Purpose
The companion paper argued that the near-absence of sustained public buffer stocks across the historical record is a governance failure rather than an information failure, and it offered a nine-item taxonomy of the mechanisms responsible. That argument has a defect its author should name rather than wait to have named: the taxonomy was assembled after the outcome was known, its items are mutually reinforcing enough that some subset will always be available, and no condition was stated in advance that would have counted against it. It is a plausible reading, not a tested one.
This paper does three things. It restates the rival explanations precisely enough that they can be wrong. It specifies eight research designs capable of separating them, each with a stated disconfirming condition. And it sets out the evidentiary rules governing use of the Genesis 41–47 material, which in the companion paper carried more argumentative weight than its warrant was made explicit for.
1. The Rival Hypotheses, Stated to Be Falsifiable
H1 — Technical sufficiency. Cumulative physical loss over a multi-year hold, combined with the labor and capital cost of construction and turning, was severe enough that a three-to-seven-year reserve was not worth building under premodern conditions. Absence is then rational and requires no governance explanation.
H2 — Institutional constraint. Multi-year holds were physically achievable at tolerable loss under available techniques, and were defeated by custodial, fiscal, and political mechanisms: raid of the reserve for other purposes, rent extraction in purchase and release, decay of administrative competence, and the temporal mismatch between cost and benefit.
H3 — Doctrinal displacement. A period-specific hypothesis: from the later eighteenth century, storage was dismantled because trade-based provisioning was believed superior, independent of both technique and custodial capacity.
H4 — Object misspecification. Reserves were absent because absolute shortfall was not the binding constraint; the operative failures were of exchange entitlement and distribution, and storage was correctly judged marginal.
These are not exclusive. H1 and H2 could each bind in different regions and periods, and the honest expected finding is a partition rather than a winner. What matters is that they generate divergent observable predictions, and they do, on one central asymmetry.
A technical constraint binds uniformly; an institutional constraint varies. If storage was defeated by grain physics, absence should be invariant across custodial arrangements holding technique constant, and should lift when technique improves. If storage was defeated by governance, absence should vary with custodial arrangements holding technique constant, and should not lift — or should perversely worsen — when technique improves. Every design below is a way of exploiting that asymmetry.
2. Eight Designs
D1 — Recovering the loss parameter
Question. What was cumulative grain loss over holds of one, three, and seven years under the storage techniques actually available in the relevant periods and climates?
Sources. Experimental archaeology on sealed pit storage, which has repeatedly demonstrated that anoxic underground storage produces losses far lower than intuition suggests once a germination layer seals the pit (Reynolds, 1974); comparative ethnographic and post-harvest work on non-mechanized storage in arid and semi-arid conditions (Gilman & Boxall, 1974; Boxall, 2001); and typological work distinguishing storage techniques by their physical regime (Sigaut, 1988). The parameter must be recovered separately by grain type, by hulled versus milled state, and by climate, since these differences are large enough to change the answer.
Discriminating prediction. Under H1, cumulative loss across a three-year hold should approach or exceed the shortfall being buffered, rendering the program self-defeating. Under H2, three-year losses in sealed or well-managed storage should be modest enough that a reserve retains most of its buffering value.
Disconfirming condition. If three-year cumulative loss in the best available premodern technique exceeds roughly half the stored quantity across the major grain types and climates, H2 becomes surplus explanation and should be withdrawn as the primary account for the pre-modern period.
D2 — The shrinkage-allowance series
Question. What loss rates did administrators themselves expect, as recorded in the accounting allowances written into granary and tax regulation?
This is the strongest available window into practitioner belief, because a stated allowance is a number an official was willing to be audited against. Qing granary regulation, Roman annona accounting, Ottoman provisioning records, and Egyptian temple and state granary documents all in principle carry such figures.
Discriminating prediction. Under H1, stated allowances should be high and target hold durations short, and the two should move together — as allowances fall with improved technique, durations should lengthen. Under H2, target durations should track political variables (war finance, dynastic security, recent famine memory) and show no reliable relationship to allowance rates.
Disconfirming condition for H2. A tight positive association between falling allowances and lengthening target durations across independent administrative traditions.
