andersen

The Public Policy Laboratory: Taking a fresh approach to policy implementation and securing policy outcomes

SUMMARY.

1. This paper proposes the use of "laboratories for public policy".

The laboratory concept offers an alternative implementation approach to proceeding largely on supposition, desk analysis and the courage of policy-makers' convictions. The laboratory concept builds on the practices of piloting, and of running 'model offices', that have become established in the implementation of policy, but it allies these practices to rigorous research methods from the private sector. And it applies them earlier in the policy-making cycle - after the principles and objectives of policy are laid down, but before the details become set.

Each laboratory for public policy would be an office, appropriately equipped and staffed, in which the business processing for its selected customers (both citizens and employers) would be performed, whilst arrangements for the general population were unaltered.

Laboratories' remit would be to devise ways to make policy in complex fields successful and sustainable, whilst avoiding unanticipated issues of the sort that damaged Child Support and the Community Charge. Laboratory research might not predict precisely how many people may decide not to comply with, say, new tax rules or not to take-up training opportunities. But it will be possible to be very clear about the degree of risk that non-compliant or apparently perverse behaviour will occur. Government can then act to mitigate this risk and plan contingencies to deal with its occurrence.

Laboratory activity would help to create a public, media and marketplace momentum for change. Laboratories would determine the answer to possible objections to policy early in the policy lifecycle, and would produce: hard evidence about citizens' behaviour in reaction to policy and about how the policy proposals can deliver the outcomes that are sought; a "shop window" of successful outcomes; and numbers of citizens who have benefited and are willing to advocate the experience to others. Laboratories would promote the "good news", and feed back "bad news" into visible remedial actions before public or media criticism could escalate.

2. In addition, public policy laboratories would re-use existing public assets (eg. administrative systems, accurate data, legislation and powers, staff skills) to prototype the new administrative arrangements, without requiring the commitment of time or resources in a grand projet on a national scale. They would often draw on assets from more than one agency or department, so creating an opportunity and a pressure for collaboration and inter-working in pursuit of policy outcomes.

3. These are bold claims that we substantiate in the body of this paper. Our case is strongest in fields where a prime objective of policy is to influence attitudes and behaviour, and where delivery requires effective inter-working between departments and agencies, both central and local. In these circumstances, government needs to be able to understand, in advance of national implementation, the likely effects of policy provisions better than has usually been possible by conventional analysis and modelling. This is doubly difficult to understand in fields where the existing policy provisions and administrative arrangements are complex - where "the devil lies in the detail". Today it is such fields that are at the heart of the public policy agenda: taxes and benefits; education; welfare, employment and income; health, social services and housing; crime.

4. In such fields, laboratories can deploy reliable research and development methods that are in common use in the private sector. Relevant research and prototyping techniques are well established "best practice" in the marketing and design laboratories of consumer industries. These techniques can successfully be applied to consumer services, including public services.

5. Laboratories can deal well with complicated topics. We are convinced that laboratories will often be the best way to devise and to fine-tune public policy and thereby to secure sustainable success for Government proposals. The laboratory concept enables government to
"think big, implement small, and then 'scale up' fast" - a sound approach to achieving radical change successfully and a major contribution to good government.
* * * * * * * INTRODUCTION.

6. For policy-makers, the "holy grail" is policy that is highly successful in achieving the intended outcomes, is efficient, equitable and without undesirable side-effects whilst in operation, and is sustainable because it secures widespread compliance and support. It is policy that converts well from the drawing board into public services and good government.

7. Occasionally, such policy may seem easy to devise. But the common experience of policy-makers and administrators has been different - devising such policy is hard. In many important areas of public policy facing Britain today, success depends heavily upon getting right the:

details of policy, so that its provisions do in practice change the prevailing culture and behaviour amongst citizens and organisations by influencing their incentives and expectation
mobilisation of change, building consensus and collaboration in support of the policy and its objectives by demonstrating its equity, effectiveness, and efficiency
execution of change, turning the substantial legacy of past public investment (for example, in administrative systems, in information technology and accurate data, and in staff skills) to new purposes - quickly, safely and with excellent value for money.

One example : taxes and benefits
8. We may illustrate with one current example: The integration of tax and benefits has been proposed by numerous politicians, academics and think tanks, in the UK and overseas, over the past 30 years. The present Government has now clearly signalled its intention to make changes to the arrangements for taxes and benefits as they apply to low-income working families, and it is possible that further reform will follow. During the public debate on this policy, we have heard at least nine different objectives advocated for tax-benefit changes, with much argument about their respective merits and likely consequences. During the implementation of past policy change we have seen the risks and costs of change rise whilst the pace and effectiveness fall, when good use is not made of existing assets. And we have seen programmes founder completely when public behaviour turned out not to respond to the provisions of policy as expected.

