

  6.  As a final exhibit of the data available from the research, Figure 1 shows trends in the proportion of recruits hired on a temporary basis. As can be seen, the trend is upwards in most plants. As noted in the text, the proportion of short-service workers in a plant's workforce (measuring insider power) is negatively associated with movements in the temporary proportion.