1 / 25

EHM Theory and Structure

EHM Theory and Structure. Behavioural Labour Supply Modelling in DWP Alan Duncan, 6 th May 2009. Motivation. Limitations to tax policy evaluation in the absence of behavioural responses

gray-maddox
Download Presentation

EHM Theory and Structure

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. EHM Theory and Structure Behavioural Labour Supply Modelling in DWP Alan Duncan, 6th May 2009

  2. Motivation • Limitations to tax policy evaluation in the absence of behavioural responses • Static tax microsimulation models operate on the premise that individual behaviour remains fixed when simulating the effects of tax and welfare policy reform • this approach is perfectly appropriate for evaluating the ‘next day’ impact of tax or welfare policy reform, and for looking at reforms that aren’t likely to affect behaviour (..but how do we know?) • however, static methods are limited when evaluating reforms for which economic responses are likely (or indeed, intended) • this motivates an MDU project to add behaviour to the DWP Policy Simulation Model (PSM) to simulate the effects of policy reform on households’ employment choices

  3. Adding Behaviour • Modelling approach • empirical implementation of a structural economic model of household labour supply • uses information from static microsimulation:- for data in the estimation of the structural model- for input into the (micro)simulation of behavioural responses- find it best to use same static model for both Features of model • structural rather than reduced-form (explain rather than describe) • discrete rather than continuous (practicality, flexibility) • probabilistic (to accommodate preference heterogeneity)

  4. Structural economic model – “as if...” • Characterise behaviour (in the first instance, employment choices) to be driven by an economic model of household labour supplyEconomic foundations • Households (more accurately, tax units) are allocated a preference function that ‘ranks’ choices over working hours & income in terms of ‘utility’ or ‘happiness’ • decisions are assumed to derive from the maximisation of this preference function subject to budget constraint that is affected by taxes and welfare payments • structure of decision-making is ‘rational’ in an economic sense (choose whichever outcome yields most ‘happiness’) • basic model can be adapted to accommodate other decisions: - welfare take-up (Moffitt & Keane, IER 1999) - childcare demand (Robins, Ribar)

  5. Structural economic model – “as if...” the basic model • assumes that families choose the number of hours they want to work on the basis of ‘preferences’ over hours hand net income y, as an expression of ‘utility’ or ‘happiness’ U=U(y,h) • any hours choice implies a certain net income, comprising earned and unearned income, taking full account of the tax and welfare system (the ‘budget constraint’) y[h]=w.h+m-t(h,w, m; X) The decision rule:choose hours to maximise U subject to remaining on the constraint: maxh U=U(y[h],h) subject to y[h]=w.h+m-t(h,w, m; X)

  6. Structural economic model – “as if...” Umax yh budget constraint h chosen h

  7. Estimation • fit model parameters to the pattern of observed choices revealed in a large and representative sample of data (FRS) • estimation process acts to rationalise observed patterns of behaviour as if they derive from choosing the ‘best’ choice among the set of alternatives presumed to exist • requires a parameterisation of preferences, and for EHM we choose a quadratic direct utility:Blundell, Duncan, McCrae and Meghir, 1999 • all parameters allowed to vary with observed factors and unobserved heterogeneity • estimate using Simulated Maximum Likelihood maxh U=U(y[h],h) subject to y[h]=w.h+m-t(h,w, m; X)

  8. Restricting hours choices (discrete) yh h

  9. Restricting hours choices (discrete) yh h

  10. Restricting hours choices (discrete) yh h*=maxh U= U( h, yh| X) h

  11. Probabilistic model • discrete approach offers practical advantages in adding behaviour (simplifying taxes in estimation/simulation, facilitating household choices, modelling take-up, adding childcare) • also allows for general forms of random heterogeneity to enter into the preference function: U(h) = U(y[h], h| X,v) + eh • for given distributions for each v and eh , this gives rise to a modelled probability Pr(h= hj | X,v)of choosing hours hj over other hours choices... • ...and a probability distribution of hours responses to tax policy reform

  12. Probabilistic model

  13. Calibration (‘alignment’) • allows behavioural simulations to be compared with static benchmark • process guarantees that, wherever possible, model predictions line up with observed choices under the base (benchmark) policy regime • requires unobserved heterogeneity terms to be drawn from a conditionaI distribution to guarantee that simulations under the base system are aligned to observed patterns of data • Need to be careful - check calibration draws

  14. Probabilistic model

  15. Probabilistic model (calibrated)

  16. Issues • Validate the model- compare model simulations with ‘known’ evaluation evidence- estimate model over periods of tax policy reform • Be sensitive to model choice- ceteris paribus (‘as if...’)- ‘rational’ model better for some groups than others • Recognise limitations- model not configured to accommodate unemployment- wages and prices taken as exogenous- no interactions with demand side of labour market To Adam...

  17. EHM Practicalities Behavioural Labour Supply Modelling in DWP Adam Richardson, 6th May 2009

  18. Policy Simulation Model • Static Microsimulation Model • Models GB tax and benefit system • Based on FRS • With additional info drawn from admin data • Uprated to current year • Financial amounts • draw-down of old benefits • grossing / calibration • Written in SAS, with graphical interface • Takes less than a minute to run • Used by analysts across DWP

  19. Incorporating Behaviour • Budget constraints (several hours) • Requires entry wages for the unemployed • Preference functions • Calibration (30 – 40 minutes) • Simulation (10 minutes) • Probabilistic • Results • Validation • Compare to known reforms • Other indicators of incentives

  20. Example Results • Change: increase level of out-of-work benefits

  21. Example Results • Change in Total Employment

  22. Example Results • Change in Spending on Benefits

  23. Example Results • Employment Transitions (Lone Parents)

  24. Way Forward • Validation • Roll out to DWP analysts • Stress-testing • Expand scope of modelling

  25. Questions?

More Related