MacroGreece

The way global landscape has been crafted the last two decades and the speed in which economic and financial conditions alter, inevitably has given rise to the construction and application of elaborate models, which nowcast gross domestic product (GDP) and monitor economic activity of a country in real time. Since GDP and other core macroeconomic indices are available at quartely basis and their value is finalized long after the end of the reference quarter, the use of such tools for an early assessment, becomes urgent. For instance, the GDP announcement for the US economy is up to one month after the end of the quarter, while for the Euro Area this time is close to two months after the end of the quarter. In the case of Greece, this period is substantially prolonged to something less than three months. Thus, nowcasting, which literally is the “prediction of present, the very near future and the very recent past” (Bańbura et al., 2013), becomes an absolute necessity. Almost all central banks of advanced economies have adopted “unofficially” some type of a nowcasting model, in order to take a flash estimate, prior to official announcements, on how growth rate, inflation rate, wage growth etc., are about to fluctuate. Below we share two figures, from the interactive webpages of the Federal Reserve Bank of New York and Bank of Slovenia, where their nowcasting tools present estimates on the current quarter, that are updated on weekly basis.

These two sites of central banks are only indicative of the tools that are publicly available. There is also the innovative application of Atlanta’s Federal Reserve Bank. There are millions of individuals who get informed on a regular basis from Atlanta Fed’s "EconomyNow" app, which provides updated estimates on GDP, private consumption, business uncertainty or unemployment. Igniting from their nowcasting framework, economic analysis is displayed through interactive and customizable charts and other advanced tools available to all interested parties. Definitely, the need of getting on time outlooks of the economic performance of a country, prior to official announcements, touches almost all economic agents, from policy makers to economists, econometricians, stakeholders, markets professionals, analysts, publicly listed firms, private sector entities, academics and especially individuals. Individuals are the ones who become massively affected by economic uncertainty, climate uncertainty and policy implementation and need a timely economic outlook. 

Motivated by these applications and trends, MacroGreece, will develop a sophisticated nowcasting modelling framework for the components of the Greek GDP with the guidance of its principal investigator Prof. Stavros Degiannakis, a leading expert in modelling and forecasting large scale data who was and still is the first in nowcasting literature to propose a disaggregated nowcasting framework for the Greek GDP (Degiannakis, 2023). MacroGreece by updating and extending original architecture of Degiannakis (2023) will dive deeper into the core of the Greek economy by further considering the perilous and uncertain conditions that lately have been emerged more aggressively than ever due to COVID-19 pandemic and climate change and transform this information to a unique multifunctional interactive tool at the hands of the Greek policy makers and concerned parties, for a more oriented decision making process. This framework would serve as an edgy reference point for the various economic sectors and the community in total. Moreover, innovate and powerful tools are the ones which diffuse targeted information and guide the lives of professionals and individuals towards more sustainable paths and refined policy implementations. 

There are many studies dedicated in exploiting nowcasting frameworks, effectively or not, most concentrating on quarterly GDP. The studies of D’Agostino, Giannone and Surico (2006), Boivin and Ng (2006), Stock and Watson (2005a, 2005b) and Giannone, Reichlin and Small (2008) which were incorporated for the U.S. economy using Dynamic Factor models (DFMs) or Factor Models (FMs), are some of the most influencial papers. the reported findings are rather blurry, concerning the proposed methodologies. Bańbura and Rünstler (2011) also adopt DFMs, in order to investigate the nowcasting accuracy for the Euro Area and the same framework apply Giannone et al. (2009). Nevertheless, DFMs, are not the only models that are part of the classic nowcasting arsenal. There is another framework that is also frequently incorporated and that is the Bridge models (BMs). Remarkable studies adopting BMs are those of Baffigi, Golinelli and Parigi (2004), Angelini et al. (2011), Antipa et al. (2012) for the Euro Area and other advanced European countries as Germany, Italy and France. Their models’ performance varies, but for Antipa et al. (2012) results favor BMs against DFMs, while the opposite is true for Angelini et al. (2011) or Barhoumi et al. (2008), also for the Euro Area, as the evidence they provide, favors DFMs. Of course, this does not come as a surprise, as BMs come with some disadvantages that DFMs seam to overcome. 

