On the transformation of children at-risk into an investment proposition: A study of Social Impact Bonds as an anti-market device

Competition

In line with existing studies of Social Impact Bonds that have raised concerns regarding the length and costs of contractual negotiation (PIRU, 2015), in Essex from first pursuing the idea, through negotiations, to then issuing a contract took around 29 months. This lengthy and costly set-up period included setting up the special purpose vehicle and deciding on the appropriate form of intervention, how it would be measured and the types of outcome payment made to investors.

The agreed aim of the Social Impact Bond was to deliver an intervention that could prevent children from being taken away from their families and placed in residential care operated by the local political authority, Essex County Council. Such care was costly to the local authority and also frequently led to children experiencing further issues in education, crime and life opportunities. The Council and investors agreed that choosing an intervention should be evidence-based to ensure the viability of the scheme. The only intervention that the parties agreed upon as providing a sufficiently compelling evidence base was developed in the US: Multi-Systemic Therapy (MST). ¹⁵ MST Inc. could provide training for UK therapists to engage with children identified as at-risk of going into care. A charity called Action for Children was awarded the Service Provider Agreement to manage the trained therapists in delivering MST through two teams of four therapists, a team manager and business administrator, dealing with four cases at a time on a rolling basis. MST would involve in-home or school therapy to try and get at the root of problems faced by the children. For children it would involve 60 hours of contact over four months.

The local authority as commissioner, investors, Action for Children as service provider working with MST Inc., and the new special purpose vehicle had to contractually agree on specific outcome measures. One key target was that over five years, 380 families should be taken through MST in 20 cohorts in order to try and prevent 110 children from going into care. A specific measurement was required to then tie this target to a successful outcome that demonstrated a cashable saving. The specific measurement for attributing savings to the intervention involved calculating ‘days of care averted’. This required that a cost was attached to each day of care, monitoring those children who went through MST and did (and did not) enter care, then tracking those children over the lifetime of the Bond. Calculating how many children went into care set against the prevention target and calculating ‘days of care averted’ were central to the contractual set-up of the Social Impact Bond – effectively tying together financial risks and returns.

The lengthy contract negotiation produced the following financial structure: £3.1m would be provided by investors (including £825,000 from Bridges Ventures and £825,000 from Big Society Capital ¹⁶ ); Essex County Council projected a cashable saving of £17m in total through a reduction in costs of children going into care, of which they would retain £10m and pay out up to £7m to investors; £120 was attached to each ‘day of care averted’ as a payment to investors, and this figure was achieved by calculating the average cost of care (in the range of £20,000 to £180,000 per child per year ¹⁷ ) and a distribution of savings that would enable Essex County Council to achieve its £10m savings target and pay investors around an 8% to 12% annual return on investment. Unlike other Social Impact Bonds where payments could only be made after a number of years upon successful outcomes, the ‘days of care averted’ calculation enabled payments to be ‘frontloaded’, meaning that investors could start to see a return immediately. Further forms of measurement were derived through comparing the success of the MST-based intervention quarterly to a ‘control’ group based on a business-as-usual counter-factual derived from historical data and the whole scheme would be evaluated by the Office of Public Management. ¹⁸

The contract was thus an important site of negotiation for setting in place the terms of intervention, the impact to be achieved, how the impact would be measured and how returns to investors would be triggered. As a single item, the contract became a focal point for drawing together multiple parties in the intervention but also became a means through which the intervention could be made to make sense.

The contractual negotiation and its outcomes were significant in two ways for the constitution of market competition. First, the authority’s ability to calculate the impact of the intervention was somewhat undermined. Core management in the local authority (rather than those delivering children’s services) took the decision to enter into a Social Impact Bond in order to bring extra finance into edge of care services as part of a package of measures to try and improve delivery following an ‘inadequate’ assessment from the government inspector OFSTED. However, by the time the Bond had started to offer MST to children, the service had improved significantly and this was recognized by OFSTED. Essex Children’s Service had been placing large numbers of children in care as a risk-averse, safety-first measure in the immediate aftermath of receiving their initial inadequate assessment. However, as the following interviewee suggests:

… um by the time this [Social Impact Bond] actually got going actually became a delivery vehicle using MST to deliver edge of care service that would eventually produce the number of days of care saved it was more than three years into the project and in that time Essex moved from being an authority in special measures with a very high rate of children in care to an authority that was well on the way to being assessed as good and in 2014 did become a ‘good’ authority for delivering children’s services with the number of children in care plummeting. So by the time the intervention actually hit the ground the authority had already made huge um strides into dealing with the problem. (Interviewee 1: Local Authority)

