Pricing the Hand on the Switch: An Event Study of Long-Dated Deterrence Bond Yields and Continuity Premia around News of the Swordholder
Abstract
A risk premium on the Swordholder, meaning the extra return that holders of long claims demand because the switch rests in one person's hand, would matter to debt managers, underwriters and the office running the succession. We revisit the proposal that such a premium exists and moves with news. Using the trade archive of the Joint Exchange Clearing Secretariat and the premium schedules of a Cape Town continuity pool, we study 18 dated releases between Deterrence Era, Year 40 and Year 61, namely 11 health reports, 5 tenure milestones and 2 announcements about the succession. Abnormal changes in a constant-maturity 25-year yield use a market-model slope estimated from intraday trade times by a Fourier-type estimator. Over the two sessions from release, the mean abnormal change was +0.27 basis points (95% CI -1.50 to 2.04, p = 0.75), no type of release differed from zero after step-down family-wise correction, and the events were no more variable than 180 placebo windows (variance ratio 0.87). The premium on ten-year continuity cover changed by +0.20 per cent (95% CI -1.22 to 1.62, n = 15). The data do not support a Swordholder risk premium that responds to news, and a premium constant across news days lies outside what an event study can see. With these samples only a mean response above 2.5 basis points could have been found reliably, and the succession itself, with two events, is untested. Windows, estimators and scripts are released.
1. Introduction
On the second morning of the fourth month of Deterrence Era, Year 52 (DE 52), the Medical Office of the UN Planetary Defense Council (PDC) posted its periodic report on the health of the Swordholder. The report ran to two pages and recorded one change in the schedule of routine examinations (UN Planetary Defense Council, Medical Office, 41–60 DE). The Joint Exchange Clearing Secretariat opened its bond list on the same morning, and its clerks stamped each trade in the long Deterrence Reconstruction Bonds with the minute of clearing (Joint Exchange Clearing Secretariat, 38–61 DE). By the close of the following session the 25-year yield had moved by 0.9 basis points after allowing for the short end, well inside the 3.8 basis points that is the standard deviation of ordinary two-session windows.
A risk premium on the Swordholder, meaning the extra yield or extra insurance charge which holders of long claims demand because the switch rests in one person's hand, is of practical use to the PDC Debt Office when it sets the coupon on new long-dated issues, to underwriters who quote continuity cover, and to the Succession Office when it judges how a public process will be received. It is hard to measure because the long bond list trades thinly and at irregular times, and because the Council releases news about the Swordholder on so sparse a calendar that few events fall inside any sample. Concretely, the archive holds 18 dated releases over some 5,800 sessions, against daily yield changes of a few basis points that have other causes.
We revisit the proposal that the long end of the bond list carries such a premium and that it moves with news of the Swordholder (Vorster-Linde, 59 DE), which estimates a level and predicts, without testing it, that the premium reacts to news, and we take the opposite side. Long yields and continuity premia did not move on health reports, on tenure milestones or on the two announcements about the succession, and we report that result flatly, together with the smallest response the design could have found.
A public process, under rules issued in DE 60, will name the next Swordholder (United Nations Planetary Defense Council, Secretariat, 60 DE), and the Succession Office has announced the candidates (United Nations Planetary Defense Council, Succession Office, 61 DE). We write in Deterrence Era, Year 61 (DE 61). The succession is pending, and every source predates the end of the first half of that year. The study of Sarkissian-Low (61 DE) scores the credibility of the incumbent and the announced candidates from declassified text, and we do not use those scores.
2. Methods
The bond data are the time-stamped trades and quotes of the Deterrence Reconstruction Bond list, kept by the Secretariat from DE 38 (Joint Exchange Clearing Secretariat, 38–61 DE). We construct two yields for each session. The long yield is a constant-maturity 25-year yield interpolated across the bonds that traded, and the short yield is the same construction at two years. Long bonds trade with a median gap of 14.0 minutes, so each yield is read at the Secretariat's clearing close by time-weighted interpolation between the trades that bracket it, not at the last price, which carries a stale quote into the close (Ngubane-Reyes, 38 DE).