D3 — Matched pairs, technique held constant
Question. Within a single technological and climatic regime, does reserve persistence vary with custodial arrangement?
Design. Paired comparison of storage systems matched on crop, climate, era, and technique, differing on custodial insulation: whether the reserve had a dedicated funding stream, whether its assets were legally transferable to general revenue, and whether the custodian’s tenure was protected. Candidate pairings include northern Italian civic abundance offices against princely storage in comparable states; Qing provinces differing in governor tenure and audit intensity; and Ottoman capital provisioning against provincial arrangements.
Discriminating prediction. Under H2, persistence should vary sharply with insulation. Under H1, all systems in a technological regime should decay at similar rates regardless of custodial form.
Disconfirming condition for H2. No detectable association between custodial insulation and stock persistence across matched pairs.
D4 — The depletion-cause audit
This is the most direct test available and the most tractable, and it should be run first.
Question. For every documented public granary system that ceased to hold stock, what was the proximate cause of the stock’s disappearance?
Design. A coding exercise over the published record. Code each terminal event into pre-specified mutually exclusive categories: spoilage write-off; sale or transfer for general revenue; diversion to military supply; embezzlement or falsified inventory; exhaustion by legitimate crisis distribution without replenishment; statutory abolition on doctrinal grounds; and destruction by war or disaster. The coding protocol, category definitions, and predicted modal category must be written down before the coding begins, and a second coder should work a subsample independently.
Discriminating prediction. Under H1, the modal exit is spoilage write-off. Under H2, the modal exits are fiscal transfer, diversion, and abolition. Under H3, abolition should cluster in a datable window and be geographically concentrated in states exposed to the political-economy literature of the period (Kaplan, 1976).
Disconfirming condition for H2. Spoilage write-off as the modal terminal cause across regions and centuries.
D5 — The technology-shock test
Question. When storage technique demonstrably improved, did public reserves expand?
Design. Identify datable improvements — sealed masonry silos, systematic fumigation, moisture management, mechanical turning — and test whether public reserve holdings rose in the following decades in states adopting them.
The confound must be handled explicitly. Rail and steam shipping arrived in the same window and are a substitute for storage, not a complement: cheap long-distance freight lets a state buy its way out of a regional shortfall rather than store against it. Any test must separate states whose freight access improved from those whose storage technique improved without comparable transport gains, or the design measures the wrong thing.
Prima facie observation to be formalized. The nineteenth century saw storage technique improve while public granary systems were dismantled in several major states. If that pattern survives the transport control, it is a serious blow to H1 and a point for H3.
D6 — The break-even reconstruction
Question. At what loss rate and discount rate does a multi-year reserve stop being worth building?
Design. An explicit expected-value model taking as inputs the recurrence interval of climate-effective eruptions from the ice-core record (Sigl et al., 2015), the shortfall distribution conditional on such an event, the D1 loss parameter, construction and labor costs, a range of discount rates, and a valuation of the averted loss that includes production collapse through seed consumption, not consumption shortfall alone. The output is a sensitivity surface, not a point estimate.
Discriminating prediction. If the break-even loss rate lies far above the D1 observed rates across the plausible discount range, H1 fails on its own terms and can be retired for the periods covered. If the break-even lies below observed rates, H1 is sustained and the companion paper’s argument must be substantially rewritten.
This design is where H1 is most cleanly resolved, and it is fully executable from published parameters.
D7 — The non-state storage test
Question. Did private, communal, or religious bodies hold multi-year stocks where states did not?
Design. Compare persistence of monastic and temple granaries, tithe barns, lineage and community granaries, and merchant reserve holdings against state systems in the same period and region. The Chinese case is unusually well suited because state ever-normal granaries and community-managed granaries coexisted and are separately documented (Will & Wong, 1991).
Discriminating prediction. Grain physics does not distinguish owners. If multi-year holds persisted under non-state custody where state holds decayed, H1 is defeated for that setting and the constraint is located precisely in the custodial arrangement — which is the sharpest possible confirmation of H2.
Disconfirming condition for H2. Uniform absence of multi-year holds across all custodial types within a technological regime.