9. As experienced implementors of the changes needed to deliver policy in both the private and public sectors, Andersen Consulting has researched the reasons why so few past tax-benefit proposals have reached successful implementation, and considered what can be done to break this cycle. Our research has concluded that experience offers some lessons if major change is to be implemented sustainably and to deliver its original objectives:

The objectives must be clearly defined and shared by all involved.
Those involved must have clear expectations about what needs to change in order to implement the policy rapidly, efficiently, and at acceptable risk.
Proposals should be proven to deliver the intended behaviours and outcomes.

10. Looking just at the tax-benefit field, many policy proposals have failed on more than one of these counts. There has seldom been clarity and consensus about objectives. Expectations have often differed about the extent to which the current frameworks, rules, administrative systems, and cross-departmental machinery must change. Proposals have been implemented without it being proven that they will achieve the intended outcomes.

Taxes and benefits are complex
11. Today's separate UK tax, National Insurance and benefit regimes are some of the most complex and intricately constructed elements of the UK public service infrastructure. The delivery organisations and skills, the computer systems, and the volume of data recorded are legacies of long and heavy investment. These assets need to be harnessed for change rather than become sources of inertia.

12. To show the scale of this legacy, consider the situation for one illustrative household - a married couple on low income, with a child under five years old. Many elements of the present tax, National Insurance and benefits regimes would need to be changed in order to implement integration for the tax-benefit affairs of this family. There are 5 benefits involved for our illustrative household. They bring with them at least 4 different workforces totalling over 10,000 skilled staff distributed around the country, who have limited technical knowledge of each other's domains. Tax and National Insurance add many more staff and locations, and 5 further domains of skill. Management and political responsibility is divided between 5 Cabinet ministers, 5 central government agencies, and many local authorities.

13. There are at least 12 Acts of primary legislation (and many associated Regulations) defining the levels of the benefits payable to our illustrative household. Further legislation defines the tax and National Insurance payments due. Each piece of legislation describes the data that is required to be known about our household for tax and benefit purposes. Definitions for interpretation of the same data may vary across this legislation. Most of the data recorded for our illustrative household are stored with that for almost all other UK households in a large number of IT systems - as many as 10 for our illustrative household. However, some data are held on paper, and Northern Ireland households are held in separate systems. The IT systems have differing technologies, many tailor-made and fragile interfaces, and a number of interdependencies related to their history rather than their future.

14. It is a veritable Gordian knot of complex cross-department, inter-agency relations - and just for this one household. This complexity will usually defeat the accurate modelling that is required if one wishes to predict the outcomes of tax benefit policy before it is implemented.

but complexity is the norm
15. Where complexity defeats modelling, the laboratory concept offers an alternative approach to proceeding largely on supposition and the courage of policy-makers' convictions. Similar knots in the machinery of government also complicate many important areas of public policy beyond taxes and benefits:

disability and employment
education, probation and social services
income, housing, and health
transport and local government.

16. Even the groupings cited above are inter-related. Few public policy problems are now understood to be simple or to stand-alone. Solutions to the important public policy challenges increasingly require cross-organisation working - in the formulation, mobilisation and execution of policy. This will involve many parties in central and local government, the voluntary sector, and private sector employers and service providers. Laboratories can deal well with complicated topics. This is a virtue not shared by many policy-making and implementation approaches we have seen. We believe the case for public policy laboratories is already strong, and set to get stronger. The rest of this paper explains that concept and discusses how it can assist Government to devise successful policy.

THE LABORATORY CONCEPT

17. Each laboratory for public policy would be an office, appropriately equipped and staffed, in which the business processing for its selected customers (both citizens and employers) could be performed, whilst arrangements for the general population were unaltered.

18. Piloting and "model offices" have become well-established techniques in several public policy fields during the 1990s. They have been effective at working through day-to-day operational problems to prove that a proposed new business process and system will work, but they have had important limitations. They have not been good at examining public behaviour, and they do not prevent the legacy of existing public service assets (administrative systems, accurate data, and staff skills, for example) being seen as a bar to the more radical policy changes that new processes could support.