Insurance premia come from the schedules of ten-year continuity cover, which we call the Pool, a label of ours for the Cape Town underwriters whose quoted rates the archive preserves from DE 44 (Cape Town Continuity Underwriters, 44–61 DE). The cover pays on loss of insured property if the Dark Forest Deterrence fails. Quotes are revised at irregular dates, on average every 3.4 sessions, and the premium series is the logarithm of the quoted annual rate per unit of sum insured, carried onto the same clearing clock.
Events are the 18 releases of three kinds. Eleven are periodic health reports of the Medical Office, of which four state a change in the Swordholder's condition in their own summary line and seven do not, a coding we fixed before any outcome was computed. Five are the notices that mark the fifth-year milestones of the tenure that fall within the bond archive, at DE 40, 45, 50, 55 and 60 (United Nations Planetary Defense Council, Secretariat, 40–60 DE). Two are announcements about the succession, namely the rules for the public selection in DE 60 and the filing of the candidacies in DE 61. The event window is the session of release and the session after it, and as a check on anticipation and slow absorption we also use a five-session window that opens one session before the release and closes three sessions after it.
Abnormal change in the long yield is its change in the window minus the change that a market model predicts from the short yield. The model is fitted on the 120 sessions that end ten sessions before the window, with the sessions of any other event window removed. Its slope is the ratio of the covariance between the two yield changes to the variance of the short one. We estimate both from the trade-by-trade price paths of the long and short buckets over those sessions, with a Fourier-type estimator that works on raw trade times, and then apply the slope to the session changes. A regression of closing changes would inherit the stale anchors of the thin long bucket and bias the slope towards zero (Mbatha-Corrin, 31 DE; Dlamini-Aoki, 22 DE; Rautenbach-Zhou, 56 DE). For the Pool no series is quoted on the same clock to serve as a market, so abnormal change in the premium is the change in the window minus the mean of the estimation window.
The unit of inference is the event. We test the mean abnormal change by a t statistic with n - 1 degrees of freedom, because every sample is below 30, with two-sided tests throughout. The family has four tests (all 18 events, and each of the three kinds that has more than two events), corrected by a step-down family-wise procedure, which multiplies the k-th smallest of the four p-values by 5 - k and keeps the running maximum. As a second test we divide each event's abnormal change by its own forecast standard error, which is approximately the residual standard deviation of the estimation window times the square root of 2 + 4/120, and sum the results. Dispersion is compared by an F test of the variance of event sums against that of 180 placebo windows, which are two-session windows starting at every 32nd session from the 131st session of the list, with windows within five sessions of an event excluded, each scored as an abnormal change with its own 120-session estimation window, and for the Pool the same grid from the first quoted session, DE 44. We state the smallest mean response that the design finds with 80% power, which is the sum of the 97.5% and 80% points of t on n - 1 degrees of freedom times the standard error of the mean.
3. Results
Over the two sessions from release, the mean abnormal change in the long yield was +0.27 basis points (95% CI -1.50 to 2.04; t = 0.32 on 17 degrees of freedom, p = 0.75). Ten of the 18 sums were positive and eight negative (exact sign test p = 0.81). The mean forecast standard error of an event was 3.74 basis points, and the largest standardised response in either direction was 1.67, so no single event stood apart from its own noise. The sum of the standardised responses gives z = 0.36 (p = 0.72).
Table 1 gives the result by kind of release. Routine health reports averaged -0.14 basis points. Reports that stated a change in condition averaged +3.05, the largest of the three means and the one sign that a premium would predict, with an interval from -3.25 to +9.35 on three degrees of freedom. Milestone notices averaged -1.74. None of the three tests survives correction, since the smallest family-adjusted p-value is 0.89.
