Theory

I consider a variety of religions with a number of structural differences, but I focus on the following key variables: (1) who chooses the new leader (the present leader or the assembly), (2) the length of the appointment (for life or for a fixed term length) and (3) how powerful the leadership position is (figurehead, does not make legislative decisions or autocrat). I work backwards down the decision tree. I first start with a voting model and then show how candidates for leadership are groomed and selected in light of the voting.Footnote ¹¹ I make the following assumptions.

Voter utility when candidate C is chosen as leader equals

if C has been a leader before, and

if C has not been chosen before.

Starting with the first line, $$${{\hat{x}}_i}$$$ is voter i's most preferred policy position. Differences in policy position may reflect differences over the appropriate church doctrine and what is best for the religion's survival, or they may represent more mundane concerns.Footnote ¹² x_c is candidate C's most preferred position (known to all).Footnote ¹³ w is a weighting factor that depends on other variables (to be discussed later). If w = 1, then x_c will be implemented if C is chosen as leader. If w = 0, then $$${{\hat{x}}_m}$$$ , the median voter's most preferred position, will be implemented. The further the implemented position is from the voter's preferred position, the less utility the voter will have. B denotes the importance to the voters of the leader's policy relative to other characteristics of the candidate.

The next variables are the leader's valence characteristics – dimensions that are perpendicular to the issue space and valued the same by all participants. In contrast to policy, where voters disagree about which policy is best, there is general agreement among voters concerning valence (regardless of the voter's preferred position).Footnote ¹⁴

$$$\hat{a}$$$ is the optimal age of a leader and a_c is candidate C's age. A leader's age has both intrinsic and strategic value. With regard to intrinsic value, the optimal age, $$$\hat{a}$$$ , is the optimal balance between declining energy and increasing wisdom as a person gets older. When a leader's age is above the optimal, the leader's increased wisdom is not sufficient to overcome his loss of energy; when the leader's age is below the optimal, the leader's energy is not sufficient to overcome his lack of experience. It is possible that on occasion, a potential leader is both young and experienced or old and energetic, but we are dealing here with averages, not particular cases. The model does not critically depend on age having an intrinsic value. Therefore, if the reader is bothered by the assumption, the term $$${{(\hat{a}\,{\rm{ - }}\,{{a}_c})}^2} $$$ can be removed from the utility function.

K_c > 0 is C's charisma – how well the members of the religion relate to the person. A larger value means that more people will be attracted to (and stay with) the religion. While I do not have an empirical measure of charisma, I include it to acknowledge that there may be valence characteristics that influence a person's vote and to show that the main propositions will hold in this more general setting.

I now concentrate on the voter's expected utility when the candidate has not been elected to office before. Not surprisingly, there is uncertainty about the candidate's policy preference, which is captured by the stochastic term (ɛ). Each voter has an unbiased estimate of the candidate's policy preference. The larger Q is, the more certain the voters are about the candidate's policy preference.Footnote ¹⁵ If an individual has been the leader previously, the die has been tossed: the person's type has been revealed (as long as w > 0) and the uncertainty no longer exists.Footnote ¹⁶

When the appointment is for a term at a time, w will be relatively small, possibly 0. The leader needs to worry about re-election, so his policies while in office will bend more toward the median voter. When the appointment is for life, w is much closer to 1, as the re-election constraint no longer applies. The longer the leader has been in office, the more powerful he is (that is, the larger his scope of control), and the larger w will be. w is a positive concave function of t (length of time as leader) and L (L = 1 when the appointment is for life and 0 otherwise). For convenience, I assume that no two candidates are identical on all dimensions.

Figures 1A and 1B demonstrate this proposition, where B = 1, Q = 1, w = 1 and K_c is the same for all candidates. A more general proof is found in Online Appendix A. There are three voters whose preferences are on a left-right continuum: L (a voter weakly to the left of M), M (the median voter) and R (a voter weakly to the right of M).Footnote ¹⁷ The valence characteristic of the candidate (age) is measured vertically. The optimal age is on the horizontal line, and the center of each indifference circle is on this line as well. If one concentrated on charisma (where more is better) rather than age, then only the lower hemisphere would be relevant. The candidate strictly closest to M is labeled x*.

