Improving Business Cycle Forecasts’ Accuracy - What Can We Learn from Past Errors?



Improving Business Cycle Forecasts’ Accuracy

11


Table 2

Accuracy of RWI-4 and GD-4 Forecasts of the Annual Growth of GDP Components
1991-2004

RWI-4

GD-4

MAFE

MSFE

BIAS

MAFE

MSFE

BIAS

Private consumption

0,91

1,12

0,61

0,71

0,75

0,30

Government consumption

0,69

0,64

-0,19

0,69

0,63

-0,20

Investment in

equipment

4,61

31,27

3,61

3,71

21,14

2,46

structures

2,78

9,26

1,72

2,53

9,05

0,84

Export

3,33

17,04

0,76

2,90

11,11

0,03

Import

3,14

22,68

1,40

3,19

18,05

1,11

GDP

0,76

0,98

0,54

0,51

0,43

0,11

Author’s computations

As already noted, short term predictions are mostly made bottom up. Even if
in the very short run (1-2 quarters) the sectoral production forecasts also may
play an important role, in the end the forecasts are dominated by projections
of the demand side of GDP. Therefore, in addition to GDP also the efficiency
of forecasts of the demand categories will be analysed. Attention will be paid
to private (PC) and public consumption expenditure (GC), investment in
equipment (IEQ) and structures (IS), and, finally, exports (EX) and imports
(IM).

Table 2 presents some statistics of the accuracy of the demand side forecasts,
taking the RWI-4 and GD-4 forecasts as examples. It clearly can be seen that
the accuracy of the GDP forecast is owed to compensating errors: For almost
each component, government consumption being the only exception, the
forecast error is by far larger than that for GDP. Investment in equipment
shows the largest error, and the forecasts are markedly biased upward. Large
errors also occur in export and import forecasts. The somewhat better per-
formance of GD-4 compared to RWI-4 is partly owed to the fact that the first
is published at a minimum of 4 weeks later than the second which allows to
make use of additional information helping to improve accuracy.

3.2 Short Term Indicators Employed

There is no clear rule at hand, which short term indicators should be employed
in the tests to follow. As a rule, forecasters use many data for their work.
Therefore, numerous variables could be tested whether they were used effi-
ciently and a selection has to be made to restrict the scope of the further
analysis. Subsequently five types of variables will be considered (for a detailed
documentation of the indicators see annex.):

- Economic sentiment variables : Three indicators will be used: ifo business cli-
mate (IFOC), ifo business expectations (IFOE) and the consumer senti-



More intriguing information

1. The name is absent
2. Wounds and reinscriptions: schools, sexualities and performative subjects
3. The name is absent
4. Multi-Agent System Interaction in Integrated SCM
5. MULTIMODAL SEMIOTICS OF SPIRITUAL EXPERIENCES: REPRESENTING BELIEFS, METAPHORS, AND ACTIONS
6. Backpropagation Artificial Neural Network To Detect Hyperthermic Seizures In Rats
7. Spatial Aggregation and Weather Risk Management
8. The name is absent
9. BUSINESS SUCCESS: WHAT FACTORS REALLY MATTER?
10. Public infrastructure capital, scale economies and returns to variety