Of the two succession events, the rules for the public selection in DE 60 were followed by an abnormal change of +5.2 basis points and the filing of the candidacies in DE 61 by -2.8. With two observations of opposite sign we can report them and do no more.
4. Checks, Dispersion and Insurance
The five-session window gives the same answer. The mean abnormal change from one session before to three after release was +0.40 basis points (95% CI -2.26 to 3.06, p = 0.75), with a standard deviation of 5.4 basis points across the 18 sums, so neither anticipation nor slow absorption supplies a response that the shorter window missed.
Dispersion tells the same story. The standard deviation of the 18 event sums was 3.56 basis points, against 3.82 for the 180 placebo windows, a variance ratio of 0.87 (F on 17 and 179 degrees of freedom, two-sided p = 0.77). A premium that made the yield jump on news would have raised the ratio above one.
Continuity premia give the same picture. Fifteen of the 18 events fall on or after DE 44, made up of 5 routine reports, 4 reports stating a change, 4 milestone notices and 2 succession events. The mean abnormal change in the premium was +0.20 per cent (95% CI -1.22 to 1.62; t = 0.30 on 14 degrees of freedom, p = 0.77). By kind the means were +0.36 per cent for routine reports, -0.65 for reports stating a change and +0.30 for milestones, each with an interval that holds zero (Table 2), and the two succession events gave +4.6 per cent (DE 60) and -2.0 (DE 61). The event sums have a standard deviation of 2.6 per cent, against 2.7 for ordinary two-session windows of the Pool; the report of DE 52, one of the five routine reports with a Pool quote, rose by 3.4 per cent in the cost of cover.
Table 2 sets the main bounds side by side. With 18 events and a standard error of 0.84 basis points, the smallest mean response found with 80% power is 2.49 basis points. Concretely, read as a spread over a bond that loses its full value on a failure of the deterrent, the upper limit of 2.04 basis points corresponds to a rise of about 0.02 percentage points in the annual hazard that the market attaches to such a failure, and a smaller loss given failure would widen that bound by the inverse of the loss share. For the Pool, 15 events and a standard error of 0.66 per cent give a smallest detectable response of 2.00 per cent.
5. Discussion
The data do not support a Swordholder risk premium that responds to news. Long Deterrence Reconstruction Bond yields did not move on health reports, tenure milestones or the two succession announcements, the premium on ten-year continuity cover did not move either, and the events were no more variable than ordinary windows.
We write within the Luo-Ji / Ye framework, in which deterrence holds only while the switch is believed to be live, and the market question is whether traders priced changes in that belief, which on this evidence they did not. We reaffirm the position of Ngubane-Reyes (38 DE) that a thin list must be read at a time-weighted close, and we add that on this list the same care returns a flat answer.
The proposal of Vorster-Linde (59 DE) puts a premium for the Swordholder's tenure into long spreads and expects it to react to news, and the reaction is what we find absent in 18 releases. A premium that is constant across news days would not appear in an event study at all, so we do not claim that the spread contains none.
For the Debt Office, a loading of a few basis points on long coupons for days of Swordholder news adds a term which the archive does not show, since the interval excludes a mean response above 2.04 basis points. Nor does it show that markets regard the deterrent as safe, only that these releases moved neither the long end nor the cost of cover.
Nothing here bears on the credibility scores of Sarkissian-Low (61 DE), which concern what a text conveys to a hostile reader, and no Trisolaran assessment of the deterrence enters this paper.
6. Limitations
Eighteen events is a small sample, and the kinds are smaller still, with 7, 4, 5 and 2 events. The design could find a mean response of 2.49 basis points with 80% power and could not find a smaller one, and the subgroup intervals exclude only large effects. The succession itself is the case of most interest, and with two events, one of them filed in the first half of DE 61, it is untested.
An event study identifies a response to news and not a premium that holds steady.
