The Department for Work and Pensions (DWP) is the UK’s largest public service department, responsible for welfare, pensions and childcare policy. A main challenge for the DWP lies in the uncertainty linked to policy implementation, for example, understanding the impact on poverty and income distribution of a particular policy choice.

Microsimulations to make fairer and more informed decisions

Under the leadership of economist Ashwin Kumar, the DWP’s Model Development Unit developed an intricate static microsimulation model, the Policy Simulation Model, which aimed to model the effects of policy changes on poverty across Great Britain. The tool combines a wide range of survey data, administrative data, defined tax and benefit rules with a series of assumptions, for example, on household structures, numbers of hours worked or minimum wage.

The simulation that is produced as a result enables the DWP to better understand the impact - both financial and non-financial - of proposed changes to the tax and benefit system on representative population samples. Outputs, such as detailed predictions of the demographics on ‘winners’ and ‘losers’, or the expected public costs or saving of a policy change can then feed into the overall DWP strategy.

Applications and limitations

The Policy Simulation Model was notably used to support and inform the introduction of the Universal Credit in 2013, replacing a range of preexisting means-based tax credits and benefits. Taking projected demographic and economic changes into account, the model was used to compare the expected benefits of introducing universal credit on work incentives and poverty levels, against the benefits of keeping the status quo.

The advantages of using simulation as a decision-making support tool are threefold. The computing power of the model allows for a more efficient breakdown of forecasts and an increased ability to observe impact of decisions over the whole benefits system. The model also enables its users to test different combinations of policies, conduct robust scenario analyses, and understand which choices might be more efficient.

Finally, decisions related to a change in the benefits system are likely to have a huge impact on the most vulnerable groups across the system, and the opportunity to experiment with more sensitive policies in a ‘safe’ environment is crucial.

However, the forecasts delivered by the Policy Simulation Model will never be identical to a policy implementation; firstly, because there will always be unexpected and therefore un-modelled events, and secondly, because the model has to make certain assumptions that may not stand up in the real world.

But models such as the Policy Simulation Model are important for teaching policymakers both the importance of experimenting, and of understanding how one change in a complex system can have knock-on effects in many other places in that system.