The time it took to set up the Social Impact Bond made it difficult to attribute a definitive effect to the MST-based intervention. The business as usual counter-factual based on historical data depended on a ‘business’ (sending large numbers of children into residential care) that was no longer ‘usual’ (as the form and practice of intervening in children at-risk had changed). For this interviewee from the local authority, other interventions that they had already put in place could be just as likely to lead to a decline in the residential care population:

Part of [our reduction in the care population] and this was recognized by OFSTED was an investment in upskilling basic social work practice, so really ramping up social work practice, really expecting social workers to be doing good quality direct work with families … (Interviewee 1: Local Authority)

Although those delivering edge of care services for children at-risk of going into care recognized some difficulties in attributing the reduction in the care population to the Social Impact Bond and to MST, they were contractually obliged to make upfront payments to the investors to cover the costs of the intervention. If the conception of a market that sat centrally in Social Impact Bonds involved the allocation of resources to the most effective and efficient provider, this set-up that insisted on payment through contractual obligation, regardless of the ability to attribute an effect, seems somewhat different. Instead, these obligations demanded that pay-outs were made regardless of the ability to measure an effect and attribute that effect to MST.

A second consequence of the contract negotiation was the agreement to use MST as the therapeutic intervention that would help children to remain outside of residential care. The terms expected of a Social Impact Bond contract (according to the Cabinet Office Centre for Social Impact Bonds) are to focus on the outcome, rather than its delivery, creating a form of measurement that enables returns to be paid to investors based on outcomes. As previously mentioned and as the following social investment fund manager suggests, the terms of the contract should enable a local market for competition in the initial choice of intervention and its ongoing use (if any problems in achieving outcomes emerged during the intervention, competition from other providers should lead to a change in the service provider). In this sense, and following Callon and Muniesa (2005) and Cochoy (2010), the Social Impact Bond should operate as a market device by bringing choices in intervention together into a single spatial and temporal frame, maintaining competition through the possibility of switching interventions, utilizing competition as a driver to generate ever more singularized offers. As the fund manager suggested in general:

… the benefit of these schemes is that all you need to do is achieve the outcomes. No one cares how you deliver them. So if you’re seeing in a year into a programme that something’s not working and you’re not achieving outcomes you could change the delivery model entirely, as long as the focus is on outcomes. (Interviewee 6: Fund manager)

However, the situation in Essex was different:

… Essex is an interesting model where there’s value to the intervention risk because we know MST works. It’s more around the operational risk. I can’t remember the contract they’re investing in, but there the intervention was prescribed. Most Social Impact Bonds are not. Even there, actually, the SPV has the power to fire Action for Children and to do something entirely different. So there is quite a lot of flexibility. It would never happen, but it could happen. There is quite a lot of flexibility. Essex is one of the more rigid ones around the intervention. (Interviewee 6: Fund manager)

Contractually it remains the case that via the Special Purpose Vehicle (SPV), the Council could replace Action for Children as the service provider, but as the interviewee points out replacing Action for Children ‘would never happen’. One reason why this change would not take place is that Action for Children were simply being trained to provide an intervention by MST Inc. and at no point during the intervention did anyone raise a concern that MST was being delivered incorrectly (even if its appropriateness as a therapy was questioned, see below). Replacing Action for Children would mean simply replacing the organization that delivered MST as the favoured form of therapy. Furthermore, MST was contractually held in place as it provided the evidence base for the key measure of success for providing a return to investors – ‘days of care averted’. Given that the Social Impact Bond was entirely built around this measure and the contractual terms were entirely dependent on this form of measurement, changing from MST to a different kind of therapy, without the same evidence base, without the means to clearly calculate ‘days of care averted’, would be impossible without entirely reworking the lengthy and expensive contract negotiation process. To change from MST would require a whole new contract which also ‘would never happen’.

The local market for provision wherein the potential to switch intervention should ensure efficiency and effectiveness was thus eliminated by the contractual requirements for a specific evidence base. Although Callon (1998) suggests, drawing on the work of Weber, that ‘agents enter into competition with one another to secure points of monopoly and domination’ (p. 43), with the result that ‘the very nature of competition is to rarefy competition’ (p. 44), and it seemed that the local competition envisaged within Social Impact Bonds might lead to a kind of ‘bilateral monopoly’ (Callon, 2016, pp. 25–26), achieved by service delivery organizations increasingly attuning or singularizing their offer to commissioners, that does not quite capture this situation. There is no prior competition that then leads to monopoly and domination; MST was the only therapy that was considered during the contract negotiation. The contract and its negotiation then resulted in the Social Impact Bond operating as an anti-market device, ruling out competition by establishing exclusivity of service to MST and Action for Children over the course of the Social Impact Bond. ¹⁹

