Skip to main content

Table 4 Out-of-sample predictive accuracy relative to benchmark

From: Capturing Swiss economic confidence

 

Full sample

Without recessions

 

Nowcasts

Forecast

Nowcasts

Forecast

Horizon

1

2

3

4

5

6

1

2

3

4

5

6

Root mean squared errors

AR(1)-model

15.30

15.3

15.30

13.00

12.90

12.90

3.57

3.57

3.57

6.10

6.08

6.08

Relative performance of SEC index

SEC Bridge

0.644*

0.644*

0.620*

0.767*

0.855*

0.898

0.909

0.907

0.835*

0.514*

0.519

0.527

SEC AR-Bridge

0.820*

0.820*

0.772*

0.791**

0.882*

0.917

1.013

1.011

0.988

0.641*

0.686

0.650

Relative performance of alternative monthly indicators

KOF Barometer

0.476

0.476

0.470

0.666

0.860

1.101

1.221

1.220

1.256

0.778

0.689

0.674

PMI Manufacturing CH

0.686

0.685

0.688

0.818*

0.867

0.889

1.157

1.149

1.081

0.571

0.556

0.536

PMI Manufacturing Foreign

0.631

0.631

0.620

0.711*

0.795*

0.884

1.118

1.116

1.059

0.579

0.551

0.542

SNB-BCI

0.410*

0.410*

0.369

0.386*

0.563

0.962

0.944

0.944

0.854

0.502*

0.672

0.834

  1. Modified Diebold–Mariano test: the alternative hypothesis states that the monthly indicator is more accurate than the benchmark. Significance levels: p value: ***< 0.01, ** < 0.05, *< 0.1 of the modified Diebold–Mariano test (Harvey et al., 1997). Horizon refers to months until the GDP release of the respective quarter. For the target variable—GDP—the real-time vintages are used. Forecast errors are with respect to the first release. The estimation sample spans from 2001:M3-2023:M6 for the full sample. For the subsample without recessions, we exclude the quarters 2002:Q3–2003:Q2, 2008:Q3–2009:Q1 and 2020:Q1–2020:Q3