Stephen Muers is Head of Strategy and Market Development at Big Society Capital. This blog post is based on his time as one of IPR's Visiting Policy Fellows while in his previous role as Director of Criminal Justice Policy at the Ministry of Justice.
Accountability is fundamental to democracy. Holding decision-makers to account for what they do and the impact they have can be seen both as a good in itself and a way of aligning their choices with the interests of the public at large. So effective democracy needs effective accountability, defined here as a system that holds decision-makers to account for things they control in a way that is meaningful and legitimate in the eyes of the public, and which is likely to promote desired outcomes. In turn, therefore, accountability needs to be based on an understanding of what different decision-makers can and should be doing, to fit with public expectations and to promote effective outcomes. Without such an understanding there is a risk that the accountability framework creates the wrong incentives and promotes neither legitimacy nor the right decisions.
What should decision-makers be doing?
In a previous piece I argued that policy outcomes are heavily influenced by culture and value systems, that governments are part of the prevailing culture and, crucially, that they can also affect it. Therefore a critical role of decision-makers is to embody and shape a culture that supports the outcomes they (and in a democracy those who elected them) desire.
One of the most important ways in which culture and values shape policy outcomes is through the individual choices and decisions made every day by the people responsible for implementation: teachers, social workers, employment advisers, police officers and so on. They interpret and act on policy according to their values and the values embodied by the organisations they work in. As I argued in another previous piece these front-line decisions create constant mutation and evolution in what policy means on the ground. As with evolution in the natural world, the resulting pattern is one of periods of stability interspersed with large and often unpredictable shifts. Decision-makers need to recognise this unpredictable dynamic of front-line evolution, and use their position to shape a culture that supports positive experimentation and learning.
If this understanding of how policy works is correct, then central decision-makers should be focusing on shaping a culture that promotes desired outcomes, and that supports front-line decision-makers in a process of learning that leads towards those same outcomes.
What are decision-makers held to account for?
Does our current system of accountability promote such behaviour? In one respect I believe that it does. The central form of accountability in a democracy, the way citizens choose to cast their votes, does reflect a view of political leadership that is centred on values and system oversight. There is evidence that views of a candidate’s cultural values and overall competence in overseeing a system are important in determining voter choice. Put simply, you vote for someone because you think they share your values and will promote them, and you think they are going to be OK at running a government that does so.
Some models of politics assume that people vote for the candidate who espouses the policies they want. However if the argument above about how policy feeds through front-line evolution into practice is correct, it would make little sense to vote on the basis of policy specifics. They will end up being reshaped and mutated by the time they reach individual voters. It is therefore entirely rational for voters to invest little in the highly costly effort of understanding detailed policy propositions (which will change through implementation) and instead form a view of whether a candidate or party has values congruent with their own. There is evidence that fundamental values (e.g. around openness, security and tradition) are significant drivers of voter choice.
Policy proposals therefore become a way of signalling what values a leader will embody and promote, rather than firm statements of what they will deliver. Politicians have always understood this: policies are announced to send a signal and create a narrative about what kind of person or party you are. A good recent example is Donald Trump’s infamous promise to build a border wall and get Mexico to pay for it. According to one poll around the time of his inauguration only 14% of Americans believed he would actually build a wall paid for by Mexico. But the announcement was an extremely strong signal of a value system (against immigration and cultural change) and an attitude (aggressiveness to other countries and a willingness to break the perceived “rules of the game”). It is those values and attitudes that people are choosing and against which they are holding people to account. In doing so they are showing a genuine understanding of what really matters for senior decision-makers.
And what are they not held to account for?
Therefore there is a fit between what it makes sense for leaders to be accountable for, given how we know policy actually feeds through into practice, and what voters use as the basis for deciding how to cast their ballots. So what is the problem?
There would be no accountability problem if political leaders did indeed devote their efforts to affecting values, culture and the overall properties of the system within which they sit: voters are good at holding them to account for that. The problem arises in that in practice they do a lot of other things. Political leaders devote considerable time and effort to designing and implementing detailed policy changes. Democratic accountability for such changes is weak. This is for two reasons.
