Beyond any specific objectives, the broad purpose of any government policy must be to make society better off, whether now or in the future. Is this true of the youth wage subsidy as proposed in the Employment Incentive Bill recently passed by Parliament? While opponents of the policy have often been characterised as operating based on ideology rather than evidence, it is difficult to discern whether that is true as proponents of the legislation have failed to adequately acknowledge how risks associated with the incentive could impact broader social welfare. In a recent book, Public Policy in an Uncertain World, the respected econometrician Charles Manski has argued that too often the producers of evidence for policy express incredible certitude about their findings, aided and abetted by the desire of policymakers and politicians for simple results that can be easily understood and easily sold. Making good policy, however, requires an honest appraisal of the strength of evidence available and adequate accounting for how uncertainty about a policy’s possible effects relate to our social priorities.
To understand why these general concerns are relevant to the youth wage subsidy, note that there are three possible direct effects of the scheme. The first is the intended one, namely that employers increase the number of jobs in response to the lower cost of young, unskilled workers. The second, a concern raised in particular by trade unions, is that employers will replace existing older workers with ones who qualify for the subsidy. Finally, and the subject of less discussion, is that the scheme will simply subsidise existing workers of qualifying age or those in new jobs that would have been created anyway. In practice, all three outcomes are likely – some new jobs will be created, some workers will be displaced and some jobs will be subsidised that either already existed or would have been created as a function of year-to-year economic growth. Whether the policy is desirable, then, depends entirely on the relative extent of these different effects, since the second trades older workers’ welfare for that of younger ones and the third simply leads to a transfer from the fiscus to a subset of existing employers without any benefit to the unemployed.
There are a number of different sources of specific evidence that could guide our expectations of what the net effect will be. Studies of other countries’ experiences are an obvious place to start but while proponents have cited a number of cases where a subsidy appeared to create jobs, opponents have noted cases where there was little effect or negative outcomes in relation to displacement of existing workers. Leaving aside the favoured emphasis of different parties, the range of experiences is a first indicator of uncertainty. Another kind of evidence is existing academic studies that have attempted to estimate the relationship between wages and employment using historical data. Although this is unfortunately not a very reliable way of determining what will actually happen if we alter wages in a particular sector, the National Treasury proposal notes that this literature produces a wide range of estimates from 0.2 to 2.2: decreasing the wage by 100% would increase employment 20% or 220%. More uncertainty. The Treasury’s forecasts of the likely effect appear to simply take an average of these findings (1.0), to assume that a given percentage decrease in wages would lead to the same percentage increase in employment.
There are two sources of more direct evidence: a randomised trial of a conceptually similar wage voucher scheme and a study that attempted to estimate the possible effects using a mathematical (general equilibrium) model of the economy. The results of the randomised trial were summarised in an article by Neil Rankin, though unfortunately the full study remains unavailable, and the economy-wide modelling study was published in the journal Development Southern Africa earlier this year. In each case the devil is very much in the detail. While randomised trials are fairly good tools for estimating the causal effect of an intervention at a small scale, there are many reasons why those effects are not likely to be representative of what will happen when an actual policy is implemented on a national scale. In this case, there are a few specific concerns. First, individuals in the experiment may simply have been more likely to get jobs because they were cheaper to employ relative to other job seekers; the jobs they got may not have been created because of the voucher. Implemented at the national level where all young job seekers are subsidised the effects are likely to be lower, possibly to the extent of being insignificant.
A second problem is that the experiment, by using a voucher instead of a tax rebate and being implemented by university researchers instead of government, is quite different from the intended policy. Arguably a large-scale, widely publicised implementation by the South African Revenue Service may be more likely to encourage employers to create jobs, but in this we are simply back to speculation. And that is the third problem with experiments: they rarely help us understand why they work, which makes it difficult to use these results to say what the effect of the actual policy will be. As an example, one strange result from this trial was that many employers did not claim the voucher. That calls into question whether the monetary incentive was the reason for the difference between those with vouchers and those without.