The local authority edge of care services manager reflected on the restrictions imposed by the evidence base requirement that ruled out potential competitors to MST:

… we’re much more constrained than we initially thought … I think if we did go forward we might not tie it to a rigidly evidence-based intervention – we might truly hold to the principles of innovation in SIBs and we might say we’re only interested in the outcomes and we’re going to work with a provider … who will get us there somehow as long as it’s legal and we’ve got defensible decision-making – but actually the outcomes and sustaining the outcomes are the really key thing. (Interviewee 1: Local Authority)

What seems apparent is that in place of a market device that would stimulate competition by disentangling entities (Callon & Muniesa, 2003) from their previous connections, enabling them to be arrayed within a single spatial and temporal frame (Cochoy, 2010), enabling choices to be made and the entangling of entities into new relations (Callon, 1998), came an anti-market device. This fixed in place providers, investors, the form of intervention, the evidence base, a measurement device and front-loaded payments to investors. Instead of competition leading to temporary bilateral monopoly (Callon, 2016), effectively the contract ruled out competition and jumped straight into guaranteeing a local monopoly through ensuring that MST was the only therapy that entered negotiations and had exclusivity over the delivery of the intervention. However, market aspects of Social Impact Bonds are not reducible to notions of competition alone. ²⁰ As I noted in the opening analysis of these Bonds, constituting an investment proposition as a means to entice investors into public service delivery as a way of managing austerity-inflected government cutbacks was a key reason for their promotion by, for example, the Cabinet Office. In the next section we explore the ways in which this Social Impact Bond generated a very specific form of investment proposition.

Investment proposition

The Essex Social Impact Bond for children at-risk involved a number of investors, including a ‘high net worth’ individual, Bridges Ventures (an investment fund specializing in social and impact investment ²¹ ) and Big Society Capital (an independent financial institution enabled by government legislation to set up a fund that draws on dormant bank accounts to make social investments ²² ). In the preceding section, I suggested that localized competition was effectively ruled out by the rigid contractual structure of the Social Impact Bond. However, this was not the only market-mimicking feature of the Bond. The aforementioned investors took part in the constitution of an investment proposition – calculating future returns and assessing risks – that put in place a set of relations that appeared superficially similar to the investment market.

The importance of the investment proposition was shaped by the approval requirements of the Cabinet Office Centre for Social Impact Bonds. For example, an intervention could only be approved if it identified cashable savings, a percentage of which would be paid to investors, thus establishing potential future earnings for the investor. Costs of the intervention and an outcome should also be clearly stated, along with the measure that would establish the relative success of the intervention, with a failure to reach the outcome threshold putting returns and capital investment at stake. The risks to investors could thus also be established in relation to potential future earnings. In this way, the proposition looks very much like other forays into the investment market. Yet there are two ways in which this market-like vision of investment, return and risk is made less clear-cut.

First, although in its very first public outing as a Social Policy Bond, it was envisaged that there would be a market for such Bonds through which investors would be able to cash their positions, this has never emerged. ²³ Partly this is because investors form part of, and lend expertise to, SPVs (along with commissioners) to oversee interventions. To cash in their investment position would require a change in the membership of the intervention and the kind and quality of expertise on offer. Furthermore, as policies, Social Impact Bonds were given stimulus and a practical shape during 2008–2009, emerging through austerity government and as a response to the financial crisis. As investors were still being publicly chastised at this time for lending without responsibility and then cashing their investment positions, possibly causing the crisis, it may have been difficult for the nominally left of centre Labour Party to gain sufficient support for a new market for social problems within which investors might similarly be given an opportunity to shed their responsibilities. As a result, there is in effect no liquidity for investors in these Bonds in the sense that an investment position cannot be cashed.

Second, there is the investment proposition itself and the constitution of upfront costs to be paid by the investors, and their returns and risks. Initially at least, the investment proposition might appear high risk: investors stand to lose all their capital and miss out on any returns if an intervention does not succeed. However, this investment proposition and its attendant risks are somewhat modified. Within the UK government’s approval mechanism, once identified, cashable savings cannot be the result which triggers payments to investors. Instead, the trigger-measure has to be something more immediate which offers an indication of a future saving (in the case of children at-risk, this is ‘days of care averted’ rather than the closure of residential care homes). This separation of measures (that trigger returns) and savings (that must be projected, but do not trigger payments) is designed to reassure investors that they will receive reasonably prompt returns. It is designed so that their rate of return will be tied to the success or failure of the intervention funded, rather than its future savings which might be reduced or eliminated by other factors beyond the intervention (including changes in government policy or an increase in other areas of costs, for example, for children’s services). Risk is thus redistributed from government to investors, but within structured terms designed to limit investors’ exposure.

