Built-in Feature for Estimating Polynomial Adjustment Cost (PAC)

[tvonbrasch]

I propose adding a built-in estimation and simulation feature for Polynomial Adjustment Cost (PAC) models to EViews.

PAC models are dynamic adjustment models that specify how agents gradually adjust a decision variable, such as investment or prices, toward a target, while minimizing both deviations from the target and the costs of changing the variable and its growth rates. The framework explicitly models inertia and anticipatory behavior by including lagged deviations from target, lagged changes, and a weighted sum of expected future changes in the target variable.

The PAC methodology is now widely used in leading policy models. It is central to the dynamic equations for investment, consumption, prices, and wages in the Federal Reserve Board’s FRB/US https://www.federalreserve.gov/econres/ ... -about.htm. The European Central Bank’s ECB-BASE and ECB-(RE)BASE models also use PAC to model sluggish adjustment and expectations formation in key macroeconomic relationships (https://www.ecb.europa.eu/pub/pdf/scpwp ... a53.en.pdf). Similarly, Saffier 3.0, developed by the CPB Netherlands Bureau for Economic Policy Analysis, implements PAC (https://www.cpb.nl/sites/default/files/ ... ground.pdf).

To implement PAC models, the following process is typically followed: the PAC equation is estimated using observed data and an initial expectation proxy, such as a VAR or AR forecast. The estimated model is then used to generate new forecasts for the target variable, which are in turn used to update the expectation proxy. This process is repeated—re-estimating the PAC equation and updating expectations—until convergence is achieved, meaning that the expectations used in the PAC equation are consistent with the model’s own forecasts. This requires integration between estimation, expectation, and model simulation objects, with each iteration updating the expectation proxy based on the most recent forecasts.

EViews’ robust capabilities in estimation and model simulation make it well suited for PAC models. Integrating PAC as a standard feature, with support for iterative estimation and expectation updating, would be an important addition that would further enhance EViews’ utility as a tool for macroeconomic modeling.

For reference on the details of implementation, please see the attached documentation from the FRB/US team.

[tvonbrasch]

Note also that PAC is now included in Dynare, see: https://www.dynare.org/manual/the-model ... c-equation

[xprimexinverse]

I would like to second this motion. Excellent suggestion that would be very useful.