The problems afflicting modelling studies are different but no less important. At the national level financing of the scheme is a critical issue and the study assumes that "the subsidy programme is financed by raising average household income tax rates by between 10.9 and 12.5%". Another key assumption, standard in modelling but relatively implausible in practice, is that any excess transferred to firms is passed on to consumers. This assumption of perfectly competitive markets means that subsidies paid for existing workers or jobs that would have been created anyway simply lead to improvement in the welfare of consumers through lower prices and more consumption or savings. However, the existence of anti-competitive behaviour and market power is the norm rather than exception in the South African economy so it appears more plausible that transfers of this kind will find their way into company profits. Depending on the extent of these, the result could be that a well-intentioned youth wage scheme leads to a regressive transfer from general tax revenue to firm profits with no effect on employment.
This casts a different light on ideological and political issues. It is obvious that from a firm perspective the policy is desirable, regardless of whether it creates any jobs at all. Similarly, for political parties with a sizeable business constituency the policy presents a win-win proposition: at best the policy will benefit unemployed youth – with all the broader social benefits of that – and businesses, at worst it will benefit only business. Treasury’s original policy proposed to subsidise existing workers (up to 600 000) with vague justification based on encouraging formalisation of enterprises. This has been omitted from the recent Bill albeit with relatively little explanation. That appears to be a good move given the concerns above and that the forecast effect on new job creation was barely 25% of this number, though unsurprisingly the decision has been met with disappointment from certain parties that may have profited from it.
That change addresses the most glaring problem with the original proposal but a second issue remains: the inevitable subsidisation of jobs that would have been created anyway. Buried in the technical appendix of the original discussion document is an observation that "about 58 per cent of the job creation would have taken place in any case". In other words: under even what are arguably optimistic projections, more money will be paid for jobs that would have been created without the incentive than are created because of it. Treasury has referred to this issue in public debates using the technical term "deadweight loss", which economists know is a possible outcome of any intervention. However, to my knowledge there has been no public acknowledgement of how this affects the merits of the policy. Perhaps the most striking issue is that the deadweight loss is essentially certain, while the actual benefits of the policy (in terms of new jobs created) are highly uncertain – as should be clear from our discussions of the evidence available. Yet reading public statements by treasury, business and certain political parties even an informed member of the public would not know that the most likely outcome of the policy may be a socially regressive transfer of tax money to profits. Perhaps proponents see this alternative scenario as a way of giving businesses a tax break? If so, they should say as much publicly.
A last set of issues concern the practical challenge of implementation. Various statements by proponents have argued that replacing older workers with younger ones qualifying for the subsidy would "not be good business sense". Interestingly, while this vague claim was not supported by any evidence it appears that estimates of the policy cost as well as forecasts of its effects have just assumed that no displacement will happen. More unacknowledged uncertainty. In fact, this claim contradicts the available evidence: a survey conducted with the randomised trial found that 20% of employers said they would replace workers in this fashion. Given a strong incentive not to admit this, that number is likely to be an underestimate. Related to that is the very real danger of large-scale fraud that would turn the policy into a massive rent seeking opportunity, not just through displacement but also the dismissal and rehiring of employees. In theory that is disallowed but despite fuzzy reassurances from treasury and political parties, no detail has been given on how such behaviour would be detected. In relation to displacement, the Bill places the onus on dismissed employees to prove in a formal forum (like the Commission for Conciliation, Mediation and Arbitration or Labour Court) that they were fired to employ a worker who would be subsidised. This may be an impossible standard to meet and effectively places the responsibility of protecting social welfare on relatively disempowered individuals.
When one looks at the policy in the above light, conclusions that the youth employment incentive will be "a relatively cheap and effective way to create some employment" are, at best, premature and optimistic. If anything, careful consideration of the facts suggests that the scheme may do more harm than good. So much for the triumph of evidence.
Sean Muller is a lecturer and PhD candidate in economics at the University of Cape Town. His PhD research examines the use of randomised trials to inform policy, with a focus on education interventions?.