How much influence could you have by becoming a politician? Common sense says that politicians have a lot of influence, and it’s a serious contender as a high impact path for someone who’s altruistically motivated. But aren’t the chances of success incredibly low? Our guess was that even though the chances are low, the potential impact is still very high. So, when we were asked about UK politics in a recent case study, we decided to make a more detailed estimate of the expected influence to feed into an overall analysis of politics as a career path.
We found that chances of success are low, but for some students they’re not low enough to offset the very large potential influence. The UK government budget is £720 billion, and even a small chance at influencing a budget that large could be highly significant (and the impact of politicians extends well beyond budgets).
We’ve extended our analysis of the chances of an Oxford PPE graduate succeeding as a politician, to make a rough estimate that such a student can expect to be able to direct £7.5 – 75 million to the causes they support from their chances of making it into elected office. For a student similar to an Oxford PPE graduate, this suggests the path is competitive with the most high potential earning to give careers – such as those in finance – in terms of financial influence, which when combined with politicians’ law-making powers and advocacy opportunities could put politics clearly ahead.
Aren’t politicians highly constrained by existing policy, what other politicians want, the desires of the electorate and other factors? Yes, but these factors have already been included in the estimate. Read on to see the full process.
Summary of the estimate
Our preliminary estimate is that an Oxford PPE graduate who aims to become a politician in the UK, could expect to influence £150 million of government spending, arising from their chances of making it into office. A number of factors decrease the impact of that money; giving a quality-adjusted estimate of £7.5 – 75 million, falling towards the lower end if you’re primarily interested in very specific interventions (e.g. supporting a certain organisation) rather than broader ones (e.g. promoting evidence-based policy). This is the amount of government spending the graduate might be able to direct towards the causes they support.
For students without the typical attributes of Oxford PPE students, chances are significantly reduced. For instance, repeating the calculation but considering students from Oxford and Cambridge as a whole suggests expected influence on the order of £1 – 10 million. More generally, the expected influence is highly sensitive to the individual’s degree of fit with politics i.e. it could be substantially higher for someone with strong success in student politics at Oxford, and near zero for many others.
Our proposed estimate is extremely coarse. We rely on a crude economic model of influence within government, assume that this influence in aggregate accounts for all public spending, and try to estimate the share of influence possessed by a number of relevant groups. We believe this model is much stronger than it appears casually, and do provide some justification for some of the simplifying assumptions at the end of the document. We also explain some important caveats, such as our uncertainty over the prominence of MPs and ministers, and focussing mainly on Oxford PPE. Nevertheless, it is certainly an extremely crude model. The error on this estimate is at least an order of magnitude or so, and if there are significant issues with the methodology they may actually be even larger.
To compensate for this we have made conservative estimates throughout, and still arrived at a remarkably high number. Since the conclusion of this calculation is also supported by the common sense position that going into politics is high potential for students with the right characteristics, we conclude that the expected influence of entering this path is indeed very large.
Overview of the calculation
We estimate the total money moved as a product of the following fractions, stating our source:
1) Fraction of influence of one capable, motivated PPE graduate amongst MPs and ministers: 1/300.
- Fraction of influence of PPE graduates amongst MPs and ministers: 1/6 (based on empirical analysis of ratios of PPE graduates in office)
- Fraction of influence of one capable, motivated PPE graduate amongst PPE graduates: 1/50 (based on empirical analysis of chances of success)
2) Fraction of total spending which is influenced by MPs and ministers: 1/16
- Fraction of influence of MPs and ministers on policy: 1/8 (an estimate informed by surveying political science literature)
- Fraction of spending determined by policy: 1/2 (an estimate informed by surveying political science literature)
3) Total public spending: £720 billion (in 2013)
Preliminary total: £150 million
4) Factors which significantly decrease the value of this influence: 1/20 to 1/2
- To the extent your interests are unique, your influence is diminished by the difficulty of realizing gains from trade, forming coalitions, coordinating, etc: ¼ to 1, depending on the idiosyncrasy of your interests. (a judgement call)
- Influence acts in a relatively complicated way over long periods, which makes it difficult to have any particular specific effect, or to implement any complex changes: 1/5 to ½, depending on the specificity of your goals. (a judgement call)
This suggests a quality-adjusted total ranging from £7.5 to 75 million. We explain our reasoning behind each of these factors in the rest of the document.