Both series are constructs of ours. The 25-year yield is interpolated from a thin list, the Pool is our label for underwriters whose schedules reach the archive only from DE 44, and the premium is the logarithm of a quoted rate and not of a traded price. The 120-session window gives few session changes for fitting the slope of a thin bucket. A release we did not code, such as a Council statement on the transmitter network, may still have moved the bond list on a day which we count as quiet. Sophons could read each bulletin while it was drafted, not only at release, and we assume that this readership was constant across the archive and that no trader on the list learned a release's content before its date, an assumption which the five-session window, opening a session early, only partly checks.
The release holds the event calendar and its coding, the estimation and placebo windows, the Fourier-type slope estimator, the interpolated yields, the Pool premium series, and the scripts behind Tables 1 and 2. The package is deposited with the Economic Reconstruction Office, whose replication archive admits accredited Council researchers, and a reader who reruns it can move the window, drop an event, or add a release we did not code, and see whether the null survives. The bond trades come from the same Secretariat file group as the earlier correlation study (Rautenbach-Zhou, 56 DE), which used its Crisis-Era ledgers and tick series, and the Fourier-type estimator is shared with it. The Pool schedules are new to this paper.
Return to the report of the second morning of DE 52. It was one of the seven routine reports and changed the examination schedule by one line. The long yield moved by 0.9 basis points after allowing for the short end and the cost of cover by 3.4 per cent over the same two sessions, each inside the ordinary spread of such windows (3.8 basis points and 2.7 per cent). The window, the estimator and the coded calendar that produced those two numbers are in the package, and anyone who recodes that report as a change in condition can see what moves.
References
- United Nations Planetary Defense Council, Medical Office (41–60 DE). Periodic reports of the standing panel on the health of the Swordholder, release series. PDC Declassified Holdings, New York, Series MH, reports 1–11.
- United Nations Planetary Defense Council, Secretariat (40–60 DE). Notices of the fifth-year milestones of the Swordholder's tenure. PDC Declassified Holdings, New York, Series CN, notices 1–5.
- United Nations Planetary Defense Council, Secretariat (60 DE). Rules for the public selection of the Swordholder. Bulletin of the UN Planetary Defense Council, declassified series, Bulletin 40, annex 1.
- United Nations Planetary Defense Council, Succession Office (61 DE). Candidacy files and public statements of the announced candidates for the succession. PDC Declassified Holdings, New York, Series SC, files 1–5.
- Joint Exchange Clearing Secretariat (38–61 DE). Trade and quote archive of the Deterrence Reconstruction Bond list. PDC Declassified Holdings, New York, file group 12, bond series.
- Cape Town Continuity Underwriters (44–61 DE). Quoted premium schedules for ten-year continuity cover. PDC Declassified Holdings, New York, file group 14, premium series.
- Dlamini-Aoki, P. (22 DE). Order splitting under presumed sophon surveillance and the decay of short-interval correlation. Deterrence Era Strategic Studies, 6, 41–66.
- Mbatha-Corrin, S. (31 DE). Fourier-type estimation of correlation under asynchronous trading in reconstituted markets. Proceedings of Applied Speculative Statistics, 14, 88–109.
- Ngubane-Reyes, K. (38 DE). Event windows in thinly traded bond lists, time-weighted closes against last prices. Deterrence Era Strategic Studies, 11, 45–73.
- Vorster-Linde, T. (59 DE). A tenure premium in long Deterrence Reconstruction Bond spreads, a level-based account. Deterrence Era Strategic Studies, 18, 9–37.
- Rautenbach-Zhou, L. (2026). Cross-Asset Correlation After the Doomsday Battle: A Comparison With the Great Ravine Regime Under Asynchronous Trading. Uncited Press.
- Sarkissian-Low, O. (2026). Credibility as a Forecastable Quantity: Scoring the Swordholder and the Announced Succession Candidates from Declassified Text. Uncited Press.
Open in Uncited Press →