DFMs and BMs seem to dominate in nowcasting literature but there are not the only ones. There is another family of frequently used models, namely, the Bayesian Vector Autoregressive models (BVARs). Studies incorporating BVAR frameworks, are those of Iacoviello (2004) and Itkonen and Juvonen (2017) for the case of Italy and Finland, respectively, but with mixed results since all tested models proved to have similar outcomes on forecast evaluation metrics. The opposite findings are reported by Cimadomo et al. (2020), who nowcast U.S. quarterly GDP, illustrating how efficient BVARs are, in tackling with mixed frequency time series, outperforming DFMs. Mixed frequencies time series is a major issue in nowcasting, that is the reason why another framework appears in the nowcasting arsenal, the mixed data sampling frameworks (MIDAS). MIDAS models have been incorporated by Ghysels et al. (2006), Andreou, Ghysels and Kourtelos (2010), Clements and Calvao (2009), who find their ability to relate response variables of quarterly frequency with explanatory variables of higher frequency, the most appropriate for extracting information from core indicators, and provide evidence of their superiority against the competitive ones. Combinations of MIDAS and Factor models have also been proposed. Marcellino and Schumacher (2010) adopt a Factor-MIDAS framework for the German GDP and highlight MIDAS superiority. The same goes for Andreou, Ghysels and Kourtelos (2013) and Kim and Swanson (2018) for the Euro Area and the Korean economy, respectively. 

Lately, we have witnessed, along the traditional nowcasting frameworks, the application of machine learning methods (ML) or neural network (NN) that are related to deep learning environments, that could possibly lead to innovations in the nowcasting literature. Chinn, Meunier and Stumpner (2023) nowcast world trade by incorporating random forest and gradient boosting models. They test them against traditional linear and nonlinear frameworks and report improved forecast accuracy in favor of the machine learning methods. Kant, Pick and Winters (2022) incorporate a Random forest model, against DFM in order to nowcast the Dutch GDP. They report that random forest delivers a shift in the nowcasting performance probably due to its ability to capture undergoing nonlinearities. Richardson, Mulder and Vehbi (2021) go a step further and adopt a pool of such methods such as boosted trees, elastic net, ridge regression, support vector machines, neural networks in order to forecast New Zealand GDP growth. They compare their performance against autoregressive models, factor model and BVAR and end up with rather mixed findings.  

Hence, literature has provided a plentiful of studies adopting classic econometric techniques, more sophisticated techniques or both of them and that in turn provides fertile soil for the in-depth realization of their functions, their limitations but also the way to benefit from their weaknesses and propose novel versions and combinations. Nowcasting GDP requires the exploitation of enormously large datasets of monthly and even daily variables, which traditionally are thought to be strongly linked with GDP and enter in the model frameworks as explanatory variables-factors. Unfortunately, the inclusion of large and diverse datasets encloses risks. One of these risks is multicollinearity. Another risk is that some of these variables do not have the required length to allow for precise coefficient estimation, valid for data that are not around for quite a long time, but certainly worth to be included at studies. Additionally, another risk generates from publication, as monthly soft data (surveys), become available earlier than hard ones. Finally a major drawback is the extreme data revisions. Thus these frameworks are not developed just for academic demonstration but for direct implementation in real applications and challenges. MacroGreece, stands at this very point by developing a novel and advanced modelling framework, interactively supported by an integrated Graphical User Interface (GUI). As such it will deal with the recorded caveats and will capture all substantial elements that miss from the Greek reality, reflecting a rational necessity spanning from the need to go with the flow and align with the global nowcasting trends.

References

Andreou, E., Ghysels, E. and Kourtellos, A. (2010). Regression Models with Mixed Sampling Frequencies, Journal of Econometrics, 158, 246-261.

Andreou, E., Ghysels, E. and Kourtellos, A. (2013). Should macroeconomic forecasters use daily financial data and how? Journal of Business and Economic Statistics, 31, 240-251.

Angelini, E., G. Camba-Mendez, D. Giannone, L. Reichlin and G. Rünstler (2011). Short-term forecasts of Euro Area GDP growth. Econometrics Journal, 14(1), 25–44.

Antipa, P., K. Barhoumi, V. Brunhes-Lesage and O. Darne (2012). Nowcasting German GDP: A comparison of bridge and factor models. Journal of Policy Modeling, 34(6), 864–878.

