1. the importance of behavioural Endogeneity for policy projections
6TH WORLD CONGRESS OF THE INTERNATIONAL MICROSIMULATION ASSOCIATION (IMA 2017)
Justin van de Ven
27/04/2017

2. Outline Aims and motivation Method of analysis Results Conclusions
11/19/2018

3. Aims and motivation AIMS Motivation
How important are the impact effects of policy change, relative to associated incentive effects?
To what extent can the over-all influence of behavioural responses to policy change be approximated by labour supply responses alone?
Motivation
Provide a sense of the overall importance of behavioural responses when analysing policy change
Indicate the practical importance of analytically convenient modelling assumptions
11/19/2018

4. Aims and motivation Two policy counterfactuals:
A rise in all rates of income tax by 10 percentage points
A fall in the value of state retirement benefits of 20 per cent
Both policies are applied from 2016, and are unanticipated
Three behavioural alternatives
The functional description of behaviour assumed under a base policy environment remains unaltered under the counterfactual
Labour supply responds optimally to the policy counterfactual, but savings behaviour remains functionally unchanged
Savings and labour supply both respond optimally to the policy counterfactual

5. Method – POLICY COUNTERFACTUALS
Projecting effects of policy counterfactuals
Generate base policy context
Policy as observed in 2016, projected to 2070
Generate counterfactual policy projection
Start with base, and project forward from 2016, using precisely the same inputs as the base projection, except the considered change in policy
Identify effects of policy by subtracting (1) from (2)
SIDD/LINDA dynamic stochastic microsimulation model (DSMM)
available from

6. Method – BEHAVIOURAL ASSUMPTIONS
- year of birth
- age
- marital status*
- number children*
- age children*
- student status*
- education*
- wage potential*
- liquid wealth
- pension wealth
- pension access
- immigration*
- emigration*
-survival*
Projects evolving population cross-section through time in two stages
solve lifetime decision problem
Based on dynamic programming methods
project population based on solutions obtained in stage (1)
Monte Carlo methods
Forecasting prediction errors
Endogenous vs exogenous considerations and model limitations
Importance of differencing
Randomness of finite samples (bootstrapped standard errors)
* denotes characteristics that evolve stochastically

7. Method – BEHAVIOURAL ASSUMPTIONShT wT
Savings and labour supply respond optimally to policy counterfactual
Default simulation approach for the model
Functional description of behaviour remains unchanged, relative to the base policy environment
Use grids from base for projecting counterfactual
Labour supply responds optimally to policy counterfactual, but savings behaviour remains functionally unchanged
Required a little re-programming of basic model
video walkthrough of changes made available on website
hT-1
wT-1
age
hT-2
wT-2
200 million h1 w1

8. Results – net government budget
Projected effects of policy counterfactuals on net government budget, by year and behavioural scenario (£2016 billions)
Behavioural alternatives have substantive impact on projected effects of change in income tax rates, but little impact on effects of change in benefits values
Highlights the context specific nature of bearing of behaviour, and the value of tailoring the characteristics of the assumed microsimulation model to the subject of interest (and vice versa)
Full explanation of difference between considered policy counterfactuals in the paper. Focus here on explaining effects of changes in income tax rates.
Standard errors are small

9. Results – Assets
Projected effects of policy counterfactuals on total household assets, by year and behavioural scenario (£2016 billions)
Reverse of previous slide in terms of behavioural sensitivity…

10. Results
Extent to which revenue effects of 10% rise in income taxes projected under the non-behavioural scenario overstate the full behavioural scenario
Effects generated by differences in projections for income tax revenues – see next slide

11. Results
Extent to which revenue effects of 10% rise in income taxes projected under the non-behavioural scenario overstate the full behavioural scenario
Importance of income taxes points toward labour supply as a key explanation for differences in generated budgetary effects between simulations, suggesting that allowing for endogenous employment may go a long way toward adjusting for the difference

12. Results
Extent to which revenue effects of 10% rise in income taxes projected under the alternative behavioural scenarios overstate the full behavioural scenario

13. Results
Extent to which revenue effects of 10% rise in income taxes projected under the alternative behavioural scenarios overstate the full behavioural scenario
Allowing for labour supply responses to policy counterfactuals does go a long way toward capturing the budgetary response generated under the full-behavioural scenario, but the gap widens as the time horizon is increased

14. Behaviour
Projected behavioural responses to 10% rise in income tax rates for families with at least one member born 1981 to 1990, by age of cohort member
Labour responses (hours per adult per week)
Consumption responses (£ per week)
Green intermediate to blue and red – importance of wealth accumulation over time

15. conclusions
Alternative behavioural assumptions can substantively alter projected effects of policy counterfactuals
Results highlight the importance of selecting an analytical tool that is tailored to the subject of interest (and vice versa)
In terms of the net government budget, considered behavioural alternatives have a more prominent impact when considering a change in income tax rates, than when considering a change in the value of retirement benefits
In terms of projected measures of household wealth, in contrast the reverse is true.

16. conclusions
A share of the over-all influence of behavioural responses to policy change can be approximated by labour supply responses alone
Extent depends upon both the subject and the projected horizon
The adequacy of labour supply responses considered in isolation tends to deteriorate with the projected time horizon, due to coincident fluctuations in accrued wealth.
Labour supply responses have limited capacity to capture fluctuations in wealth
Best practice should compare projections from alternative analytical approaches to ensure that stylisations associated with a preferred methodology do not drive centrally important results

17. Method – BEHAVIOURAL ASSUMPTIONShT wT
hT-1
wT-1
age
hT-2
wT-2
200 million h1 w1