19. The laboratory concept takes the best elements of piloting and model offices and adds more, to improve on both. The important differences are that:

A laboratory follows a more structured research process with greater rigour in methodology. It still uses real offices, staff and IT to explore how implementation can best be accomplished, and it allies these to the right research techniques.

A laboratory works with real citizens and employers as participants, and they are entirely free to act as they choose in accordance with the law. A laboratory would not script their behaviour as has been done in model offices before - and which greatly reduces the validity of any conclusions one can draw.

A laboratory would operate earlier in the policy-making cycle - after the principles of policy are laid down, but before the details become set. There would be enough time planned for the conclusions from laboratory research to be acted upon, prior to any national rollout of policy provisions.

20. A laboratory's remit would be to help the Government devise ways in which specific public policy programmes could be advanced. In particular, it would help to determine and to demonstrate how new policies can work - when tested against the Government's objectives and values. Laboratories would have three goals to fulfil this remit:

1. To produce hard evidence about people's behaviour in reaction to policy changes and how the policy proposals can best deliver the economic, financial and social outcomes that are sought by Government - - there are reliable methods to do precisely this, and they are in common use in the private sector.

2. To find ways to exploit the legacies (legislation, processes, skills, culture, IT systems, etc.), and to deal with the obstacles, that may otherwise appear to make the proposals impossible to execute - - a laboratory would deliberately re-use these assets to prototype new arrangements, without requiring the huge, new investment on a national scale. Such prototyping using existing components is a well-established "best practice" in the design laboratories of consumer industries. The concept is equally applicable to the business design of consumer services, including public services.

3. To help address the root causes of any institutional or citizen resistance to policy and to secure public support - - a laboratory could determine and publicly demonstrate the solution to many of the objections and criticisms of policy, early in the policy lifecycle, helping Government to create consensus and build the momentum for change.

LABORATORIES AND RESEARCH ON BEHAVIOUR AND OUTCOMES.

21. In many public policy fields today it is a prime objective of policy to influence people's behaviour in order to help achieve the social and economic outcomes that are sought. Such policy-making depends on judgments about the behaviour effects of policy provisions:

citizens' reactions to their experience under the policy
their perceptions of the benefits and dis-benefits of that experience
their evaluation of the remedies offered for the dis-benefits
and, most important, their bias and intent toward significant behaviours such as co-operation, non-compliance, advocacy to others, and repetition of the experience themselves.

22. Based on desk analysis, these judgments may be wrong. In the extreme cases, non-compliant behaviours can undermine policy, can negate its effects, and can lead to its complete retraction. The Child Support Act and the Community Charge are two "headline" examples. We use the term "non-compliant" to mean all forms of undesired behaviours that were not intended by policy. For example, we include: not claiming entitlements; committing fraud and abuse of the rules; giving obstructive responses to enquiries; raising unwarranted queries, complaints and appeals; and delaying compliance; as well as outright evasion of legal liabilities or responsibilities.

23. Some of these "non-compliances" do not imply wrongful behaviour by citizens, but they each reduce the effectiveness and efficiency of policy if they are not anticipated and prepared for. In less extreme examples, citizens' silence instead of advocacy, or their avoidance of future contact with the administrators of policy (for whatever reasons), can prevent the desired outcomes of that policy being achieved. Income Support entitlement amongst pensioners, and Disability Working Allowance, are examples of policy weakened in this way.

24. Laboratories can identify these behaviour effects of policy amongst citizens in advance. Of particular importance, we argue below that:

a laboratory can give meaningful advance information on public behaviour in reaction to policy, including non-compliant behaviour

small exercises are sufficient for useful results

many possible location(s) for a laboratory can be "representative" of the entire national picture.

Relevant techniques for researching behaviour
25. The use of specific research techniques to explore behavioural issues is well established in many arenas outside public policy. For example, in the worlds of retail, consumer goods and services, and advertising/merchandising, market-research companies now use an increasingly sophisticated range of attitudinal research methods. These are applied to "soft" topics such as customer loyalty, perceptions of customer service and product performance, brand identity and buyer values. The same methods are frequently applied to these topics and others by independent third parties, including consumer organisations, regulators and lobby groups, for example.