1. Influence of one PPE graduate amongst MPs and ministers
We want to estimate the expected influence of one PPE graduate. To do this we need to estimate the influence of PPE graduates (captured by how many MPs and ministers studied PPE). We then need to estimate the influence of one graduate amongst all PPE graduates.
a) Influence of PPE graduates amongst MPs and ministers
According to the BBC about 5% of MPs are graduates of Oxford PPE, and Wikipedia provides a list of PPE graduates which covers roughly this fraction of current and (living) former MPs. The same sources suggest that the cabinet and shadow cabinet are 20-25% Oxford PPE graduates, and the density amongst PM’s may be even higher.
It would be difficult to estimate the significance of PPE within party politics in the UK more broadly, but it seems reasonable to presume that these numbers are crudely comparable.
How we should weigh these different proportions—5% to 25%—depends on how we think the influence is distributed between different levels of government. If we thought the majority of influence was concentrated amongst MP’s we would use the 5% figure (or lower); if the majority of influence is concentrated amongst ministers we would use the 25% figure. In between we would extrapolate arithmetically between these extremes, e.g. if we thought that half of the influence was with MPs and half was with the cabinet, we would obtain an estimate of 15%.
Given a broad distribution over the behavior of the government, it seems sensible to assign significant influence to each category. Anecdotally, influence on UK spending appears to rest disproportionately with the highest-ranking politicians (although it can be hard to determine the true influence, since influence may be expressed through many channels). Combined with the concerns about regression to the mean expressed below, it seems sensible to take a middle-of-the-range – and likely conservative – estimate like 15%.
Regression to the mean
Oxford PPE appears to represent a larger fraction of political influence than we might expect, for example considerably more than almost any other group which can be easily identified. To the extent that this estimate is surprising, we should expect the outlier that is Oxford PPE to regress to the mean. That is to say,we should expect other estimates of PPE’s influence to be lower.
In particular, I expect the estimate to be lower at different times (the influence of PPE seems to have held up over recent history, but I don’t know how it holds up across decades and would expect it to decrease in the future). Moreover, I expect it would be lower if we were to take a more comprehensive inventory of influence in a political career, including more non-parliamentary roles.
It’s not quite clear what to make of this consideration; fortunately, however, the fraction of MPs and cabinet ministers seems to be a strong proxy for political influence, and even more promisingly the available evidence suggests that these fractions may understate the real influence of PPE graduates, given the trend towards increasing over representation of PPE graduates as we examine more influential positions. This would suggest that our 1/6 figure is a conservative, reasonable estimate. Overall, concern about regression to the mean seems to be a modest consideration. It would also be relatively easy to eliminate by obtaining more robust estimates of PPE influence.
b) Influence of one PPE graduate amongst her academic year
To get a first approximation of the influence of one student amongst their cohort of PPE graduates, we can simply estimate the size of their cohort. There are about 200 UK students in each year of PPE, suggesting a first estimate of 0.5%.
Beyond this, estimating the impact of any particular student clearly depends on the student. For the purpose of analysis, we can divide the relevant characteristics into intention and aptitude. Reasoning about these characteristics and their impact is hard, but fortunately a large part of this work has been done for us by the selection process for Oxford (interviews) and the self-selection process for PPE (choosing to apply).
The great majority of PPE students have an interest in politics, but a majority do not intend to pursue a career in politics. The group which is committed to pursuing a career in politics is even smaller, representing perhaps 1/5 to 1/3 of students (based on discussions with PPE students). It is hard to know exactly what to make of this number, without collecting more detailed information about the political intentions of UK politicians—how likely is it for a student interested in politics to ultimately find their way into a successful political career?
The majority of members of parliament (and a greater majority of cabinet members) follow explicitly political career tracks, especially working for a party, working as a special advisor or political researcher, working at a think tank, or working as a lobbyist. This suggests that students with explicit political aspirations are significantly more likely to end up with a successful career in politics.
It seems plausible that graduates with high intention and aptitude represent between 1/40th and 1/70th of the total political influence within their year of PPE. This best guess is based on multiplying the size of the PPE class by the fraction with comparable characteristics.
Getting in vs. having influence
The above discussion concerns an individual’s ability to obtain a role in the political process, but can also affect their influence within that role. So estimates of how much these characteristics observably affect an individual’s role in politics will tend to understate the real magnitude of the effects.
Beyond those characteristics, politicians with clear, well-thought-out, and ambitious objectives seem likely to wield a disproportionate share of influence. To the extent they are more interested in influence (as compared to e.g. social consequences of political involvement) they may be expected to make trade-offs that further their own influence. To the extent they have a clear policy agenda they may have more such opportunities available to them (and can always sideline such an agenda if appropriate, minimizing the possibility of negative effects). Anecdotally, such considerations can easily make more than a factor of 2 difference, merely by changing the nature of interactions within the core executive and leadership of individual departments.