Baffigi, A., R. Golinelli and G. Parigi (2004). Bridge models to forecast the Euro Area GDP. International Journal of Forecasting, 20(3), 447– 460.

Bańbura, M. and G. Rünstler (2011). A look into the factor model black box: Publication lags and the role of hard and soft data in forecasting GDP. International Journal of Forecasting, 27(2), 333–346.

Bańbura, M., Giannone, D., Modugno, M., and Reichlin, L (2013). Nowcasting and the real time data flow, European Central Bank, Working Paper Series, 1564.

Barhoumi, K., Benk, S., Cristadoro, R., Reijer, A.D., Jakaitiene, A., Jelonek, P., Rua, A., Rünstler, G., Ruth K. and Nieuwenhuyze, C. Van (2008).  Short-term forecasting of GDP using large monthly datasets: a pseudo real-time forecast evaluation exercise, European Central Bank, Occasional Paper Series, 84, 1-25.

Boivin, J. and S. Ng (2006). Are more data always better for factor analysis? Journal of Econometrics, 132(1), 169 – 194.

Chinn, M., Meunier, B. and Stumpner, S. (2023). Nowcasting world trade with machine learning: a three-step approach. European Central Bank, European Paper Series No. 2836.

Cimadomo, J., Giannone, D., Lenza, M., Monti, F. and Soko, A. (2020). Nowcasting with large Bayesian vector autoregressions. European Central Bank, Working Paper Series No 2453.

Clements, M.P. and Galvão, A.B. (2009). Forecasting US output growth using leading indicators: An appraisal using MIDAS models. Journal of Applied Econometrics, 24(7), 1187-1206.

D'Agostino, A., Giannone, D., and Surico, P. (2006). (Un)Predictability and macroeconomic stability. European Central Bank, Working Paper, 605.

Degiannakis, S. (2023). The D-model for GDP nowcasting. Swiss Journal of Economics Statistics 159, 7 (2023). https://doi.org/10.1186/s41937-023-00109-8.

Ghysels, E., Santa-Clara, P. and Valkanov, R. (2006). Predicting volatility: Getting the most out of return data sampled at different frequencies, Journal of Econometrics, 131, 59–95.

Giannone, D., Reichlin, L. and Simonelli, S. (2009). Nowcasting Euro Area Economic Activity In Real Time: The Role Of Confidence Indicators. National Institute Economic Review, 210, 90–97.

Giannone, D., Reichlin, L. and Small, D. (2008). Nowcasting: the real-time informational content of macroeconomic data, Journal of Monetary Economics, 55, 665–676

Iacoviello, M. (2001). Short-term forecasting: Projecting Italian GDP, one quarter to two years ahead, International Monetary Fund, IMF Working Papers 01/109.

Itkonen, J. and Juvonen, P. (2017). Nowcasting the Finnish economy with a large Bayesian vector autoregressive model, BoF Economics Review, No. 6/2017, Bank of Finland, Helsinki.

Kant, D., Pick, A. and Winter, J. (2022). Nowcasting using machine learning methods. DeNederlandscheBank, Working Paper No. 754.

Kim, H.H. and Swanson, N.R. (2018). Methods for backcasting, nowcasting and forecasting using factor‐MIDAS: With an application to Korean GDP. Journal of Forecasting, 37(3), 281-302.

Marcellino, M. and Schumacher, C. (2010). Factor MIDAS for nowcasting and forecasting with ragged-edge data: a model comparison for German GDP. Oxford Bulletin of Economics and Statistics, 72(4), 518–550.

Richardson, A., Mulder, T. and Vehbi, T. (2021). Nowcasting GDP using machine-learning algorithms: A real-time assessment. International Journal of Forecasting, 37(2), pp. 941-948.

Schumacher, C. (2010). Factor forecasting using international targeted predictors: The case of German GDP, Economics Letters, 107, 95–98.

Stock, J. H. and M. W. Watson. (2005a). Implications of dynamic factor models for VAR analysis, National Bureau of Economic Research, Working Paper No. 11467.

Stock, J. H. and M. W. Watson. (2005b). An Empirical Comparison of Methods for Forecasting Using Many Predictors. Princeton University, Working paper.