Fig. 1A

Every point that is preferred by a majority to x* is preferred by M: x* is to the right of M and to the left of R Note: any point preferred by L and R to x* must also be preferred by M to x*. If x* is the closest candidate to M, then there is no other candidate preferred by a majority of voters

All three voter-indifference curves are drawn through this point. For visual clarity, M's indifference curve is drawn with a heavier line. In Figure 1A, x*'s policy position is between the preferred positions of M and R; in Figure 1B, x*'s policy position is to the right of the preferred positions of L, M and R. It is readily apparent that the only set of points preferred by both L and R to x* are also preferred by M to x*. But we have assumed that x* is the closest candidate to M's most preferred position. So x* is the candidate that is preferred by a majority of voters to all other candidates.

Fig. 1B

Every point that is preferred by a majority to x* is preferred by M: x* is to the right of L,M and R. Note: any point preferred by L and R to x* must also be preferred by M to x*. If x* is the closest candidate to M, then there is no other candidate preferred by a majority of voters.

While the main focus of this article is on age of appointment being used as a strategic variable in exercising control, the model presented contributes to election theory qua theory. Groseclose has shown that with one issue dimension and one valence dimension, there are no pure-strategy equilibria when candidates care only about winning.Footnote ¹⁸ The intuition for the lack of equilibrium in his case is that the higher-valence candidate will try to choose a position similar to the lower-valence candidate, thereby getting all the votes. At the same time, the lower-valence candidate will want to distance himself from the higher-valence candidate in order to get some votes from those who gain sufficiently from the issue to compensate for the inferior valence. So there is no pure-strategy equilibrium, as the high-valence candidate chases the low-valence candidate, who runs away from the high-valence candidate. In contrast, in my model, candidates care about policy, and promises to act differently are not credible (except to a limited degree, as outlined in the model). So there is no chase. In Online Appendix C, I outline how the model can be extended to the multiple-issue case. This too is a contribution to theory. The standard way of handling a multidimensional issue space is to assume that voters vote probabilistically (that is, they sometimes vote for the candidate that makes them worse off). But in the n-dimensional version of the model presented here, there is an equilibrium even when voters always vote for their preferred candidate because credibility limits each candidate to one point.

There are two natural extensions to the basic framework: (1) allowing for a more complex relationship between age and utility, and (2) allowing different cohorts to have different optimal age preferences. The first is easily taken care of. Let f(a) be a positive monotonic transformation of a. Then one can simply substitute $$${{[f(a)\,{\rm{ - }}\,f(\hat{a})]}^2} $$$ for $$${{(\hat{a}\,{\rm{ - }}\,{{a}_c})}^2} $$$ . The vertical axis is now measured in differences between f(a) and $$$f(\hat{a})$$$ instead of differences between a and $$$\hat{a}$$$ . Proposition 1 still holds under these more general conditions.

The second extension requires a bit more work. Assume k cohorts of n voters, where each cohort lives for k generations (equivalently, k terms of office). Both n and k are odd. Hence, at the end of every generation, one cohort dies and a new one replaces it. Assume further that the median voter in each cohort has the same preference; otherwise, the distribution of policy preferences may differ from one cohort to another. Finally, allow different cohorts to have different age preferences that are monotonically increasing or decreasing according to the age of the cohort, although within-cohort preferences are the same regarding age preference.

For the proof, see Online Appendix B.

I now turn to the comparative static results. Assume initially that the leader is not just a figurehead (that is, B is large).

There are several strategic reasons for this proposition. Candidates elected for life compromise less on their policy preferences than candidates elected for a term who would like to be re-elected and therefore compromise more on their preferences. That is, candidates elected for life have larger w values, and therefore their policies are likely to be farther away from the median voter. This increases the potential for conflict between the leader and the voting body. At the margins, there is a tug of war between the electing body and the leader over control. The longer the leader is in power, the more power the leader amasses and the less influence the electing body has on the margin. An important method for the electing body to curb the leader's power is to elect the person when he is old and therefore less dynamic, and with less time to accumulate power that distorts policy from the median. Other methods of controlling the elected-for-life leader, such as writing pre-election contracts, will be covered when I discuss particular religions.