2. Influence of MPs and ministers on how the budget is spent
In addition to MPs and ministers, there are other influences on government behaviour. When policy is being formulated the civil service, think tanks, and other party members will normally have influence. Typically unions, business interests, lobbyists, etc. will also contribute to the policy-making process. When policies are implemented, the civil service and other actors play a key role in determining how the budget is actually spent. All these different actors have an influence over how government pounds are ultimately spent.
It seems sensible to talk about the fraction of influence which each of these groups wields. Unfortunately, there seems to be little academic or elite consensus on the influence of MPs and ministers as opposed to other actors, or about the extent to which policy determines the actual implemented government behaviour. We have therefore made some fairly conservative estimates. These estimates could be sharpened by a more systematic investigation of historical policy decisions; but again it is worth pointing out (in our defence) that it is very difficult to trace the actual influence of individuals involved in the policy-making and policy-implementing process, because it is almost entirely mediated by informal mechanisms.
We approach this question by considering party politics’ share of influence as a product of two factors: the relative influence of party politics on high-level policy and policy directives, and the influence of such high-level directives on actual behaviour.
a) Influence of MPs and ministers on policy
Formally, most policy influence is given to the cabinet and the heads of the various departments. However many other actors seem to collectively have a good deal of influence as well. After brief searches I have not been able to find any credible work that sheds much light directly on this question; this estimate is informed loosely by reading a few political memoirs, the news, and a textbook on party politics. There seems to be no academic consensus on the proportion of influence to apportion to MPs and ministers as opposed to other actors.
We can nevertheless give a rough, conservative estimate of the influence of MPs and ministers on policy. I am inclined to give about ½ of all influence to the electorate, by casting votes and substantially constraining the range of acceptable behaviours. Within party politics, I am inclined to give about ½ of all influence to MPs (and others on the same career track as MPs) and to give the other influence to other players within party politics, financial interests, and so on. And finally, within policy-making I am inclined to give about ½ of all influence to party politics, and the remaining influence to bureaucrats, regulatory bodies, international bodies, and so on. Overall then, we estimate the influence of MPs and ministers on policy to be ?.
Note that as in our discussion of the representation of PPE amongst MPs and ministers (1.a), this figure seems likely to be a conservative estimate. This is because anecdotally, ministers are generally viewed as substantially more influential than all other actors in the policy-making process.
b) Influence of policy on how the budget is spent
Policy decisions have a large effect on how political pounds are spent, but they don’t have complete control. Even specifying a fairly narrow budget item often leaves much leeway to the civil servants who actually spend the budget, and typically there are many additional steps between the determination of policy and the implementation of the budgeted activities (the actual creation of the budget, the environment in which people engage in the budgeted activities, etc.)
Again there seems to be no academic consensus on this question, but we can give a rough, conservative estimate. I feel like ½ is an appropriate fraction of influence to allocate to policy, based on intuition guided by similar evidence. If one was more optimistic about how effectively policy is implemented one might give a more optimistic estimate, but it seems reasonable to adopt a conservative figure.
Comparison to other opportunities to move money
Whenever we spend or move money, we face some similar issues in implementation. For example, if I give money to a charity, I have direct influence over which charity I pick, but I still need to leave many important details to the charity. If I run a company, I can manage my employees but ultimately must leave them some slack, etc. In light of this it isn’t clear if we should count the second factor of ½ above, or whether we should treat money moved by politicians as the same as money moved in other ways.
We do tentatively think it is reasonable to count this factor as an extra cost, though: many of the steps between policy-makers and policy implementation really are unique to government spending, and many of these factors can be much attenuated when giving or running a company by employing like-minded people, by applying lots of selection power which isn’t available to politicians, etc.
3. The budget
The UK’s budget in 2013 was £720 billion. This figure is very likely to increase in subsequent years.
4. Quality-adjusting money moved
It seems intuitively clear that the influence of an MP or minister is more constrained than the influence of a philanthropist. ‘Spending’ one’s notional influence over the budget in the policy-making process is not the same as spending your own money. Some of this simply comes because influence is somewhat diffuse in politics. But this is not the only issue, and there seem to be other considerations that further reduce the value of political influence.
a) Coordination and unusual interests
Making anything happen politically typically requires a high degree of coordination, e.g. buy-in or at least acceptance from an entire party, cooperation between implementers and designers of policy, or other politicians refraining from making a stink about a pet issue. To the extent that you try to pursue unusual interests, achieving this coordination requires implicit or explicit bargaining, application of political capital, etc.
