European Commission Directorate General for Economic and Financial Affairs The harmonised EU investment survey: What can it tell us about investment growth in the euro area? Roberta Friz and Christian Gayer Joint EC/OECD Workshop on Business and Consumer Surveys 12-13 November 2007, Brussels European Commission 2007 1 Motivation • Business investment is a key GDP component • GFCF accounts for approx. 20% of euro-area GDP • High volatility significantly lager share of GDP growth • Since 1966: EC investment survey • Representative samples useful information on overall investment in industry sector • Released before statistical data should also be useful for forecasting • But: few studies on the usefulness of the data European Commission 2007 2 Outline 1. 2. 3. 4. Overview of the EC Investment Survey Descriptive analysis Usefulness in forecasting Conclusions European Commission 2007 3 1. Overview of the EC Investment Survey The investment survey is carried out twice a year : • • in March/April (“spring”): percentage change in investment of the company from year t-2 to t-1 and from year t-1 to t in October/November (“autumn”): percentage change in investment of the company from year t-1 to t and from year t to t+1 four consecutive estimates of investment growth are available for each year. Example for year 2006: First estimate Second estimate Third estimate 'definitive' estimate Date of survey Oct/Nov 2005 Mar/Apr 2006 Oct/Nov 2006 Mar/Apr 2007 Date of publication of survey results: end-Jan 2006 end-Jun 2006 end-Jan 2007 end-Jun 2007 Eurostat data available at the time of survey results publication First estimate 2006Q1 2nd estimate 2006Q3* 2nd estimate 2006Q4** * 1st estimate available around end-Nov 2006 ** 1st estimate available beginning March 2007 European Commission 2007 4 2. Descriptive analysis The data: • Investment survey covers only the manufacturing sector therefore we choose equipment investment (metal products, machinery and transport) as reference series to approximate investment activity in the manufacturing sector. • Equipment accounts for approximately 40% of total investments and the two series are highly correlated • Focus is on nominal investment growth European Commission 2007 5 Equipment investment growth and surveyed change of investment in the euro area – nominal value Equipment investments Survey data (investment 'definitive' value) 19 86 19 87 19 88 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 19 85 20 15 10 5 0 -5 -10 -15 -20 6 15 5 0 -2 0 O ct-8 4 A p r-8 5 O ct-8 5 A pr -8 6 O c t- 86 A p r- 87 O ct-8 7 A p r-8 8 O ct-8 8 A p r- 8 9 O ct-8 9 A p r-9 0 O ct-9 0 A p r-9 1 O c t- 91 A p r- 92 O ct-9 2 A p r-9 3 O ct-9 3 A p r -9 4 O c t-9 4 A p r-9 5 O ct-9 5 A p r-9 6 O ct-9 6 A p r- 9 7 O ct-9 7 A p r-9 8 O ct-9 8 A p r -9 9 O c t- 99 A p r- 00 O ct-0 0 A p r-0 1 O ct-0 1 A p r- 0 2 O ct-0 2 A p r-0 3 O ct-0 3 A pr -0 4 O c t- 0 4 A p r- 05 O ct-0 5 A p r-0 6 O ct-0 6 Investment reported in each survey for the euro area – nominal value 20 198 5 1 989 198 6 10 1 995 19 90 19 88 1996 199 4 -5 19 98 19 97 20 00 2 007 198 7 2006 1991 200 1 2 004 1 999 199 2 200 2 20 03 2 005 -1 0 1993 -1 5 7 Average revisions vis-à-vis the definitive survey estimate 1992 to 2006 Difference vis-à-vis definitive value definitive value 3rd estimation 2nd estimation 1st estimation Mar/Apr t+1 Oct/Nov t Mar/Apr t Oct/Nov t-1 Euro area 1.1 -0.5 2.7 2.1 Germany -0.7 2.3 4.3 3.3 Spain 2.5 -2.2 3.2 6.5 France -1.4 2.0 5.2 2.3 Italy 6.7 -8.1 -7.8 -5.3 Minus previous estimate was lower by … pp than the definitive estimate Plus previous estimate was higher by … pp than the definitive estimate 8 Actual versus estimated Euro area 20 15 10 5 0 Ameco equipment inv Nominal survey - definitive value survey - 3rd estimation survey - 2nd estimation survey - 1st estimation -5 -10 -15 -20 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Euro