26. This provides a clear body of evidence that public attitudes can reliably be assessed, and behaviours and motives examined, using established research techniques. We do not claim that predictions based on such research would always be accurate to the last percentage point - no method of analysis is, and this is not what is required. Laboratory research might not predict precisely how many people may decide not to comply with, say, new tax rules or not to take-up training opportunities. But, using qualitative methods, it will be possible to be very clear about the degree of risk that such non-compliance or apparently perverse behaviour will occur. Government could therefore act to mitigate this risk and to plan contingencies to deal with its actual occurrence. For both Child Support and Community Charge, acting on such a risk assessment (rather than on a precise, but wrong, prediction of numbers not complying) would have been of obvious value but does not appear to have been done.

27. The research techniques we believe are relevant to a laboratory include both qualitative and quantitative methods. Qualitative and quantitative methods are different, but complementary. Selecting the right combination of techniques is an essential part of Research Design, and must be driven by the objectives. We are not aware of a laboratory or pilot yet designed with the objective of researching the behaviour effects of policy, in the UK public sector.

28. There are many cases of pilots exploring behavioural issues, both in the UK and overseas. They are to be found in both the private and public sectors. Examples include:

Water metering trials. These took place from 1990 to 1993 and centred on 50,000 households on the Isle of Wight. The authorities wished to measure public reaction to metering within a discrete area, both in terms of how household consumption changed and whether there was any political outcry.
Betreibskrankenkassen (BKK) pilot. This began in Berlin and Bavaria in 1996, in order to pilot a combined healthcare budget for an integrated network of healthcare providers. One behavioural aspect of the pilot is that quality of care will partially be measured by the number of patients choosing the pilot healthcare provision scheme over the existing model of healthcare provision, which will continue to operate in parallel.
Domestic gas competition pilot. The first pilot for domestic gas competition began in the southwest of England in April 1996, and exposed problems not only with some business processes but also in the behaviour of some companies' sales staff.
Information for lone parents. This project was a pilot for the delivery of information on childcare, job vacancies, benefits entitlement and training. It was conducted through kiosks and PCs in over sixty public locations throughout Cambridge. This allowed testing not only of the technology but also of how people used the service. For example, it became apparent that employers as well as individuals were using the service - a behavioural outcome which had not been originally intended, but which was in this case not undesirable.
Mondex pilot. This trial of smartcards in Swindon from July 1995 allowed the testing of both the technology and how consumers would use it. It discovered that the cards have been most popularly used by consumers in supermarkets, fast food outlets, newsagents, tobacconists, car parks and pubs (ie. outlets where consumer behaviour is usually to pay cash, not with plastic). Mondex are now building their strategy for a nationwide launch and 30 organisations have bought franchises.

Size of the exercise
29. A laboratory aims to ensure that routine outcomes and general trends, ie. situations occurring in many cases, will occur as intended by policy. The more prevalent the outcome or behaviour trend, the smaller the sample needed to disclose it.

30. It is a common objection mounted to research exercises that the number of participants would have to be very large in order to be useful in researching behaviour. This is not so. As illustrations, we would consider 20-30 participants as the effective minimum necessary for a useful application of qualitative methods to a simple research topic (but most public policy issues would need more) and 1,000-2,000 as the usual maximum required for quantitative methods.

31. The key to obtaining useful qualitative and quantitative information lies in expert Research Design and conduct of the exercise - choosing the right techniques and performing them correctly on the right size and composition of sample to be representative.

"Representativeness"
32. What composition of sample is 'representative' must be judged against the aims of the exercise. We consider the generic research objectives of a public policy laboratory would be to evaluate:

What day-to-day burdens (or conveniences) will the detailed policy processes place on people ?
Are the business processes and interfaces workable for the participating staff and organisations ?
How can one re-use existing assets to make this policy change happen quickly and effectively ?
What will be the actual effects of the new arrangements - amongst citizens and organisations - and do they match the Government's objectives ?

33. We believe that in many instances government could establish public policy laboratories that remain representative, and at the same time:

citizens' and employers participation could be voluntary (avoiding the need for legislation)
only one or a few locations need be chosen (keeping the exercise manageable).

Voluntary participation versus compulsion
34. Whether a public policy laboratory could operate with voluntary participation is a critical issue, because compulsion may require new legislation. Most policies create apparent "winners" and "losers". In some cases, the populations of winners and losers are clearly separate, as when taxes are levied on one segment of the income distribution to finance programmes of benefit to another segment. However, for some policies the populations of winners and losers may not be so separate, or there may be few apparent winners at all. There may also be some people who perhaps expected to be winners but who are not, and other people who do not believe, or do not care, that they will be.