According to Liebman (2011) and Warner (2013), Social Impact Bonds require this kind of structure to create the conditions that might inspire innovation in interventions (or at least prove attractive to investors). The UK government has attempted to enhance such conditions for investors through Social Investment Tax Relief. Since April 2014 investors ‘can deduct 30% of their investment from their income tax liability’. ²⁴ The distribution of risks shifts again here. Although investors’ capital might be wholly at stake in a Social Impact Bond, a proportion of that risk can be deducted from tax liabilities that they would have paid anyway. Adding in the possibility that investors can actively promote their involvement in attempting to reduce recognized social problems, what might initially appear a significant financial risk, could be recast as a tax offset, publicity opportunity and a chance to make a return on investment.

The Essex Social Impact Bond for children at-risk went further. The investment proposition became more attractive once it was agreed that ‘days of care averted’ could act as a measure of the success of the intervention and that this could trigger payments to investors. This meant that as soon as a child entered into MST, money could begin to accumulate and returns to investors could be made. Rather than waiting on a lengthy and uncertain outcome as investors have had to do in other Social Impact Bonds, in Essex the returns were quite immediate. Also, frontloading the payments from the moment when a child entered into therapy ensured the investors received a partial return immediately, regardless of longer-term future outcomes.

The problems with this investment proposition were recognized by many of the interviewees who took part in the research. As the following self-named, philanthro-capitalist suggests:

One of the great perverse incentives in a thing like – Essex was a very good example of it, but it happens in all Payment by Results things. We want to make sure the children who should go into care do go into care, not that they’re kept out of it because then they [the investors] get a payment. (Interviewee 7: Philanthro-capitalist)

The ‘perverse incentive’ appears to rank the need to give investors a guaranteed return ahead of assessments of the appropriateness of keeping children out of residential care. ²⁵

How did this investment proposition come about? I have already noted that the contract negotiations were lengthy. These upfront activities were also costly for the local authority, who required a combination of legal, economic and policy expertise in order to take part in the negotiations. In the same way that others (PIRU, 2015) suggest that money ‘saved’ through a Social Impact Bond is effectively merely a redistribution of costs from the intervention back to its set-up, Essex County Council core management faced significant set-up costs. Furthermore, the contract negotiations appeared to bring together parties into the same space with differential calculative agency (Callon & Muniesa, 2005). This uneven distribution of capacities to calculate seemed to result in certain parties (such as the investors) being able to forecast the future with certainty. Meanwhile, for other parties (such as the local authority), their ability to calculate the future with any certainty, retrospectively certainly seems to have been limited, meaning their projections of, for example, cashable savings, were later opened to question as the Social Impact Bond began operating.

We can see this calculative asymmetry and differential confidence in projecting future effects in quite straightforward ways. MST was not cost-free. In US interventions using MST ²⁶ the marginal cost was US$4246 ²⁷ (approximately equivalent to £3418). Investors knew their fixed costs. Yet returns, that formed the principal cost for the local authority, were less fixed, being based on ‘days of care averted’ – the more days averted, the higher the cost for the local authority. At the same time, the local authority had to cover further costs, including for example a high turnover of staff during the intervention leading to extra recruitment and training costs. These extra costs to be met by the local authority had not been foreseen. Although it is unclear to what extent an uneven distribution of calculative agency meant that one party (the investors) could foresee potential general categories of costs (such as staff training) and contractually avoid them while another party (the local authority) did not foresee such costs and was stuck with them, the contract certainly instituted these terms. In terms of the investment proposition, the costs for investors were fixed and the costs for the commissioner were not.

Costs for the commissioner continued to rise. Front-line workers in children’s services held the view that MST was only suitable as an intervention for a small number of children:

One of the things we’ve found with MST there are a number of exclusions with MST … for example young people on the autism spectrum are excluded from it. … it actually isn’t an intervention designed for crisis edge of care work [if you had an urgent problem] MST can’t deliver that. (Interviewee 1: Local Authority)

As a result, Essex County Council faced what a former UK Treasury advisor called ‘a double spend’ problem: having to maintain children’s services for cases where MST was not suitable or failed, and having to make frontloaded payments to investors. ²⁸ The double spend was absent from the local authority’s projected costs.

