Skip to main content

Table 5 Post-lasso logit results for job tasks

From: FDI and onshore task composition: evidence from German firms with affiliates in the Czech Republic

Dep. variable: FDI in 2 years

Post-lasso logit

Manufacturing

Services

(1)

(2)

Job activities

 Researching, developing

–

12.12***

 Teaching, training

3.105

–

 Acquiring customers, PR

–

− 2.443

 Analyzing, investigating

3.136

10.97***

 Buying, procurement, selling

− 11.76***

− 3.082**

 Organizing/planning for others

0.353

− 0.327

 Consulting, informing

–

− 2.483**

 Measuring, checking

–

2.820*

 Negotiating

2.889

2.831*

 Serving, caring

− 2.377

–

 Surveilling, monitoring

− 2.095

− 4.322**

 Repairing

− 1.929

–

 Producing

–

5.604***

Associated knowledge

 Management

19.39***

9.080**

 Computer engineering

–

− 10.54**

 Giving presentations

–

1.148

 Foreign language

–

− 8.301

 Legal/law

135.1***

− 31.12***

 System analysis

1.013

–

 Maths

–

− 4.941*

 Other technical knowledge

–

2.431

 Labor legislation

88.54

42.12**

 Design

− 44.82***

–

 Use of software

− 0.312

0.339

 Marketing, sales

–

–

 Finance, tax

41.853

− 5.075

 Native language (German)

–

–

 Labor prot., environm. reg.

2.760

–

 Other specialized knowledge

− 5.669**

− 10.04*

 Medical knowledge

–

− 20.64***

Nonpenalized in lasso

 # of establishments

0.516***

0.299***

 Employment growth

0.0795**

0.166***

 Share of women

1.163***

0.196

 (log) wage bill

− 0.0070

0.585***

 Mean wage growth

0.0353

0.071*

 (log) employees

Yes

Yes

 Region FE

Yes

Yes

 Industry FE

Yes

Yes

 Year FE

Yes

Yes

 Observations

57452

51022

 log \(\lambda\) of min. MPSE

2.72

2.13

  1. This table reports the estimates of a post-lasso logit model. The set of included task intensities is selected by cross-validating the findings of the model with the lowest MSPE. The covariate employment size and the industry fixed effects control for the stratification of the sample of non-MNEs. Standard errors are clustered at the treatment level, i.e., the firm level, following Abadie et al. (2017)
  2. *\(p<0.1\), **\(p<0.05\), and ***\(p<0.01\)