On the other side of the equation, when candidates are elected for one term at a time, an option value arises. This option value is greater for younger candidates. The logic is as follows: once elected, the uncertainty about the candidate's preferences is resolved. If the median voter is pleased with the outcome, then the leader is likely to be re-elected. This option is more valuable when the leaders are younger and more likely to live long enough to serve several terms in office. Therefore, other things being more or less equal, younger candidates will be preferred over older candidates when neither candidate has been elected to office before. This option value is missing for leaders elected for life. It is true that being re-elected increases the leader's power, but in order to be re-elected, the leader's choices must be reasonably aligned with the electorate. Therefore, the leader would have used his power to pursue actions that are in line with the median voter's preferences.

Of course, other things need not be equal – a younger candidate for life leadership may have preferences closer to those of the electorate than an older candidate would. But over the course of many elections, we would expect those who are elected for life to be older than those who are not.

The formal election model is embedded in a more general informal ‘model’ in which similar issues arise. Candidates for leadership do not simply appear. They typically have revealed some of their views beforehand. When mistakes are irreversible, as is the case when a leader is elected for life, there is a need for a better and more time-consuming pre-election screening process that reduces the uncertainty (ɛ) about each candidate's preference to a reasonable level and eliminates people whose preferences are far from the median. More screening means that the electorate will have more information about the candidate's preferences than otherwise, but this takes more time. As a consequence, candidates will be older in religions that elect for life. This reinforces the results of Proposition 3.

Recall that the leader has a similar preference structure to (but potentially different preferences from) those held by voters. That is, the leader also values policy and charisma. Choosing a younger person as the next leader gives the present leader influence over the church long after his death. The alternative is for the leader to choose an older person as a successor, who will then choose an older person as a successor, etc. But this method tends to be more risky than choosing a younger leader in the first place, as each successor is only an imperfect agent of the previous leader. When a ‘leader for life’ chooses a successor, there is a random walk of preferences that is exacerbated by choosing older leaders who live for fewer terms, thereby creating more random walks and more risk.

To illustrate, suppose that in each period t, a leader chooses a person who is old and will only lead in the next period, t + 1. To further simplify, suppose that the leader in time t with the most-preferred position $$${{\hat{x}}_t}$$$ can choose a new leader in time t + 1 with expected preference $$$E({{\hat{x}}_{t\, + 1}})\, = \,{{\hat{x}}_t}$$$ . Then a leader with the most-preferred position $$${{\hat{x}}_t}$$$ will have a Q chance of choosing a leader with a most-preferred position $$${{\hat{x}}_{t\, + \,1}}\, = \,{{\hat{x}}_t}$$$ , a 0.5[1 − Q] chance of choosing a leader with a most-preferred position $$${{\hat{x}}_{t\, + \,1}}\, = \,{{\hat{x}}_t}\, + \,{\epsilon}$$$ and a 0.5[1 − Q] chance of choosing a leader with a most-preferred position $$${{\hat{x}}_{t\, + \,1}}\, = \,{{\hat{x}}_t}\,{\rm{ - }}\,{\epsilon}$$$ .

Under this scenario, the present leader knows that there will be a random walk of policy with $$${k}-{\rm{1}}$$$ steps over $$${k}-{\rm{1}}$$$ generations. Because utility is a concave function of the difference between the present leader's preferred position and the future leader's preferred position, the present leader is risk averse. With regard to policy, a less risky strategy for the existing leader would be to choose a new leader who is only one generation old. Then the random walk takes place only once over the $$${k}-{\rm{1}}$$$ generations – when the young leader is appointed and rules for a long period. In contrast, when there is an election, the assembly is still intact after the old leader dies, so the problem of the random walk disappears over time as it restarts from the most-preferred candidate of the median voter in the assembly, which changes more slowly (if at all). Appointing a younger leader means that the leader will garner more power over time. This is desirable from the old leader's perspective, but, as argued earlier, when the assembly appoints a leader, it prefers one who is weaker vis-à-vis the assembly, which requires appointing an older leader. There will be greater uncertainty about a younger person's preferences regardless of who is appointing the new leader, but as argued above, there are benefits that accrue to the leader that do not accrue to the assembly when appointing a younger leader.