In this sense, the intervention was only partially effective (it suited certain children in certain circumstances) and not very efficient (it did not allow the County Council to make the kinds of cashable savings they projected or that would be apparent if, for example, they could close residential care units). For the UK government Cabinet Office, such cashable savings through Social Impact Bonds remain a ‘question to answer’. An investment ‘facilitator’ who took part in the research also suggested that for the local authority most savings were actually quite ‘theoretical’. These costs were absent from the investment proposition as they did not fall to the investors.

How could the local authority not have projected this double spend at the negotiation phase, before they were committed to such costs? What can explain the calculative asymmetry between investors and the local authority? According to those tasked with delivering children’s services, this problem stemmed from front-line staff being excluded from contractual negotiations.

There is a disconnect between the strategic needs and the operational needs of the County Council. They [central managers who took part in the negotiation] don’t understand [edge of care needs], just as I don’t understand many of the nuances of their practices. They assumed that edge of care evidence [provided by MST] was fine. It doesn’t do that, it can’t do that. … So they can make assumptions that an ordinary person would make – ‘you just refer someone to MST’ – but no you can’t. (Interviewee 1: Local Authority)

Calculative asymmetries between investors and the local authority in contract negotiation could partly be explained by this absence of front-line workers, with County Council managers assuming that MST was well-suited to all children at-risk. The investors were happy to see the contract built around MST as it enabled them to clearly project their costs (the level of investment based on costs of MST and the likely number of children that would go through MST) and their return (based on frontloaded payments as soon as children entered MST). Once the Social Impact Bond was up and running, an initial response from those delivering services to children at-risk was to not refer children to MST, eliminating half of the double spend. If it had continued, this position may have shifted significantly the investment proposition as fewer ‘days of care averted’ would have been amassed. However, not sending children to MST led to antagonism in SPV meetings between children’s services, local authority managers and the investors keen to start building up frontloaded payments. This resulted in higher numbers of participants in MST in order to meet targets for the number of children that would go through therapy during the life of the Bond. Meeting these targets meant the investment proposition remained secure: the costs for each child remained fixed and returns began to build up for each child in a predictable manner as soon as they entered into MST, regardless of the outcome.

For Essex County Council’s frontline delivery team, their assessment of costs and savings was mostly negative:

Would Essex County Council look at this intervention and say this is saving us money? It probably wouldn’t at the moment. It would probably not do that. At the moment it would probably say ‘would we buy MST if the SIB was not here?’ which is the key question, is it that important – and the answer is probably not. (Interviewee 1: Local Authority)

Through a close examination of the constitution of the investment proposition and the various activities required to put this into action, we can start to see a variety of issues with costs, projections, calculations and returns. The costs for the investors are fixed and returns begin to accumulate, while for the local authority savings are not made, costs expand significantly and returns to investors still have to be paid. But is this still an investment proposition? It is not an investment that can be easily capitalized (Muniesa et al., 2017) as it cannot be cashed in a Bond market and the proposition is certainly modified by the availability of social investment tax relief, the measurement device with its partial dislocation from cost savings, and frontloaded payments to investors. The asymmetrical distribution of calculative abilities also seems to suggest that rather than facing investment risk, potential future problems for investors are recognized during the contract negotiations and protections against these risks are built into the contract. In the absence of risk, there is no need to discount future earnings (Doganova & Muniesa, 2015). Instead, this looks more like rent than an investment proposition, whereby the local authority are paying to use the financial resources of the ‘investors’ without any loss to the latter’s original resource. It seems to provide an example of what Bear terms: ‘the hidden rentier regimes through which political and economic influence is maintained by the tiered brokering of access to resources’ (2013, p. 394). However, reversing the conventional logic of a rentier state that enables external parties to pay rent to access indigenous resources, the rentier in Social Impact Bonds is the ‘investor’. The contract negotiation operates as the basis for investors to take part in solving social problems and as a form of brokerage, deciding who will take up what position on costs and risks within a tax and incentive structure that seems to lean heavily towards the ‘investors’. The latter are paid rent by the commissioner to access their resources, with the rent drawn from tax income. ²⁹ Whereas in the first part of the analysis I suggested that local competition was ruled out by the contractual demands built around MST, here the central components of the investment market (liquidity, discounted cash flows, risk) are eliminated by the protections for ‘investors’ built into the contract. In this sense, the Social Impact Bond maintains its position as an anti-market device, preventing the intervention from mimicking the investment market. In the Conclusion I explore what this means for future interventions that anticipate a widespread development of Bonds.