Experimental/lifetime draws

[andrewbaxter439]

What

• closes Add probability of re-sampling random draw from labour transitions to stabilise employment transitions #176 (also in reference to Labour supply is currently producing strange estimates for employment transitions #163)
• adds probability of re-sampling labour supply random draw
• Adds new labour supply estimates

Why

• aiming to stabilise employment transition rates

[Copilot]

Pull Request Overview

This PR implements an experimental lifetime draws mechanism to improve employment transition rate stability in the labor supply model. The main purpose is to add probability-based persistence for labor supply random draws rather than generating completely new random values each period.

Key changes:

• Introduces employment and unemployment persistence probabilities to control how often labor innovations persist from previous periods
• Adds new labor supply regression variables for education-work hour combinations
• Updates model parameters including starting year range and column counts for regression matrices

Reviewed Changes

Copilot reviewed 3 out of 4 changed files in this pull request and generated 4 comments.

File	Description
Innovations.java	Adds new constructor to support single-draw double innovations for the persistence mechanism
BenefitUnit.java	Implements labor innovation persistence logic and adds new regression variables for education-work combinations
Parameters.java	Updates persistence probability constants, year ranges, and regression matrix column counts

Comments suppressed due to low confidence (1)

[andrewbaxter439]

Explanation:

Probabilities of 'persisting' a previous labour supply innovation are set in Parameters.java at the top:

    // Determine probability of yearly labour supply matches persisting from previous year
    public static double labour_innovation_employment_persistence_probability = 0.9;
    public static double labour_innovation_unemployment_persistence_probability = 0.5;

It then runs a new method with this to determine whether to assign a new labour innovation value or persist the previous one:

   private double getLabourInnovation(double persistenceProbability) {

        double newLabourInnovation = innovations.getDoubleDraw(5);

        // persistenceRandomDraw - random value to determine if we keep old innovation
        double persistenceRandomDraw = innovations.getDoubleDraw(6);

        int currentYear = model.getYear();

        // Initialize on first call
        if (lastLabourInnov == null || lastYear == null) {
            lastLabourInnov = newLabourInnovation;
            lastYear = currentYear;
            return newLabourInnovation;
        }

        double labourInnov;
        if (persistenceRandomDraw < persistenceProbability) {
            labourInnov = lastLabourInnov;
        } else {
            labourInnov = newLabourInnovation;
        }

        // Store values for next period
        lastLabourInnov = labourInnov;
        lastYear = currentYear;

        return labourInnov;



        }

These allow differing 'churns' for those entering/exiting employment.

[andrewbaxter439]

• Change to "not in employment" rather than "unemployment"

[andrewbaxter439]

@matteorichiardi thats it working with terms updated to note "not in employment" rather than "unemployed". Seems ready to merge then can do further tweaks as new labour supply estimates come along?

[andrewbaxter439]

Hi @justin-ven and @matteorichiardi - have noted you as 'reviewers' but just to sense check from last few weeks' discussion whether this is now ready to merge in?