area Germany Spain France Italy Difference vis-à-vis actual outcomes Nat. Account definitive estim 3rd estimation 2nd estimation 1st estimation Outocmes Mar/Apr t+1 Oct/Nov t Mar/Apr t Oct/Nov t-1 2.7 -1.6 -2.1 1.2 0.6 0.9 -1.6 0.7 2.7 1.7 7.6 -5.1 -7.2 -1.9 1.4 1.7 -3.1 -1.1 2.2 -0.8 4.5 2.3 -5.8 -5.6 -3.1 Minus underestimation Plus overestimation 9 • The mismatch can partly be due to: – different coverage of the survey (manufacturing industry) and reference series (equipment) – Revisions in the National Account data: • Considering the years from 2000 to 2006, the mean absolute error has been 0.68 pp and the RMSE 0.91 pp. • In the period 2000 – 2006 the first releases have always been revised upwards. The negative bias amounts to -0.68 (mean error). 10 Direction of the changes and correlation coefficients Percentage of correct directional changes indicated in the survey Survey data definitive value 3rd estimation 2nd estimation 1st estimation Mar/Apr t+1 Oct/Nov t Mar/Apr t Oct/Nov t-1 Euro area 0.80 0.73 0.87 0.73 Germany 0.80 0.87 0.80 0.80 Spain 0.80 0.74 0.79 0.89 France 0.68 0.77 0.91 0.73 Italy 0.95 0.64 0.68 0.76 Correlation between investment survey data and actual investment in equipment Survey data definitive value 3rd estimation 2nd estimation 1st estimation Mar/Apr t+1 Oct/Nov t Mar/Apr t Oct/Nov t-1 Euro area 0.92 0.85 0.86 0.72 Germany 0.78 0.71 0.73 0.59 Spain 0.68 0.73 0.69 0.49 France 0.82 0.82 0.78 0.54 Italy 0.71 0.81 0.63 0.61 11 • Benchmark with other questions in the BCS From Industry survey asked in January, April, July and October – factors limiting production • “none”, “insufficient demand”, “shortage of labour force”, “shortage of material and/or equipment”, “financial constraints” and “others” – Assessment of current production capacity • “more than sufficient”, “sufficient” or “not sufficient” – Capacity utilisation • quantitative figure reporting the percentage of full capacity they operate at 12 Correlation between quarterly industry survey data and actual growth in equipment investment Euro area Germany Spain France Italy October Industry Survey Capacity utilisation Assesment of capacity Demand limiting productionEquipment limiting production 0.83 -0.83 -0.87 0.62 0.76 -0.74 -0.87 0.48 0.58 -0.65 -0.69 0.69 0.37 -0.49 0.22 0.33 0.57 -0.58 -0.33 0.48 Euro area Germany Spain France Italy July Industry Survey Capacity utilisation Assesment of capacity Demand limiting productionEquipment limiting production 0.80 -0.84 -0.81 0.72 0.81 -0.80 -0.77 0.56 0.48 -0.66 -0.72 0.74 0.21 -0.28 0.19 0.29 0.55 -0.58 -0.33 0.61 13 • Signs of the correlation coefficients are correct – Investment is negatively correlated with the factor "insufficient demand limiting production” and with the “capacity assessment” – The correlation is positive for the results obtained from the capacity utilisation question and the other factors limiting the production. • For the euro area, correlation coefficients are high using October Industry survey results: • 0.83 with capacity utilisation • -0.83 with the assessment on production capacity and • -0.87 with "demand" as factor limiting the production • • • For Germany, high correlation coefficients with capacity utilisation as reported in the July (0.81) and in the October industry surveys (0.76). In addition, correlation is -0.87 with "demand" as factor limiting the production in the October survey. For Spain, coefficients with factors "demand" and "shortage of material and/or equipment" with the July industry survey are -0.72 and 0.74, respectively. For France and Italy, correlation coefficients are lower than with the investment survey 14 3. Usefulness in forecasting (explanatory power) • • • • Basic regression method Correction for bias by using a constant Augmentation of AR model by investment plans Degree of improvement of fit over