In principle such bargaining need not reduce the efficiency of pursuing unusual interests. But in practice bargaining or coordinating may be difficult, and making explicit trades may be significantly less efficient than coordinating by virtue of shared values and goals. There also may be norms against such bargaining (rather than focusing on collective interests), which force it to occur implicitly and may further reduce its efficiency. One case where such bargaining is likely to be particularly difficult is across time—if you pursue goals that future regulators don’t share, then you are less likely to coordinate successfully with future policy-makers and implementers.
This is more likely to be a problem the more diffuse influence is, both amongst contemporaries—if the PM holds much of the influence then such bargaining is less important—and across time—if current policy-makers unilaterally set current policy, then you should expect most of your influence to occur in the short term and you shouldn’t much care whether future policy-makers share your values or like your policies.
If your policy goals are as broad as any, so that you can achieve coordination as well as anyone, then this factor should be neglected—it’s not important whether coordination is perfect, just how well you can coordinate relative to other influencers (e.g., if coordination was very hard and things were very rarely changed, then whatever changes you were able to implement would stick around for a long time).
If your goals are quite unusual, such that few other political actors share them, then you may suffer a very large cost from the inefficiency of coordination (and we might expect an effectiveness-minded person to have unusual interests). I am inclined to model a separate cost for the effects of bad coordination across time and at the current time. Each of them seems reasonably likely to impose significant losses, say a factor of ½. It is easy to imagine much larger losses. It is also possible to imagine smaller losses, primarily because of uncertainty about the actual helpfulness of common goals—it seems quite plausible that almost all political activity must be achieved by more or less explicit bargaining anyway. This uncertainty, and the prospect of heterogeneity amongst different modes of influence, makes it hard for the expected fraction to fall far below ½.
Combining these two factors, we obtain a total factor of ¼.
Note that regardless of the considerations in this section, we should expect to see the political process preferentially furthering common goals, so we have to be careful when getting empirical evidence. We expect to see popular projects succeed much more frequently, but this is largely because they enjoy the support of many influencers (each of whom can only claim a small fraction of the credit). Similarly, we expect to see politicians focus on those of their interests which are broadly held and agreeable to others, but this would be true even if there was frictionless bargaining (as long as focusing on big common issues rather than a variety of unusual interests is actually socially efficient).
b) Influence over time and specific projects
Much of the influence that policy-makers wield seems to be somewhat indirect, via changing discourse, norms, precedent, and so on. This especially true to the extent that political influence extends over longer periods than an individual’s political involvement. Although this doesn’t change the total influence of politicians, it makes it harder to achieve any fixed goal with that influence.
For example, we might imagine that each minister can change their department’s policies only a short distance from the status quo. This doesn’t directly change the total influence available to ministers (though it might open a window for other actors to have more influence). But it does mean that much of a minister’s influence comes from their effect not on current policy but on future policies (since their successors will be operating under a similar constraint). So if your goal is a relatively immediate one you may have much less ability to achieve it; instead, we need to understand the long-term effect of the precedent you are setting.
Similarly, we might imagine that political actions are largely constrained by the expectations of political peers, maybe because policy-makers are unwilling or unable to risk disapproval or push against established norms. This doesn’t directly decrease the total influence of a politician, since the reduction in their direct influence is compensated by an influence over norms. But it does make it harder to understand the effects of actions, which may be mediated mostly by slightly shifting social expectations, and it also means that it is much easier to have broad effects than specific effects.
This pattern seems to be common to many simple models, and anecdotally to be true of policy-making in practice.
If you have a very general and broad goal, like increased reliance on quantitative estimates or a larger welfare state, then these may not be huge constraints. Simple changes can have a foreseeable impact at this broad level. You can predict that precedent will behave one way or the other, you can support people and groups that share your goal, you can push discourse in the appropriate direction, etc. Each of those things nevertheless is somewhat unpredictable and has some losses, and even in extreme cases I am inclined to recognize significant losses, perhaps a factor of ½ in efficiency.
For very specific goals the situation is worse, and it seems that much of the politician’s influence might be unavailable. E.g., influence extending significantly beyond one’s’ political tenure or social influences may be hard to apply to specific projects. I don’t know how to reason about the share of influence that will be inaccessible, and more empirical research could shed light on this issue, but a factor of 1/5 seems plausible (perhaps ½ – 2/3 of influence extending beyond political tenure, perhaps ½ of influence mediated by social effects that are too broad to accomplish specific goals).