35. A voluntary laboratory may be expected to self-select participants from amongst the expectant winners only, and to omit losers, the apathetic, and the socially excluded. On the other hand, a compulsory exercise may be expected to raise issues of differential treatment of participants in comparison to the general population, perhaps leading to accusations of inequity. Are these problems?

36. For the first three research objectives, there is no reason why a laboratory need be very representative of the general population's characteristics. Instead, the way in which it operates should be as inclusive as possible of the different situations which might be encountered, by both citizens and staff, and one needs to ensure that the sample size and structure is sufficient to reveal most such situations. Voluntary participation by winners only can achieve this and can therefore achieve the first three aims of a laboratory. It may sometimes also be possible to offer carefully designed incentives or guarantees to participants, to increase the rate of volunteering without distorting the results.

37. For the fourth objective, self-exclusion of those people who expect to be losers or do not care is not completely unhelpful. The extent of self-exclusion indicates citizens' immediate perceptions of the policy, and some of these perceptions may usefully be followed up by attitudinal research amongst non-participants.

38. Despite this, self-exclusion does reduce the scope for assessment of low uptake and of evasive or obstructive behaviours. So for policies where non-compliance was a priori a concern we would advocate following a first, voluntary stage with a second, compulsory stage. We would advocate this in preference to starting immediately with compulsion because a lot can be achieved to create a durable consensus in support of the policy by working first with the volunteers to prove that there are beneficial outcomes for them. This reflects the fact that the laboratory is not "neutral" - its remit is to help devise successful policy and to build momentum toward the desired conclusions, not dissipate it.

Compulsion
39. Concern about compelling people to participate in a laboratory follows from the different treatment that laboratory participants are given in comparison to the general population. This is a particular criticism in those policy fields where Britain is perceived to have a unitary system of policy endowing equal treatment on all.

40. Taxation is perhaps the best example of this, with social security close behind - in both fields "winning" and "losing" tends to be expressed solely and coldly in cash, diluted only slightly by the other outcomes of the tax and benefit systems (eg. incentives, stigma). We believe that in most other public policy fields there is not a unitary UK system. Even in fields where the relevant legislation is unitary, people know that there can be great diversity of treatment in practice - health, housing, education, probation, social services - particularly where the day-to-day authorities are local not national, and also where the outcomes of policy are not measured predominantly in cash.

41. In social security, the practical day-to-day system is perhaps a bit less unitary than citizens tend to believe. The Ombudsman has twice advocated the need to pilot social security arrangements ahead of policy implementation, recommendations that imply recognition of the need to deal with the issue of differential treatment during piloting. The House of Commons Social Security Select Committee has also commented on the benefits from greater flexibility and opportunity for experimentation.

42. Any such differential treatment would not be one-sided. Those differences in treatment that create winners are the up side, helping fuel demand for policy implementation more widely. In a compulsory laboratory, we would argue that these beneficial differences are therefore advantages, not criticisms. Differences in treatment that create losers are the downside. To address them we would suggest it is possible to underwrite cash losses in some, and perhaps all, cases of those compelled to participate.

Location
43. For many public policy laboratories, we believe that the most appropriate location would be a single "travel-to-work" area contained within just one or two local authorities. To obtain an efficient density of population for a laboratory would rule out rural areas. To allow a relatively simple boundary to be drawn round a small participating population would rule out conurbations. This elimination suggests provincial towns would be likely locations. But would this lead to results which were unlikely to be reproduced nationally when conurbations will be the main environment in which much public policy must operate ?

44. We do not see this as a significant problem for the first three objectives of a laboratory (see above). Although conurbations do have some different characteristics from provincial towns, many of the differences would not be relevant to these three objectives. For the fourth objective - What will be the actual effects of the new arrangements (amongst citizens and organisations) and do they match the Government's objectives ? - one would need to be more careful.

45. In conurbations there may well be a much greater density of people in the demographic and socio-economic groups at whom policy is addressed. To remain representative, one would collect information on the relevant characteristics of those volunteering to participate. One would use quotas (for inclusion in the sample, rather than for participation in the laboratory) to structure the sample in such a way as to take account of most or all situations relevant to the policy under examination. This might mean the sample should omit some participants from some groups, and seek additional participants from other groups of people.

