Financial and Macroeconomic Leading Indicators of Industrial Production: A Financial Markets' Perspective
Abstract
IntroductionFinancial market practitioners advise clients and investors on market research opportunities through the coordination of economic research and portfolio management. The coordinated approach sees economists... [ view full abstract ]
Introduction
Financial market practitioners advise clients and investors on market research opportunities through the coordination of economic research and portfolio management. The coordinated approach sees economists providing research to be used for trading by portfolio managers and the emanation of research to clients. In turn, this allows for the implementation of investment strategies based on the research to match the risk profile and liabilities of their client base. Practitioners rely on real-time, high frequency information, leading to decision-making within a high-paced environment. In light of the time constraints and market expertise they tend to focus on a limited set of leading indicators when making their economic activity forecasts. Generally financial market practitioners focus on a sub-set of the vast pool of economic and financial data when making their growth forecasts. This paper examines whether the small datasets used by practitioners are more effective at forecasting activity compared to the vast datasets used in the literature.
This paper constructs a series of leading composite indices to forecast economic growth in the period prior to the financial crisis in 2008, post the crisis, and during the recovery period up to 2016. The financial and macroeconomic indicators driving the composite are derived from structured interviews with economists working in the International Financial Service Centre in Ireland (IFSC). The IFSC harbours a wide spectrum of global financial institutions and it provides therefore a strong representative sample of the global population. The interviews took place between 2013 and 2016 and included large market players such as Bank of America, Barclays, BNP Paribas and Deutsche Bank as well as local players such as Bank of Ireland, Allied Irish Bank, Merrion Capital and Davy’s. We build two sets of composites for each county. The first is built using the vast array of macro and financial data suggested by the literature while the second set uses the indicators selected by the market practitioners. Each set includes composites for the pre-Lehman, post-Lehman and recovery periods.
Literature Review
A common feature of the literature is to include disaggregated versions of indicators from multiple sectors. In the US, Stock and Watson (2004), Boivin and Ng (2006), Giannone, Reichlin and Sala (2004), D’Agostino and Giannone (2006) and Castle, Clements and Hendry (2012) use similar datasets covering multiple indicators from housing, manufacturing, employment, trade, consumption and survey sectors. In Germany and the UK, Schumacher (2007) and Artis et al (2005) incorporate disaggregated indicators from the same sectors as the US literature. The literature finds mixed results in both factor composition and forecasting performance of both static and dynamic models. Giannone, Reichlin and Sala (2004) find that only two factors explain 60% of the variance in their dataset for the US, while Artis et al (2005) find that six factors explains 50% of the movement in their dataset for the UK.
The literature reports mixed results comparing the relative forecasting performance of static and dynamic models. Schumacher (2007) uses a static and dynamic factor model to forecast German growth, using rolling and recursive forecasting schemes. The results show optimal forecasting performance using the recursive forecasting scheme but no additional benefit of using the dynamic model over the static model. Similarly, D’Agostino and Giannone (2006) report similar forecasting performance for static and dynamic models in forecasting US growth. The literature also finds conflicting results for the size of the dataset used to forecast growth across Germany, the UK and the US. Bahroumi et al (2010), fails to find additional forecasting performance using disaggregated data compared to aggregated data in forecasting French GDP. Boivin and Ng (2006) find that a small dataset produces better or even results when compared to a larger dataset in the US. Stock and Watson (2004) argue that the actual indicators in the dataset have more importance to forecasting German, UK and US growth. The improved forecasting performance is attributed to the inclusion of financial indicators rather than actually using a smaller dataset.
Methodology
The datasets in this paper are pre-screened to ensure that only leading indicators are used in the analysis and two approaches are implemented to isolate leading indicators for the static and dynamic models. Granger causality and cross-correlation tests are implemented to isolate leading indicators within the static method, while cross-correlation and mean delay analysis are used for the dynamic approach. This paper uses PCA to extract factors using both static and dynamic methods to construct composites, in which each leading indicators is assigned a weight based on the PCA results. The composites are based on the datasets derived from the practitioner structured interviews and the literature. In addition, this paper compares the performance of composites constructed using the literature by benchmarking forecasting errors of both sets of composites using out-of-sample recursive forecasting windows. The relative forecasting performance is conducted based on Clark and McCracken (2001) forecast encompassing.