Diverting funding from other projects
Overall, whether this impact is positive—and how positive it is—depends on how the money would have been spent. If you believe that that government spending is typically highly effective, then this may significantly reduce your impact. If government spending typically has a positive impact, but one which is significantly less positive than possible, then this effect is often quite small. For example, if your intervention increases the impact of the spending by a factor of 3 then the value of intervention is 2/3 what it would otherwise appear, which is quite large.
In practice, it seems that you can typically divert money towards projects which you consider significantly more valuable. This is not surprising, given that government spending serves a wide variety of interests.
Main sources of uncertainty
Our estimates of the amount of money a PPE graduate can be expected to be able to divert towards particular projects are quite uncertain – I would expect them to decrease rather than increase with further inquiry. The two largest sources of uncertainty in the analysis are points 2 and 4, that is to say:
the fraction of total government spending which is influenced by MPs and ministers, and
ad hoc adjustments for the difficulty of using politicians’ indirect and constrained influence to achieve a particular, potentially unusual, end.
The estimates used here are judgement calls informed primarily by secondary information about UK politics. The integrity of the analysis primarily rests on a modelling approach which aims to minimize the influence of these judgement calls on the final conclusion, hopefully limiting the impact of uncertainty.
Limited to Oxford PPE
In this article we are solely considering the outlook of an Oxford PPE graduate. We recognise that this is limited in scope. However we can offer three defences. First, this estimate was produced for an Oxford PPE student. Second, PPE is noticeably overrepresented, making this analysis far clearer – no other single degree at a single university is so well represented. Third, this analysis could be extended to other groups.
For example, we could estimate the influence of a randomly selected Oxbridge graduate. In a similar analysis to this article, Carl Shulman noted that “Oxbridge graduates formed 32% of Gordon Brown’s last Cabinet, and 69% of the recent coalition Cabinet”. Additionally, 24% of the 2010 cohort were from Oxbridge. Somewhere between these figures, let us assume that the fraction of influence of Oxbridge graduates is 1/3. There are around 4000 undergraduates in each academic year at both Oxford and Cambridge. We can therefore estimate:
1) Fraction of influence of one Oxbridge graduate amongst MPs and ministers: 1/24,000.
* Fraction of influence of Oxbridge graduates amongst MPs and ministers: 1/3
* Fraction of influence of one graduate amongst Oxbridge graduates: 1/8000
2) Fraction of total spending which is influenced by MPs and ministers: 1/16
* Fraction of influence of MPs and ministers on policy: 1/8
* Fraction of spending determined by policy: ½
3) Total public spending: £720 billion
Preliminary total: £1,875,000
4) Factors which significantly decrease the value of this influence: 1/20 to 1/2
This suggests a quality-adjusted total ranging from £93,750 to £937,500. Note that we have not taken into account the intention to go into politics, and we know that most Oxbridge students don’t try. We guess taking intention into account would boost the estimate by at least a factor of 10.
Money moved vs other kinds of influence
This analysis explicitly concerns money moved, largely because it is so much easier to treat quantitatively. Other kinds of influence seem likely to be subject to very similar considerations, and perhaps the same analysis can be applied almost verbatim.
The most important and fundamental change in the analysis would be obtaining a different estimate for the total impact of government activity of a different type. For example, we might look at how significantly regulation could distort private activity (e.g. how large an economic gradient can it compete against) and then do a similar analysis to estimate the total regulatory impact of an individual.
There will be additional complications in other cases (in addition to being more easily quantified, money moved is more fungible and generally easier to reason about), and overall I expect the analysis to be much more difficult.
Parliamentary terms and peer groups
In reality a single MP typically serves for many years, influencing a total budget much higher than £720 billion. Similarly, there are many cohorts of PPE students rather than just one.
However, by attributing one year of government spending to one year of PPE students in our model, it seems we can avoid either of these effects: the total influence of politics over N years is simply N times annual spending, while the total number of PPE students over that period is just N times the annual number of students, so the impact per student can be obtained by using the yearly versions of each figure.
This analysis suggests that we can safely reason about a single politician as representing (at least) a proportional share of political influence, even when that influence is mediated by complex processes which we don’t understand. However, it is worth keeping in mind that this influence might be exerted in the future, or in a more diffuse way than we anticipate.