Econometrics

 

For Econometrics is the process of quantitatively analyzing economic data. Its goal is to detect patterns and quantify relations between variables. By constructing this information, econometrics allows us to infer the future evolution of variables.

 

There are several econometrical tools and they have a myriad of applications.

 

First, when forecasting the price of a commodity, econometrics helps us to quantify the possible impact of each factor on the price. Since a price in the financial world is obtained from a complex structure, its prediction has to take into account several forces that can have opposite directions and can be related to each other. With econometrics we can disentangle these forces and forecast with more accuracy.

 

Second, we can estimate the relation between our client’s revenues and several other variables, in order to forecast them in our model. More particularly, we can assess how much our client’s revenues increase with the increase in GDP, or any other relevant variable. Third, we also know that, in bad times, the economic environment becomes more volatile, so how much do our client’s revenues become more volatile with recessions? This type of questions is important for our simulations, and this is quantified with Econometrics.

 

At Watson & Noble, we apply the state of the art econometric techniques. We apply customized econometric models to each problem, which enables us to minimize prediction errors as well as to assess the most plausible relation between variables:

  • VAR models: to estimate the impact of several time-series in each other

  • GARCH models: to clean the impact of heteroskedasticity in data

  • Markov-Switching models: to estimate the existence of several different regimes

  • Smooth-transition models: to make the transition between regimes smooth

  • Spatial econometric models: to account for spatial factors when  space has impact on the relations

  • Robust regression: to handle outliers

  • Panel data models: to take into account the cross-sectional and time series effects

  • Binary outcome models (Logit and Probit): to infer about binary variables

  • Network analysis: to understand and infer the impact of a network structure

 

Moreover, when dealing with a continuous project with our client, our econometric methods can have a learning mechanism, in which we make the model learn with its own prediction errors.