Results
The PCA results show a change in the composites within the time periods as we find intertemporal variation among the assigned weights of key indicators and the content in each of the composites. There is a clear shift from manufacturing in the pre Lehman period to financial and interest rate indicators in the post Lehman period, with surveys remaining prominent in both periods. The static composites reveal that yields are prominent in both time periods in the US, showing the importance of financial markets in the US as a global market driver in financial markets. In the UK, housing indicators are important in both time periods, thereby providing evidence of the role of credit in driving the UK economy and overlapping with the research of Reinhart and Rogoff (2009) and Ng and Wright (2013). The dynamic composites for Germany show that a high weight is given to surveys and manufacturing within practitioner and literature composites. However in the post Lehman period, a shift to interest rates and financial indicators is evident. In the UK, aggregated and disaggregated FTSE and prices are the key drivers of the composites in both time periods, with disaggregated manufacturing retail more prevalent for the post Lehman period. The key sectors driving the composites in the US are financial, surveys, housing and manufacturing.
The forecasting results for Germany and the US reveal that the practitioner and literature composites outperform the benchmark models in both periods. In the UK, the practitioner composites are the most consistent, outperforming the benchmarks in both time periods. The static model generally outperforms the dynamic models, particularly in the US and in the UK. The results provide strong evidence showing robust forecasting power of the composites constructed using the leaner pool of indicators availed of by market practitioners. The estimated factor models explain the observed fluctuations in real economic activity leading us to use the composites for forecasting, both prior to and in the aftermath of the Lehman Brothers crash in 2008.
Contribution
This paper makes a number of unique contributions. Firstly, the novel dataset constructed, following the structured interviews with market practitioners, allows us to gain insight into the practitioners’ approach to forecasting. This paper uses the expertise of these practitioners to attain a unique set of indicators, rather than relying on indicators suggested by the literature. Next this concise pool of indicators is used in both static and dynamic PCA to build the series of composite both prior to and in the aftermath of the Lehman crisis in 2008. Using forecasting tools, these composites are compared to the standard literature composites thus identifying the additional insight practitioners have into market activity.
REFERENCES
Artis, M. J., Banerjee, A. & Marcellino, M. 2005. ‘Factor forecasts for the UK’. Journal of Forecasting, 24, 279-298.
Barhoumi, K., Darne, O. & Ferrara, L. 2010. ‘Are disaggregate data useful for factor analysis in forecasting French GDP?’ Journal of Forecasting, 29, 132-144.
Boivin, J and Ng, S 2006. ‘Are more data always better for factor analysis?’ Journal of Econometrics, 132, 169-194.
Castle, J.L., Clements, M.P & Hendry, D.F. 2012, ‘Forecasting by factors, by variables, by both or neither?’ Journal of Econometrics, 177, 305-319
Clark, T.E. and McCracken, M.W., 2001. ‘Tests of equal forecast accuracy and encompassing for nested models’. Journal of econometrics, 105(1), 85-110.
D’Agostino, A. & Giannone, D 2006, ‘Comparing alternative predictors based on large-panel factor models’. ECB Working Paper, No. 680
Giannone, D., Reichlin, L., & Sala, L. (2004). Monetary Policy in Real Time. NBER Macroeconomics Annual, 161-200.
Ng, S. & Wright, J. H. 2013. ‘Facts and challenges from the great recession for forecasting and macroeconomic modeling’. National Bureau of Economic Research.
Reinhart, C.M. and Rogoff, K.S., 2009. ‘The aftermath of financial crises’. American Economic Review, 99(2), 466-72.
Schumacher, C 2007, ‘Forecasting German GDP using alternative factor models based on large datasets’. Journal of Forecasting, 26, 271-302.
Stock, J.H. and Watson, M.W., 2004. ‘Combination forecasts of output growth in a seven‐country data set’. Journal of Forecasting, 23(6), 405-430.
Authors
- Austin Brady (University College Cork)
- Geraldine Ryan (University College Cork)
- Ella Kavanagh (University College Cork)
Topic Area
Topics: Accounting, Finance and Corporate Governance
Session
AFCG - 1 » Accounting, Finance and Corporate Governance - Session 1 (15:45 - Monday, 3rd September, G02)
Presentation Files
The presenter has not uploaded any presentation files.