46. Other characteristics that are more common in the inner city environment (such as social exclusion) may create a "micro-culture" that influences public attitudes and behaviours to be different to those in provincial towns. This greater density may create a greater likelihood of mass swings in attitudes that one would not see when looking in isolation at a provincial town. If this is a priori a significant concern, we believe it is appropriate to conduct baseline research in different areas before commencing the laboratory itself. This baseline would establish whether there were different attitudes evident in local people's responses to hypothetical policies and provisions. If this research revealed differences, we would expect to follow a first, provincial stage of a laboratory with a secondary stage conducted in another area. Once again, our preference follows from the non-neutrality of the laboratory - its remit is to help devise implementable policy. It would tackle locations in the sequence most likely to lead to positive conclusions.

LABORATORIES AND THE RE-USE OF ASSETS

47. Public policy laboratories would develop experimental arrangements for the administration of new policy and would establish the extent to which the existing legislation, organisations and infrastructures can be exploited or must be replaced. Where replacement is necessary, a laboratory would explore the migration paths available.

48. It would perform these roles by deliberately re-using the existing assets, resembling the way in which makers of consumer goods first build trial- or prototype- products from the componentry of their existing models. This is intended to minimise obsolescence amongst those components and to reduce the investment risk from that of the grand projet which aims to replace all existing components with new.

49. A laboratory would be equipped with computer systems and data from the existing public services, including those available under public-private partnership. Personnel who are drawn from the best of each participating department and agency would staff it. It would usually operate under legislation, discretionary powers, and the delegated authorities and management freedoms that already exist - new legislation is neither attractive, nor always necessary.

50. The various staff would be appropriately co-located to administer the policy for participating citizens. The laboratory would experiment with cross-skilling of those staff, with job re-design, and with organisation and human performance factors, including the appropriate division of administrative processes between employers, citizens and public servants. The laboratory would also use the co-location of work to identify opportunities for process re-engineering. And it would make changes to computer systems as necessary to support the limited scale of laboratory operations rather than at "industrial strength".

51. The laboratory would create a business and operational model for a new policy regime that works, and develop a good understanding of the major investment and obsolescence issues of implementation.

LABORATORIES AND THE MOMENTUM FOR CHANGE

52. Once up and running, each public policy laboratory could act as a showcase for the operation of policy proposals and for the outcomes they will produce. Laboratories would provide a way to demonstrate the benefits flowing from policy, to identify and emphasise the winners, and to show how any unintended adverse effects on losers can be ameliorated. They would promote this good news and feed back any bad news into immediate remedial actions.

53. The small scale of a laboratory would help make this promotional activity more controllable. It would be possible to secure the engagement of relevant local bodies - media, voluntary and community organisations, employers - whose contribution and buy-in could then be leveraged into advocacy to others.

54. The results of this activity would be to help create a public, media and marketplace momentum for change. Laboratories would produce: hard facts that can be cited in support of policy; a shop window of successful outcomes; citizens who have benefited and are willing to advocate the experience to others.

AN ILLUSTRATION : THE TAX - BENEFIT LABORATORY

55. Andersen Consulting has recently proposed, to the House of Commons Social Security Select Committee, the creation of a tax-benefit laboratory established through public-private partnership . We did so because:

the formulation of tax-benefit policy must deal with issues of incentives and behaviour that are not easy to predict by conventional analysis and modelling

the mobilisation of change to taxes and benefits must overcome an absence of public consensus about objectives and means

the execution of that change must find a way to turn the legacies from the very large, recent investment in the social security and tax administrations to new purposes.

56. These are precisely the things we believe a public policy laboratory can accomplish. The difficult issues in many proposed tax-benefit schemes arise from their technical complexity and from their anticipated effects on public behaviour. Technical issues include whether one can avoid joint taxation of couples, and whether periods of tax and benefit assessment can be integrated. Behaviour issues include whether the policy provisions will create the expected incentives to work, how powerful these incentives will be, and what the reaction of citizens will be if they must keep more, or more complicated, records and provide this information to their employers.

57. When addressing these issues, the great scale and complexity of the current tax and benefit regimes make it difficult to analyse what outcomes, both intended and perverse, may result from specific proposals. This scale and complexity is also presented as a challenge to the execution of change. In the absence of a laboratory, we see the great majority of the debate being conducted by argument that is necessarily based on advocacy and desk analysis, not facts.

The laboratory
58. A small, clearly defined population of citizens and employers is appropriate to a tax-benefit laboratory. This could be achieved by focusing the office on a particular geographic catchment - perhaps a single travel-to-work area within one local authority - and selecting further from the population within that. Depending on political will, people's participation in the laboratory could be left voluntary, or be made compulsory. Since compulsion is very likely to require legislation, we do not find it attractive in the first instance.