First, a policy change can be completely disastrous in its own terms without seriously impinging on the welfare of individual voters. Very large sums of money lost to the public purse are hardly noticeable to individuals, especially if such losses are from future potential value rather than current income. A good example would be under-valuation when privatising an asset: no voter feels an immediate loss even if they are in fact worse off because their share of a valuable asset has been given to someone else. But this is true of any large policy failure that wastes money, makes a large service incrementally worse or damages the environment in lasting but not immediately apparent ways.
Second, even if voters are aware of a policy (usually not the case) and are affected by it, it is unlikely to change how they vote. In fact there is evidence that the causation often runs the other way: how someone is inclined to vote affects their understanding of what a policy has achieved. Whether or not someone is aware of a policy and what they believe its effects to be are influenced by their political starting point. This is down to confirmation bias: we interpret information in line with our starting positions. A recent piece of research showed that people’s ability to interpret statistics correctly is dramatically worse when the same statistics are used to describe a divisive political topic (immigration) rather than a neutral one (effectiveness of a skin cream). When political control changes after an election, partisan perceptions of other events changes dramatically. To use another contemporary US example, there was an 82% net positive swing among Republican voters in perceptions of how the US economy was doing in six months Trump’s election as president, at a time when objective economic indicators were fairly stable. So if we believe a political leader is acting in line with our values, that shapes how we see policy and we will tend to register information that implies they are being successful. This dynamic gives politicians considerable leeway to implement policy that is damaging as long as a majority believe they have the right values.
In an attempt to remedy this weakness, we have created a structure of accountability intended to expose policy failure and thereby create incentives for politicians to do the right thing. In the UK this includes the National Audit Office and Select Committee scrutiny. The media also performs an important function in exposing policy failure. Part of the thesis behind such structures is that the appearance of competence is vital to politicians and so methods that expose the opposite will create a strong drive towards successful policymaking.
These structures are, however, unlikely to succeed. Media coverage of a damning NAO report on a multi-billion pound policy will still go unnoticed by the vast majority of the population. And there is no evidence (or even a very plausible theoretical case) for arguing that such a report is likely to contribute to any significant change in party perception and voting behaviour. Confirmation bias is important here too: even if a hypothetical damning audit of a policy became widespread news, people would interpret that news according to their existing cultural framing of the political situation and use it to confirm their existing biases about what policies are or aren’t desirable. It is much more psychologically plausible to believe that a report is biased or wrong than to change your view of a policy and political leader with whom you identify in a deep-seated cultural sense.
Audit and scrutiny take a long time, because of the understandable desire for thoroughness and rigour. This creates two further difficulties. The first is that, because ministers and civil servants – especially in the UK system – move around frequently and fast, by the time a major project or policy is evaluated, those needing to explain it are probably not those responsible for implementation. Again, this means that the incentive to make good policy created by this part of the accountability system is limited.
The second is a more fundamental point. I argued previously that the evolutionary nature of policy means that rapid feedback is critical. It is important to know immediately whether a deliberate or inadvertent change has started to make a difference, allowing the front-line policy implementers to adjust accordingly. The clearer and faster the feedback, the more likely it is that people will learn and iterate towards improvement. In technology projects it has become the norm to use “agile” techniques: build something small, test it with users, learn fast and make repeated small changes. Such techniques have evolved as a way of coping with uncertainty about how people will behave in the face of change and the fact that requirements and goals shift as we find out more about what the front-line users actually want. Such uncertainty and changing goals are a strong feature of policy implementation and so these techniques, and the rapid feedback on which they depend, could have major benefits. However the norm in the public policy field is one of long-term detailed studies that aim to assess, retrospectively, the impact of a whole programme against its stated objectives once enough time has passed to measure progress against them.
What should we do differently?
In an ideal world, political and administrative leaders would devote their efforts to the kind of activity that they are able to carry out effectively and for which we have accountability mechanisms that work: embodying and shaping values and setting the overall parameters for how policy systems work. Alongside this, we would develop rapid feedback mechanisms that enable empowered front-line policy implementers to innovate and iterate within those overall system parameters.