It is reasonable to believe that the correlation of a leader's preferences with his/her adult child's preferences is likely to be positive, even if not perfectly so. This means that in comparison to appointing an unrelated successor, there is less risk in appointing for life the leader's adult offspring while they are still relatively young. Coupled with other added benefits of hereditary succession, it is not surprising that offspring are often chosen as successors, even if there is no such explicit rule. When one's offspring step into the leadership role, grooming for the leadership role can start at a very early age. At the same time, screening is limited to the leader's children, which allows for a relatively rapid elimination of the weaker candidates for leader. This reinforces the results of Proposition 4. Particularly when life expectancy was short, such hereditary succession resulted in the appointment of very young leaders.

When leaders cannot be removed from office, the leadership position can sometimes be stripped of power. For example, the devolution of power of the English monarchy and, more germane to the topic of this article, the Protestant Reformation's denial of the pope's power. In Protestant religions, power is considerably devolved so that the leader is much weaker. In extreme cases, the leader is only a figurehead and the voters’ utility functions no longer depend on the leader's ideological position, x_c ; that is, B = 0. In such cases, the most important characteristics are the leader's age and charisma, K_c . The leader is the human embodiment of the abstract religious principles. This human connection between the leader and the followers becomes stronger the longer the leader has been in this role. We thus have the following proposition:

Because the leader's power is weaker and can be controlled, the risks of electing a younger person are greatly reduced. On the other side of the equation, the longer the leader has been the embodiment of the church, the stronger the connection the followers feel. Hence, symbolic leaders will tend to be appointed at a younger age.

All of the above deal with succession in existing religions. But founders of new religions do not have to conform with the existing power structure, because there is none. In particular, they do not have to travel up the steps of the pyramid. Instead, they compete with other religions to gain members. Therefore it should not be surprising that founders will tend to be younger when first recognized as the leader of a new religion than subsequent leaders in the established religion.Footnote ²⁰

The Structure of Churches and the Age of Their Leaders When First Appointed

The theory developed in the previous section will now be used to provide insight into particular religions. The starting line for the data is the closest appointment to 1776, as that was about the time that the Episcopal Church of the United States split from the Church of England; since that time, a number of American churches have been created.Footnote ²¹ The relevant data is found in Table 1, which shows that the pair-wise comparisons between religions have statistically significant differences in the age of (first) appointment to the leadership. The data behind Table 1 is found online in Tables 1A and 1B.

Table 1

Differences in Age and T-values

Note: age and age differential are in bold.

Politics More Generally

I have provided a convenient metric, age, which allows for comparisons both over time and across religions. While I have focused on politics within religious institutions, the same analysis should hold for politics more generally. An insightful comparison is between the doges of Venice and Genoa.Footnote ⁴⁹ The former were appointed for life, while the latter (after 1527) were appointed for two-year terms. According to the theory presented here, the Venice doges should have been appointed at an older age. Unfortunately, for Genoa, I have only been able to obtain birthdates for those who became doges after 1700. For Venice, a consistent series starts a century earlier. Napoleon ended the reign of both doges in 1797. For Genoa the average age at appointment was 60.4, and for Venice since 1700 the average age of appointment was 66.9; the difference between the mean ages is three standard deviations, making these results highly statistically significant. For Venice since 1600, the average age of appointment was 69.2; the difference between the mean ages of appointment between Genoa and the extended series on Venice is 5.9 standard deviations. The average reign for Venice was 6.2 years since 1600 and 8.8 years since 1700. So here, too the age of appointment was used to rein in the power of the leader.

