Selection and Redistribution in the Irish Tontines of 1773, 1775, and 1777

• Date: 01 September 2020
• Published date: 01 September 2020
• DOI: http://doi.org/10.1111/jori.12274
• Author: Yikang Li,Casey Rothschild

©2019 The Journal of Risk and Insurance. Vol.XX, No. XX, 1–32 (2019).

DOI: 10.1111/jori.12274

Selection and Redistribution in the Irish Tontines of

1773, 1775, and 1777

Yikang Li

Casey Rothschild

Abstract

We construct and analyze a new data set on the mortality experience of the

nominees of the 1773, 1775, and 1777 Irish tontines. The active participation

of Genevan speculators in these Irish tontines has been well documented.

We use our new data to quantify both the extent to which these nominees

werelonger-lived and the financial consequences of their enhanced longevity.

The Genevan nominees were indeed notably longer-lived than non-Genevan

nominees—particularly so for the 50 nominees selected by a Genevan in-

vestment syndicate. Their enhanced longevity had only trivial financial con-

sequences for the issuer, but it led to significant redistribution from non-

Genevan to Genevan investors. We highlight the implications of this across-

group distributional risk for modern proposals to introduce tontine-like ele-

ments into modern retirement pensions.

Introduction

Tontines—also known as “annuities with benefit of survivorship”—have a long and

storied history. They were invented 1650s, used extensively in the late 17th and 18th

centuries as a source of public funds by Britain and France (Weir, 1989), and were

popular consumer products in the U.S. insurance market in the late 19th century

until an adverse legal decision against tontine providers in the early 20th century

effectively killed them off (Ransom and Sutch, 1987). Recently, a number of scholars

and policymakers have argued for resurrecting tontines, on the grounds that they can

play a useful role in financing retirement.

This article makes two complementary contributions to the economic literature on

tontines. The first contribution is historical: we construct and analyze a new data set

Casey Rothschild is at Wellesley College. Yikang Li was a student at Wellesley College while

completing this research;she does not currently have any academic affiliation. Rothschild can be

contacted via email: crothsch@wellesley.edu.Wethank Leticia Arroyo-Abad, Martin Boyer, Eric

Hilt, Moshe Milevsky,Kyung Park, and seminar participants at Wellesley College, Middlebury

College, ETH Zurich, Amherst College, and St. John’s University for useful comments; we

thank Crysti Wang,Beryl Larson, and Yining Li for excellent research assistance; and we thank

Wellesley College’s Jerome A. Schiff Fellowship for financial support for Yikang while she

worked on an early version of this article for her Senior Honors Thesis. Finally, we thank the

editor and two anonymous referees for insightful and constructive feedback.

1

719

719

Vol. 87, No. 3, 719–750 (2020).

2The Journal of Risk and Insurance

on the mortality experience of the nominees of the 1773, 1775, and 1777 Irish tontines,

and we use this data set to quantify the differential longevity of participants who

were selected by known Genevan speculators, and the financial consequences thereof.

Our results indicate potentially significant redistributive consequences of speculation

within the Irish tontines. Our second contribution is to observe that these historical

results also speak to a potentially important but heretofore overlooked risk associated

with using tontines in finance modern retirement systems: the potential for nontrivial

across-subgroup distributional risks.

The recent renaissance of interest in tontines among economists and policy advocates

stems from a potential advantage they have relative to traditional annuity-based re-

tirement systems. Economists have long recognized the value that traditional defined

benefit pensions and life annuities provide in insuring retirees against the idiosyn-

cratic risk of outliving their resources(viz. Yaari, 1965; Davidoff, Brown, and Diamond,

2005), and pensions and life annuities also protect retirees from aggregate longevity

risk by effectively offloading this risk onto annuity providers. Insofar as providers

have limited ability to hedge their exposure to this aggregate risk (Blake, 1999), how-

ever, they may be exposed to a higher probability of default risk (Forman and Sabin,

2014) and may be forced to charge higher annuity prices (Brown and Orszag, 2006).

Tontines can potentially alleviate these concerns.

