Offshoring data

The offshoring intensities for each industry and year are calculated using a method similar to Feenstra & Hanson (1999). In particular, the measures are constructed to represent the share of imported intermediate inputs in total industry output - a slight variation of the original measure following Geishecker (2006). The necessary data on input-use by both industry and year are taken from the import matrices that are part of the input-output tables provided by the Statistical Office of Germany. It is important to note that only inputs which originate in the same foreign industry j^∗ as the home industryj are included. Thereby it is ruled out that traditionally imported intermediates, that do not reflect an offshoring decision, are counted in. This makes the offshoring indices constructed here resembling the ”narrow” measure of Feenstra

& Hanson (1999)¹⁷:

OF F_jt = IM P_jt ×Ω_j^∗_jt

Y_jt . (4.12)

Ω_j^∗_jt represents the share of imports from a specific industry j^∗ used in the same industry j at home. IM P_j measures all imports of industry j.¹⁸ These imports can be split in order to differentiate among different offshoring destinations. To

16It may be problematic to assume fixed task compositions of occupations if some occupations change their task intensities and are relocated abroad not due to a general fall in offshoring costsβ, but due to themselves becoming idiosyncratically more tradable. Besides the argument that short-run compositional changes are inhibited by binding contracts, there is also little evidence of such change. From a similar task-survey for 2005/06, there is evidence to be found that is strengthening the validity of the assumption taken here. On average, there was hardly any movement in within-occupationalR-task intensity over the sample period considered and the average share of routine and non-interactive tasks within an occupation decreased by only -1.91%. The ranking of occupations in terms of theirR-task intensity also remained mostly the same. Estimation of a Spearman’s rank coefficient yields a value ofρ= 0.934 indicating a very high persistence of the occupational ranking over time.

17An alternative specification would be to use the share of intermediate inputs from all foreign industries. This variant is called ”wide” in Feenstra & Hanson (1999).

18These trade data are taken from the OECD STAN database. See the appendix for details.

provide for the closest fit with theory-based offshoring considerations, which are rooted in labor cost differentials across countries, a non-OECD country specific offshoring intensity is calculated to complement the worldwide measure.¹⁹ This region specific calculation entails the assumption that Ω_j^∗_jt does not differ across country groups, an assumption which is frequently taken in the empirical offshoring literature (see Geishecker 2006). Y_jt is the industry j’s output in year t as supplied by the OECD in the STAN data base. Output and import values are deflated using industry specific producer price indices, from the EU Klems data base, and an ag-gregate import price index supplied by the German Statistical Office, respectively.²⁰²¹

Table 4.1 gives an overview of the constructed variables. Offshoring intensities vary substantially between industries and over time, with most industries showing an increase over the sample period. Offshoring to non-OECD countries is much less important from a levels perspective, yet, it does show much stronger growth and has increased for all considered industries. It is of no surprise to observe relatively high offshoring intensities in the wearing and apparel and textile industries as well as in basic metals. This holds true for both worldwide and non-OECD offshoring. The highest growth is found in the radio, television and communication industry. At the same time, the cost share of routine and non-interactive tasks fell in most industries.

Employment shares also fell, but are excluded from this table due to expositional reasons. Employment shares of routine and non-interactive tasks are slightly higher.

This is intuitive since using employment shares is equivalent to assuming equal wages across occupations. With R-task intensive occupations being rewarded at a lower wage, the industry level wage cost share of these tasks is lower compared to their

19See the appendix for details regarding the data sources and the methods used.

20For additional details, see the appendix.

21Using output, which is equivalent to value added plus inputs, has the advantage of better accounting for domestic outsourcing. If firms outsource inputs domestically, the input measure will rise but this increase will be counterbalanced by a decrease in value added.

employment share. While falling relative labor demand in the face of increased off-shoring intensities point to a possible relationship between the variables, it will be left to the empirical analysis in the following sections to uncover the strength and significance of this link.

Table 4.1: Descriptives: offshoring and task intensity

industry name offshoring world offshoring non-OECD cost share ofR-tasks 1998 2007 change 1998 2007 change 1998 2007 change Food Products And Beverages 3.27 4.64 1.37 0.50 0.86 0.36 36.63 37.17 0.54

Textiles 11.19 9.26 -1.93 4.21 5.08 0.87 45.54 43.41 -2.13

Wearing Apparel; Dressing 17.99 13.74 -4.25 6.77 7.54 0.77 32.32 26.25 -6.06 Wood Products, Except Furniture 4.80 3.54 -1.26 0.98 1.20 0.22 45.56 45.39 -0.17 Pulp, Paper and Paper Products 9.19 14.07 4.88 0.45 1.26 0.82 44.57 45.75 1.19

Publishing, Printing 0.09 0.88 0.80 0.00 0.08 0.08 33.82 29.61 -4.21

Chemicals and Chemical Products 11.46 14.57 3.11 0.78 1.23 0.44 34.32 33.83 -0.48 Rubber and Plastic Products 0.97 1.61 0.63 0.09 0.23 0.14 47.77 47.20 -0.57 Other Non-metallic Mineral Products 2.26 2.09 -0.17 0.22 0.40 0.18 43.18 41.92 -1.26

Basic Metals 13.24 23.52 10.27 2.17 5.64 3.48 48.44 49.86 1.42

Fabricated Metal Prod., excl. Mach. 1.48 1.89 0.40 0.25 0.46 0.21 46.38 46.76 0.38 Machinery and Equipment NEC 6.10 8.33 2.24 1.05 2.34 1.29 40.37 38.73 -1.64 Office Machinery and Computers 10.75 10.86 0.11 2.65 5.14 2.49 25.26 22.46 -2.80

Electrical Machinery 5.78 6.16 0.38 0.99 1.52 0.53 36.94 35.77 -1.17

Radio, Television, Communication 5.16 16.90 11.75 1.20 6.55 5.35 32.80 28.80 -4.00 Medical, Precision and Optical 4.38 5.30 0.92 0.65 0.99 0.34 33.44 32.08 -1.36

Motor Vehicles, Trailers 6.90 9.71 2.80 0.20 0.62 0.42 44.24 41.56 -2.68

Other Transport Equipment 8.16 11.71 3.55 0.37 1.23 0.86 36.67 32.93 -3.74 Furniture; Manufacturing NEC 7.23 8.82 1.60 2.12 4.17 2.05 42.18 40.36 -1.82 Notes: The table shows the offshoring intensities for all industries included in the sample - both for worldwide offshoring and non-OECD country offshoring. Offshoring is the share of intermediate inputs (originating from the same industry abroad) in industry output. It also shows the cost shares which are calculated as total wage bill of routine and non-interactive tasks (occupation wage-sum multiplied by tasks intensity) over total wages. All values are expressed in %. Changes are simple differences. Some industry names are abbreviated.

4.4 Does offshoring affect relative labor demand