Extending the analysis to succession in non-democratic governments is not an easy task, as the power relationships may not at all resemble the stated legal relationships. Can the leader really be removed by the parliament? Is the real power in the hands of the presidency or the generals? And has the structure of power changed dramatically over short periods of time, despite there being no change in the legal basis for the allocation of power?

With these caveats in mind, let us look at China. The Politburo Standing Committee (PSC) of the Communist Party of China is the center of power in the country.Footnote ⁵⁰ The size of the committee has varied between five and eleven members, and like the Jehovah's Witnesses’ governing body, many of the internal workings are mysterious. However, since Deng Xiaoping became leader, it is clear that the former leaders (and, to a lesser extent, the former members) are very influential in choosing the new leader and the other members of the PSC.Footnote ⁵¹

The term ‘paramount leader’ is often employed by commentators because the true leader is not necessarily connected to a particular office. The paramount leader has typically been the chairman of the Communist Party of China (and later, the general secretary of the Communist Party) and sometimes the chairman of the People's Republic of China (later, the president of the People's Republic of China) and the chair of China's Central Military Commission. However, Deng Xiaoping never held the first two of these offices, and he was considered the paramount leader even during 1987–92, when he was no longer a member of the PSC and Jiang Zemin was the general secretary of the Communist Party of China. Indeed, Deng continued to exert influence after his retirement from all government positions in 1992 until his death in 1997. For reference, the paramount leaders have been Mao Zedong, Hua Guofeng, Deng Xiaoping, Jiang Zemin, Hu Jintao and now Xi Jinping. The power of the leader has devolved since Deng so that the party leader is now closer to being ‘first among equals’.

As noted earlier, the founders of a religion tend to be young. The same holds true for the Chinese Communist Party. Even after engaging in a long war of ‘liberation’, the average member age at the start of the ten-year de facto PSC beginning in 1945 was forty-nine: Mao Zedong was fifty-two and Zhou Enlai and Liu Shaoqi were both forty-seven. However, Mao held onto power in his role as chairman of the Communist Party of China until his death at age eighty-two in 1976. Zhou Enlai also died in 1976 at age seventy-eight (Liu Shaoqi had been expelled from the party in 1968 and died shortly thereafter).

By the end of 1978, Deng Xiaoping had become the de facto leader of China until his retirement at age eighty-eight in 1992. There were three areas, all in reaction to Mao, that needed to be resolved by Deng and those who followed him: (1) to implement a more pragmatic economic policy and have a more stable political climate, (2) to prevent a continuation of leadership by gerontocracy and (3) to prevent another Mao from arising. I now consider each in turn.

1. (1) After the trauma of Mao's chaotic and ideology-driven policies, Deng shifted to a more pragmatic economic approach by opening China to foreign investment, the global market and limited private ownership. These policies continue today. The desire for a more stable political climate and the dominance of the Communist Party also resulted in the military suppression of the protests in Tiananmen Square and the continued control of political information.

2. (2) With the exception of Hua GuoFeng's 2-year rule, between 1963 and 1992 China's paramount leaders have been in their seventies and eighties, including Deng. This long rule by the very old was certainly suboptimal. As early as 1982, Deng argued for the mandatory retirement of older state (but not party) leaders.Footnote ⁵² In 1997, Jiang Zemin officially instituted forced retirement at age seventy so that all members who had reached that age, except Jiang himself, were ineligible to serve on the fifteenth PSC. In 2002, China instituted a rule requiring the retirement of all Politburo members after they reached the age of sixty-eight.Footnote ⁵³ Specifically, a person could not be appointed or reappointed to the next five-year term of the PSC if he would reach age sixty-eight before the next party congress.