AR European Commission 2007 15 Table 1: In-sample statistics, euro area adj. R² const. p-value invest. plan p-value AR(1) 0.11 1st plan 0.47 -0.81 0.606 1.24 0.0084 2nd plan 0.71 -0.07 0.939 0.80 0.0002 3rd plan 0.67 2.44 0.029 0.69 0.0006 definitive 0.83 2.18 0.010 0.77 0.0000 N=15 European Commission 2007 16 In-sample results: euro area • Including 2nd plans improves R² to 0.71, slightly lower for 3rd plans • All survey vintages are Granger-causal • Significant underestimation of investment growth in t in the January and June releases of t+1 • Based on 16 yearly observations only (since 1992) purely indicative results • 23 observations available for FR and IT European Commission 2007 17 Table 2: in-sample fit France and Italy adj. R² France Italy AR(1) 0.06 0.05 1st plan 0.23 0.31 2nd plan 0.51 0.33 3rd plan 0.66 0.61 definitive 0.67 0.47 N=22 European Commission 2007 18 In-sample results: France and Italy • Goodness of fit increases monotonously with maturity of investment plans • But: “definite investment plans” surveyed in t+1 (Mar/Apr) for t do not carry more information than 3rd plans (Oct/Nov) • For Italy, explanation becomes even worse General problem: in-sample statistics overstate forecasting power European Commission 2007 19 Out-of-sample forecasting • Euro-area time series too short • some indicative results for France and Italy • Division of sample: initial estimation over 19852000, forecast for 2001 • recursive extension of sample until 2006, forecast for 2007 • RMSE to reflect forecast uncertainty over 20012007 European Commission 2007 20 France 16 16 16 EQUIP_NOM PROG_1 EQUIP_NOM EQUIP_NOM PROG_2 12 12 12 8 8 8 4 4 4 0 0 0 -4 -4 -4 -8 -8 -8 86 88 90 92 94 96 98 00 02 04 06 PROG_3 86 88 90 92 94 96 98 00 02 04 86 06 88 90 92 94 96 98 00 02 04 06 Italy 20 20 EQUIP_NOM 20 PROG_1 EQUIP_NOM PROG_2 EQUIP_NOM 15 15 15 10 10 10 5 5 5 0 0 0 -5 -5 -5 -10 -10 -10 -15 -15 86 88 90 92 94 96 98 00 02 04 06 PROG_3 -15 86 88 90 92 94 96 98 00 02 04 06 86 88 90 92 European Commission 2007 94 96 98 00 02 04 21 06 Table 3: RMSEs for France and Italy RMSE France Italy 1st plan 3.53 3.60 2nd plan 3.63 4.62 3rd plan 3.32 2.43 N=7 (2001-2007) European Commission 2007 22 Out-of-sample results (1) • RMSEs computed with respect to final NA figures • RMSE of first NA estimates with respect to final estimates is 0.91 for the euro area (total business inv.), likely higher for individual MS • Second investment plans perform worse than first • Superiority of third investment plans, esp. for IT • Forecasts look reasonable • Benchmark?? European Commission 2007 23 Out-of-sample results (2) • Commission’s Autumn Forecast, iterative-analytic, produced in October, published early November • i.a. forecast for real growth in equipment investment in t+1 (amplitude and general pattern very close to nominal growth) • RMSE over 2001-2007: 2.81 (FR) and 4.89 (IT) • Simple regression using 3rd investment plans: 3.32 (FR) and 2.43 (IT) Comparable or even superior accuracy (IT) But: results are released in late January only European Commission 2007 24 Summary of forecast evaluation • Evidence for usefulness in forecasting is mixed • High in-sample explanatory power • But: low benchmark (past investment, low autocorrelation) • Many potential macroeconomic/survey data that might encompass info of investment survey • Out-of-sample: indication of usefulness for forecasting, but results are needed earlier European Commission 2007 25 Conclusions and possible improvements Investment survey contains useful information For structural analysis and for forecasting But data are available very late Release the results earlier? Survey covers manufacturing industry only Extend the survey to other sectors? Quarterly survey would allow estimating quarterly investment growth (see INSEE) Increase the frequency (4 time per year)? European Commission 2007 26