59. Nor do we believe it is necessary in this case. In what we anticipate will be the Government's 1998 Budget proposals on tax-benefit policy, we expect that the populations of winners and losers will be separate. The intended winners, for example receiving a Working Family Tax Credit (WFTC), will be low-income working families. The intended losers, paying higher income tax amounts and/or National Insurance Contributions, will come from further up the earned income distribution.

60. Given the Government's stated objective to create incentives towards work, we assume that it is the actual effect (both intended and unintended) of policy on the incentives and behaviours amongst the winners that merits investigation in this case. So a laboratory need be representative only of the winners. It is unlikely that their voluntary participation will be a problem to obtain.

61. In a voluntary laboratory, those citizens invited would elect to participate because they perceive it offering a "one way bet" on improved service and improved opportunity. In exchange, they would formally agree to accept any consequent changes in the effective cashflow of their benefits and after-tax wages, or diversion of cash into non-cash benefits. Our belief is that this can be done without requiring legislation, for example by using existing appointee rules. Under compulsion, participation would be required of all citizens fitting the candidate criteria. In that event, non-compliance would need to result in sanctions, probably requiring new legislation. We do not envisage this would be necessary for the Government's anticipated 1998 Budget proposals on taxes and benefits.

Proving the policy outcomes
62. The tax-benefit laboratory would simulate the effects of proposed changes to the current features of tax and benefit regimes (eg. changes to thresholds, tapers, disregards and allowances, non-cash entitlements, and passporting). It would thereby gather evidence of the likely policy outcomes of integrating taxes and benefits in the particular way(s) proposed.

63. It would work with the volunteering employers and citizens to simulate the effect on them of proposed changes to the composite tax-benefit regime, and to collect empirical data on the transition from welfare to work. For cash elements, this simulation can be done by employers making additional payments or deductions in citizens' wage/benefit stream. For non-cash elements (eg. childcare, special needs transport, employability training, subsidised public transport), simulation can be done by independent funding of these services, accompanied by deductions from wages and/or benefits of the equivalent cash to the proposed charge for the service, if any.

64. Empirical data could be collected on benefit entitlements, benefits actually claimed, taxes paid, net cashflows, non-cash services received and net charges levied, and also on citizens' perception of the regime and their measurable behaviours under it. (For more detail, see Laboratories and Behaviour below.) To avoid the need for immediate legislation, any advances and additional payments could be made ex gratia, from independent funds established by the participating agencies and/or the private sector partners.

65. The tax-benefit laboratory would operate within the existing framework of legislation both primary and, initially at least, secondary . However, it would be able to exploit existing discretionary powers in tax and benefit decision-making and would be encouraged to amend those official forms and publications not prescribed in law. This approach would enable:

wide ranging experimentation with information provision and collection processes
preservation of a unitary national regime of tax and benefit law during the experiment
demonstration of the extent to which granting discretionary powers can minimise the need for comprehensive re-writing of legislation.

66. The results of this activity would be the validation of tax-benefit proposals for working families against the Government's policy objectives. Does integration have any unintended or perverse micro-economic or social policy side effects for families? Are the required tax returns and calculation really too complex for typical low-income households and small businesses to cope with? Will married and cohabiting couples co-operate with a household-based approach to income tax assessment? Which features of the regime actually determine the desired outcome and which are ineffective?

Exploiting existing assets
67. The tax-benefit laboratory would develop trial arrangements for the consolidated processing of a citizen's tax and benefit information, including the accumulation of all relevant skills in one tax-benefit workforce, and the amalgamation of tax and benefit records in a "citizen's account".

68. Personnel who are drawn from the best of each participating agency and co-located to process the tax and benefit affairs of the participating citizens would staff the office. The office would use the co-location of this work to identify opportunities for process re-engineering to create:

common processing steps (such as simultaneous processing of WFTC and Housing Benefit/Council Tax Benefit)
data commonality, and convergence between computer systems and databases
opportunities for unsolicited processing (such as automatic determination of eligibility for WFTC based on PAYE and JobSeekers Allowance data)
new opportunities for public revenue generation.

69. We believe there are perhaps 25 key staff currently involved in the tax and benefit affairs of a typical individual citizen. Several of these roles are substantially clerical and could easily be combined in far fewer different people than today. Other roles are considerative or technical, based on deep skills or experience, and it may be more difficult to develop all the skills in one individual. Some roles deal primarily with employers and others with citizens. Both tax and benefit work involves front- and back- office processes that are not all currently distinguished. The office would experiment with cross-skilling, job re-design, and organisation and human performance factors, including the appropriate division of advice, verification, eligibility and payment/collection processing between employers, citizens and the office staff.