In a prototypical tontine, each of the Nmembers of the tontine pool “buys in” with

a lump sum payment R/N, and the tontine provider promises a periodic interest or

“dividend” payment R to the pool, to be divided up evening among the members

who are still alive at the time the payment is due. As members of the tontine pool die

off, the periodic payments to each surviving member grows, but the total payment

due from the provider remains constant at R. A number of scholars and practitioners

have noted that this constant-total-payment feature greatly mitigates providers’expo-

sure to aggregate longevity risk (relative to traditional life annuities) and can thereby

potentially improve pricing and limit default risk. These include Piggott, Valdez,and

Detzel (2005), Rotemberg (2009), Forman and Sabin (2014), Newfield (2014), Milevsky

and Salisbury (2015), Wettstein (2018), Macdonald (2018), and Milevsky’s (2015)

book King William’s Tontine: Why the Retirement Annuity of the Future Should Resemble

Its Past.

The aggregate-longevity risk-hedging benefits of tontines are most easily seen from

the point of view of issuer. Consider, for example, a cohort of newly retired 65-year-

old individuals “locking in” a pension for the remainder of their lives, and consider

a government with a fixed sum of resources Rset aside to fund these pension pay-

ments. Suppose first that, as in the U.S. Social Security system, the government pays

a pension in the form of a life annuity: it promises each individuals a fixed annual

income for as long as that individual remains alive, with the magnitude of the annual

income calibrated so that the expected discounted cost equals R. In this case, the gov-

ernment is fully exposed to aggregate mortality risk: if the cohort turns out to have

significantly lower mortality than expected, then the provider will see their annual

pension payment outlays rise more or less immediately, and the cost of providing the

pensions will greatly exceed R.

2The Journal of Risk and Insurance

720

Selection in the Irish Tontines 3

Contrast that with the case of a pure tontine, where the government earmarks a reg-

ular annual payment to the entire cohort of individuals, and, each year, divides that

annual payment up among the surviving members of that cohort. In this case, the

government is minimally exposed to aggregate mortality risk. If the cohort turns

out to be longer-lived than expected, the fixed annual payments due in early years

will, by construction, be completely unaffected from the point of view of the gov-

ernment. The only cost to the government will be due to the longer, far-future tail of

the payments—as the last-to-die member of the tontine pool will be longer-lived, in

expectation. By virtue of being in the far future, these extra payments will typically be

small in present value terms. Moreover,and importantly, most serious modern policy

proposals modify the prototypical tontine structure by prescribinga declining annual

income stream, calibrated so that the dividends per survivor remain approximately

constant over time under baseline mortality projections (so that, from the point of

view of a tontine participant, the expected income stream will closely resemble the

income stream provided by a standard life annuity). The present value of the risk to

the government from a positive longevity shock to the cohort is virtually nil under

such proposals.

The fact that tontines can effectively eliminate aggregate mortality risk from the point

of view of an annuity or pension provider such as a government does not necessar-

ily make them desirable, of course. Indeed, relative to a traditional pension annuity,

a tontine exposes the pension recipients—who may have less ability to bear it—to

greater aggregate longevity risk. This is well understood in the literature. Another

potential drawback to tontine schemes appears to be less well understood: shifting

from traditional annuities to tontine-based schemes may introduce or magnify re-

distributive risks caused by group-specific longevity shocks. For example, the small

mortality improvements experienced by lower earners relative to high earners in

the United States during the 20th century (National Academies, 2015) would poten-

tially have had significantly different across-class distributional consequences under

a tontine-based pension scheme as opposed to an annuity-based pension scheme like

the current U.S. Social Security system (see the “Modern-Day Implications” section

for an illustrative calculation).

We use the unique historical “experiment” providedby the Irish tontines to illustrate

and to quantify these potential distributional consequences. The Irish tontines have

been discussed in earlier literature, notably by Gautier (1951) and Jennings and Trout

(1983), but without the benefit of the micro-data we have now gathered from several

archival sources. We have a complete list of all lives nominated for these tontines,

their ages at the time of nomination, their domicile, and substantial but—because of

imperfect record keeping—incomplete information on their death dates. Despite the

partially missing death dates, our data set nevertheless allows us to make precise

computations of the mortality experience of different subgroups of participants in

these tontines and, for the first time, allows us to make precise estimates of the financial

returns to investments therein.

As discussed in Gautier (1951) and Jennings and Trout (1983), a Genevan banking

syndicate subscribed heavily to the Irish tontines. These bankers took advantage of the

Does National Flood Insurance Program Participation Induce Housing Development? 3

721