3. (3) Mao is the horrible history lesson for the Chinese Communist Party. To prevent the rise of another leader as powerful as Mao (or even as powerful as Deng Xiaoping), other means had to be devised. As argued earlier, a major method of reducing the power of the autocrat is to reduce his time in the leadership position, thereby reducing his ability to accumulate more power. Since Mao's time, only one leader, Jiang Zemin, has been general secretary of the Communist Party of China for more than ten years (Jiang was head for thirteen years). And even if we consider paramount leaders, only Deng and Jiang served more than ten years (nearly fourteen and twelve years, respectively).Footnote ⁵⁴ Now that there is a retirement age for party leaders, the accumulation of power is curtailed by appointing leaders who are too old to serve more than two terms (because they will have reached retirement age before a possible third term) and by appointing most of the other PSC members when they are too old to serve more than one or two terms. I will now go into greater detail.

I start with those who were appointed leader after 1997, when the retirement age became official policy. I have already mentioned Jiang Zemin. Hu Jintao reached sixty-eight before the end of his two five-year terms and therefore could not be on the PSC, let alone be head, for a third term starting in 2012. The same will hold true for Xi Jinping, as he will reach sixty-eight before 2022. Next looking at the members of the PSC (including the leaders), there have been twenty-five members from 1997 until now. Of those, only four could have served more than ten years before being forced to retire. And of those, two were Hu Jintao and Xi Jinping. It makes sense that they served on the PSC before they became leaders, because in this way they could serve as apprentices. At least half of the rest of the members of the PSC have not been eligible to serve more than one term.

The youngest member to be appointed to the PSC since Mao is Hu Jintao, who was appointed at age forty-nine in 1992. As argued earlier, when a leader anoints his successor, the leader wants his successor to be in power as long as possible. Deng Xiaoping and Jiang Zemin appointed Hu Jintao to the PSC in 1992. In retrospect it was clear that the plan in 1992 was for Jiang to be in charge until 2002 and Hu Jintao to be in power from 2002 until 2012.Footnote ⁵⁵ In this case, Deng not only chose his successor, Jiang Zemin, but also his successor's successor, Hu Jintao. Hu Jintao was not as powerful as Deng was, but during his last five-year term there were only two Politburo members who could be in power for at least ten years after he retired – Xi Jinping and Li Keqiang. The latter could be in power for five more years after Xi Jinping retires in 2022 – or longer if Xi dies before then and Li is his replacement.Footnote ⁵⁶ Essentially, the fear of creating another Mao means that the leadership position is held for ten years. Given that constraint, the present leader influences future leaders by choosing a member of the PSC who is young enough to take office after the first successor has to retire. The more powerful the about-to-retire present leader is, the longer the shadow he can cast on the future. In any event, the present leader and his fellow members of the PSC cast a shadow on the future as they choose their immediate successors.Footnote ⁵⁷ And as already noted, the immediate successors are young enough to remain in power for ten years, but too old to remain in power for longer.

Concluding Remarks

I have shown how different institutional arrangements affect the choice of a leader. One might argue that a more fundamental question is why the particular institutional arrangement was chosen in the first place. But this ‘more fundamental’ question cannot be fully answered until one knows the effects of various institutional arrangements. That is, without knowing the effects of different institutional arrangements, one cannot explain why a particular arrangement was chosen. While it is conceivable that institutions do not matter, it has been shown that when it comes to the age of the leader when first chosen, they do matter.

Some religions that are relatively similar in religious doctrine have significantly different power arrangements, while other religions that differ greatly in religious doctrine have similar methods of appointing leaders. This variation means that religious institutions are a natural laboratory in which to study elections and the politics of power.

Some religions limit the power of the church leader by placing policy in the hands of a synod and electing the leader for a set term. In other religions, the power of the church leader is less constrained because there is no legislative body and the appointment is for life. In such cases, the main methods of control are longer screening periods and electing a leader who will be in power for a relatively short time. Today, the main method of accomplishing these objectives is to elect older leaders; at other times and places, murder was another option.

In contrast, the leader's main way of exerting influence beyond the grave is to directly choose the next leader (or if he cannot directly choose the next leader, to choose the electors who will choose the next leader). In the former case, this leads to younger leaders being chosen; in the latter case, leaders tend to ‘stack the deck’ of electors by increasing the size of the electoral body and possibly eliminating some of the electors who were chosen by previous leaders.