Resumen:

La inversión es una variable que, por un lado, determina en gran medida las posibilidades de crecimiento a largo plazo de la economía y, por otro, al ser el componente más variable de la demanda agregada, sus oscilaciones condicionan de manera decisiva las fluctuaciones cíclicas de la producción y del empleo. El objetivo de este trabajo es construir un número reducido de indicadores sintéticos mensuales y de modelos de predicción de la inversión para Galicia, a partir de un conjunto de indicadores componentes. Los modelos propuestos permiten seguir la evolución mensual de la inversión en la economía gallega y satisfacen las necesidades de información a corto plazo del análisis económico.

Abstract:

Productive activity requires a certain level of capital so that the production processes of new goods and services can be carried out, the 'stock' of capital. The national accounting measures one of the flows that modify this level through gross capital formation (FBK), which is understood as the acquisition of machinery and buildings (Gross Fixed Capital Formation-FBKF), plus the change in inventories. Investment is a variable that, on the one hand, largely determines the long-term growth possibilities of the economy and, on the other, being the most variable component of aggregate demand, its fluctuations decisively condition cyclical fluctuations of production and employment. Therefore, knowledge of its evolution and its determinants will make it possible to design efficient economic policies that help growth and stabilize economic fluctuations. Investment in Galicia accounts for 17% of the GDP (IGE, 2021), which is lower than that of Spain, which stands at 21%. Investment is characterized as a variable with high volatility that amplifies fluctuations in the cyclical profile of economic activity. Despite this, it holds certain value; it can anticipate activity peaks and confirm its lows. Additionally, it plays a subsidiary role in monitoring the savings component of the GDP in a regional statistical-economic space such as Galicia. Regarding labor factor, there are an association between the average propensity to invest and the employment rate (employed/population aged 16 and over). Both exhibit a similar profile, especially before the great recession, although the employment rate may be more seasonal. However, recently, the fraction of GDP allocated to investment is lower, more in line with more developed economies, whereas previously this fraction was higher due, in part, to Galicia starting from a less developed situation, hence the need for European funding. In Galicia, the results of economic accounts are quarterly (IGE, 2021) and are published with a two-month delay. One month before they are released, many of the monthly indicators used to estimate the data provided by accounting would already be available. Therefore, the publication frequency may slow down the adoption of short-term measures and strategies, making it very useful to construct indicators and models that allow for monthly estimations and predictions. Due to the difficulty of employing a theory that explains economic cycles, it is necessary to use indicators that reflect their fluctuations. Indicators allow us to approximate the behavior of a particular variable, either because it is not available, or because there is a delay or it is not available with the necessary frequency. Therefore, the objective of this work is to build a reduced number of monthly synthetic indicators and investment prediction models for Galicia, based on a set of component indicators. With this analysis, the increasingly immediate information needs required for economic analysis are satisfied. Consequently, the treatment in this article will be phenomenological, in the scientific sense, without delving deeply into economic theory issues. The initial nine series or component indicators used to build the synthetic indicators provide a measure of some aspect of the evolution of investment. After that, we will reduce the size of the series using two statistical techniques, dynamic factor analysis (DFA) and dynamic principal component analysis (PCA). The issue with the series used is that they do not have the same timeframe; in other words, many series start too late, and some end too early. Therefore, for the estimation of the final indicators, an iterative procedure will be followed, which is partially based on the one used by Cuevas & Quilis (2012). Two investment models are proposed using the factors or components obtained from the previous procedures. We use for this purpose two generalized additive models (GAM). In order to conduct the analysis that relates the FCIS or the principal components with the FBK, it is necessary for the observation frequency of all variables involved in the analysis to be the same. The response variable is the FBK, which is published quarterly, but the explanatory variables have a monthly frequency. Therefore, in order not to lose information, the FBK was converted into a monthly frequency. In order to assess the predictive capacity of the two models, a simulation study has been designed in which the root of the relative mean square error (RRMSE) was calculated. A clear gain is observed in the model that uses the DFA compared to that used by the PCA, in addition to the fact that the factor is easier to interpret. Then, this paper presents two methodologies that are used to monitor investment in Galicia on a monthly basis, one of them based on the DFA and the other on the PCA. Therefore, the novelty of this study lies in the use of available and publicly available information to characterize the monthly behavior of the investment. This allows a better approximation of the monthly evolution of investment in the Galician economy and, in addition, satisfy the short-term information needs of the economic analysis. In terms of results, it is necessary to point out the good behavior of the FCIS when compared to the evolution of the FBK, calculated by the IGE. The high degree of coupling between the evolution of both (FCIS and FBK), with a Pearson correlation coefficient of 0.79 (95% confidence interval of (0.74; 0.82)), demonstrates their coherence. Both indicators exhibit very similar behavior in all phases of the economic cycle within the considered time frame. Ultimately, this constitutes an empirical validation of the calculated indicator. Indeed, the FCI, together with the DFA-GAM model, serves as an indicator of investment evolution and allows for the adoption of short-term measures and strategies. Within a quarter, with one month's advance notice, it will be possible to anticipate the evolution of the FBK in the respective quarter. Also, the FCI serves as a tool for monitoring investment evolution, which, in our opinion, remains somewhat constrained after the great recession. In the initial years following the crisis, this seems logical due to the saturation effect of the new housing market with the excess supply in preceding years, but in more recent years, there perhaps should have been a greater recovery. Finally, a easily replicable methodology is presented, both with the information available in another region and with information at the national level or even in other countries. It should be noted that if the objective is a country, there is usually more data series related to investment available, therefore, it would be advisable to expand the initial series from which this work starts