70. The results of this activity would be a business and organisation model for an integrated tax-benefit regime that works, and an understanding of the major investment and obsolescence issues of implementation. How much IT investment is needed? How quickly can staff be re-skilled and deep change achieved? What propositions could be made for public-private partnership to deliver tax-benefit integration in practice?

Winning support
71. The tax-benefit laboratory would prototype how an integrated, modern tax-benefit regime would impact upon citizens and employers. The office would therefore offer alternative methods of contact and information provision to citizens and employers. These might include:

telephone call-centre handling of all tax and benefit advice and queries
(secure Internet or) Electronic Data Interchange handling of PAYE transactions with employers
secure Internet and kiosk enquiries on each citizen's tax-benefit data
counter- and tele- services for fast-track processing of adjustments, refunds, grants, etc.
"bulletin board" IT applications for job placement, child care, etc.

72. It would work with the volunteering employers and citizens to assess their reaction to the proposed changes to the tax-benefit regime, and to collect empirical data on changes to burdens and compliance requirements. This would enable presentational, "intrusion" and "burden on business" issues to be examined.

73. The laboratory would also provide a showcase for the operation of the policy and for the outcomes it produces. It would provide a way to demonstrate the benefits flowing from the policy, to identify and emphasise the winners, and to show how any unintended adverse effects on losers can be ameliorated. But would not be neutral in a scientific sense - the laboratory's remit is to devise ways to make policy successful. It would therefore promote the good news and feedback the bad news into immediate remedial actions. And it would tackle cases and issues in a sequence likely to build momentum rather than dissipate it.

Public-private partnership
74. Those public agencies which deal with individual citizens' tax and benefit affairs under the current regime would need to participate in the laboratory - Benefits Agency, Contributions Agency, Inland Revenue, Child Support, Local Authority. The basis for funding the administrative costs of the office could include capitation on these agencies in relation to the workload they each transfer to it, direct funding by Treasury, and ex gratia resource contributions by private sector partners. Any private sector partners would be expected to "bring something to the party" in terms of assets that may be used to constitute the laboratory. Such assets would include:

the ability to deliver existing tax and benefit IT applications to the laboratory
personal networks with stakeholders in the laboratory's catchment zone
credibility and trust within the selected citizen population
potential components of future tax-benefit IT applications
finance for policy experiments (eg. to fund cash flow effects).

75. Some form of participation would also need to be solicited from stakeholders in the current tax and benefit regimes that are outside Government:

representative employers (as likely co-administrators of any new regime, and for experimentation with policy features and outcomes)
the voluntary sector (eg. debt advisors, child carers, citizens advice bureaux)
the training and education sector (eg. TECs).

Conclusion
76. There have been numerous proposals for tax and benefit integration, matched by numerous objections to them. Most proposals and criticisms have been based solely on academic argument or on advocacy, and we think it is time for a fresh, more practical approach.

77. We propose the public policy laboratory concept as a means of ensuring that objectives are met by the proposals that Government advances; that the many obstacles are assessed and overcome; and that useful opportunities to achieve more rapid and secure progress are exploited. The tax-benefit example illustrates the concept and its potential to exorcise the devil that lies in the detail. From Andersen Consulting's perspective as professional implementors, the laboratory concept enables government to "think big, implement small, and then 'scale up' fast" - a sound approach to achieving major changes successfully.

About Andersen Consulting
Andersen Consulting is a leading management and technology consulting firm, with specialist skills in the implementation of large-scale change. In the public sector, we have significant experience of the operation of tax and social welfare regimes world-wide, and with justice and police services. We have worked in health and education.

Andersen Consulting has worked very closely with the DSS Contributions Agency, Benefits Agency, Child Support Agency, War Pensions Agency and IT Services Agency over many years. This work has included the re-development of the Contributions Agency's National Insurance Recording System and its construction and operation under the first Private Finance Initiative (PFI) information technology contract. We have assisted in the development of many of the major social security benefits systems, in conjunction with the Benefits Agency, War Pensions Agency and IT Services Agency. We have worked on projects to improve the delivery of public services in disability, and to implement new policy for unemployment and incapacity. We have also worked with Child Support.

Our Public Policy Group uses this practical experience, allied to formal research methods, to help policy-makers close the loop between operational success and policy-making.