MVED: Multi-variate Earning Dynamics Package for R

A multivariate approach on Income Decomposition

This is a first implementation of the research on Earning Dynamics modeling improvement, which models the dynamics of an earning process. Here we implement an multi-variate approach.

Equations

yit = αi + pit + τit

pit = φppit−1 + ξit,

τit = θ(L)�it

• yit: is individual i’s log-earnings (residuals) at time t;

• pit: is the permanent component (random walk if φp = 1);

• τit: is the transitory component: MA(1), ARMA(1,1), AR(1), or iid;

• αi: is an individual fixed effect

  MARIMA Model

1. We pre-train an MARIMA model to obtain a full AR and MA coefficients of the multi-variate process.

2. Then, we train the model with p,q,d (1,0,1) as shown in the literacy.

3. We have that a model of the permanent effect of the shock is equivalent to a Random Walk model first order of differentiation, we simulate this with the previously obtained model and thus we obtain this resource.