This is, of course, an unrealistic ideal. Political leaders are not going to deny themselves the possibility of pushing through large scale detailed policy change, despite the evidence that complexity and uncertainty means the results will almost certainly not be as promised. They may do this because, contrary to that evidence, they believe that they will be able to deliver the specific promised results. However there is also a more sophisticated view available: a leader can believe that although uncertainty means that exact results are unpredictable, launching a large systemic change is one tool for beginning the process of shifting culture in a policy system. Therefore the real goals of the policy are not specific changes in some output metrics but a much longer-term, more fundamental and hard-to-measure shift in attitudes, values and the way people behave.
One good example of a large, detailed change programme with cultural objectives was the privatisation programme in the UK in the 1980s and 1990s. There were plenty of technocratic claims and counter-claims about the merits of access to private capital and management, and whether the receipts realised in the sales were appropriate. But the bigger-picture goal, explicit in some of the government communications at the time, was to change culture both within the companies themselves and in society more widely by spreading share ownership and a sense of investment in the capitalist system.
So if one of the aims of major policies should be, and often is, cultural change, what do we need to do to make accountability more effective? For a start, accountability and audit could look at the impact policy has on culture, rather than a pure financial and service quality focus. To take the privatisation example, did attitudes to private capital and entrepreneurship change more among people who received shares than in the general population? The tools for this kind of evaluation exist: attitudes and values can be measured and compared over time (e.g. through the British Social Attitudes Survey or World Values Survey), and specific behaviours that are symptomatic of underlying values can also be identified. Of course it is hard to attribute cultural change to a specific intervention, but attribution difficulties also apply to many of the effects (e.g. behavioural or economic) that policy evaluation is usually concerned with. This is not a reason to put culture to one side.
Such an approach would make the issue identified above about the time that scrutiny takes even worse: it only makes sense to think about culture change over a long period. Also, just because an audit report looks at cultural change won’t give it any additional traction with voters in determining how to exercise their powers of democratic accountability. However if the real aim of a policy is (or should be) culture change, evaluation done in those terms is at least more honest and more likely to lead to valuable learning.
As well as shifting assessment and audit towards culture, we could shift it towards the sort of front-line feedback that agile and iterative working needs. If large-scale technical analysis of policy is ineffective in delivering accountability, as argued above, then the resource could be better deployed. Emerging technology offers plenty of opportunities to create ways for front-line workers to get more rapid feedback on the effectiveness of their practice and new initiatives. For example, schools are already using sophisticated and continuous tracking of how pupils are progressing and to inform and improve teaching practice. The “Friends and Family” test in the NHS was introduced to provide immediate and specific feedback on patient experience. If government gave a clear signal, backed by a financial commitment to invest in new tools, that such real-time accountability was a priority then it is very likely many more methods would emerge.
These two suggestions – switching resources to front-line feedback and putting greater weight on evaluation and audit in understanding impact on culture – are not revolutionary. They will not immediately solve fundamental issues around what a democracy actually holds leaders to account for, or how that relates to what they are able to do. But they may help us make better policy within that environment, and have a more honest conversation about what that policy is trying to achieve.
 “Is your policy a dodo?” Civil Service Quarterly October 2014
 Agendas and Instability in American Politics Baumgartner and Jones 1994
 “Personality and Politics: values, traits and political choice” Capara et al Political Psychology 2006 Vol 27 Issue 1
 “Personality, Authoritarianism, Numeracy, Thinking Styles and Cognitive Biases in the UK’s 2016 Referendum on EU Membership” onlineprivacyfoundation.org 2017
 Marquette University Poll in Wisconsin reported at http://www.jsonline.com/story/news/blogs/wisconsin-voter/2017/04/15/donald-trumps-election-flips-both-parties-views-economy/100502848/
 “The United Kingdom: Privatisation and its Political Context” David Heald West European Politics 1988 vol 11 issue 4