D8 — The negative-case register
Question. Which regimes possessed the technique, the fiscal capacity, an insulated custodial form, and recent famine memory, and still did not store?
Negative cases must be selected on a stated possibility principle rather than gathered opportunistically — the candidate set is regimes where storage was a genuine option, not every polity that lacked a granary (Mahoney & Goertz, 2004). This design exists to correct the companion paper’s Section 5, which selected its examples on persistence and therefore could only discover what persistent systems have in common. The unit of analysis should be adoption events, with persistence as the outcome, so that adoptions that failed enter the sample.
3. Pre-Registration Block
Stated before any of the above is executed.
The companion paper’s thesis will be treated as disconfirmed if D4 returns spoilage write-off as the modal terminal cause, or if D6 places the break-even loss rate below the D1 observed rates, or if D7 finds no multi-year holds under any custodial form within a technological regime.
It will be treated as partially disconfirmed and requiring redefinition if the results partition by period — for instance, if H1 holds for above-ground bulk storage in humid climates while H2 holds for sealed storage in arid ones. In that event the object of study is not “why regimes did not store” but “under what physical conditions storage was an available political choice,” which is a different and probably better question.
It will be treated as sustained only if D3 shows persistence varying with custodial insulation and D4 returns fiscal or doctrinal exits as modal and D6 places the break-even well above observed loss.
The expected outcome, stated honestly in advance, is the middle case. The most likely revision is that the companion paper’s Section 4 collapses from nine causes to three or four that survive, with the remainder demoted from causes to correlates.
4. Evidentiary Handling of Genesis 41–47
The companion paper used this material without stating which of three distinct uses it was making, and the uses carry different burdens.
Use A — as evidence of Egyptian administrative practice. This is the heaviest burden and the paper does not need it. Independent Egyptian evidence for state granary administration exists — architectural remains, funerary granary models, and accounting documents (Kemp, 2006) — and establishes that centralized grain storage under state officers was a real institutional form. It does not establish the seven-year program as an event. Any attempt to corroborate the narrative through the Sehel famine stela should be handled with care rather than enthusiasm, since the stela is a much later composition and the direction of dependence is unresolved. Concordist argument here would import a burden the policy claim does not require.
Use B — as evidence of an available conceptual scheme. The text demonstrates that the multi-year buffer concept, with a proportional levy, a named custodian, an execution layer, distributed depots, and an earmarked purpose, was articulable and transmissible at an early date. This is what the companion paper’s rarity argument actually needs, and it is fully carried by the text’s existence and circulation. It is the load-bearing use.
Use C — as a normative design template. Legitimate, and for a biblicist the text’s authority does not depend on external corroboration. But the argument should be built so that the policy claim rests on the paleoclimate and famine evidence and the text supplies the design articulation. Constructed that way, the recommendations survive intact for a reader who grants the text no authority at all, and the argument does not become hostage to an unrelated dispute.
Three further rules. First, the foreknowledge disanalogy must be carried through every recommendation rather than conceded once and dropped: Joseph had duration and magnitude, and the entire detection-and-release apparatus of recommendation 6.8 exists because no modern regime does. Second, chapters 41 and 47 must be treated as a single unit of analysis. The text supplies a design and its failure mode in the same document, which is rare and valuable, and citing one without the other is not abbreviation but misreport. Third, the outcome in chapter 47 should be read as evidence about monopoly reserves without distributional rules, not as evidence about reserves as such — the mechanism is the sale-on-terms structure, and that is a specifiable and separable design feature.
5. Scope Conditions and the Audit Schedule
The prescription in the companion paper holds only while certain conditions do: that severe shocks remain serially correlated over multiple years; that they are supra-regional, so trade partners are afflicted simultaneously; that private storage horizons remain shorter than the shock duration; and that freight and substitution cannot close the gap within a growing season.
Each has an indicator and a plausible failure. Global staple stock-to-use ratios, the count of independent major producing regions, freight redundancy on the principal grain routes, and the frequency of export restriction during price spikes are all observable. If producing regions become sufficiently numerous and climatically decorrelated, and freight sufficiently redundant, the trade-substitution argument becomes correct in the tail as well as in the ordinary case, and the storage prescription should be withdrawn rather than defended.
A five-year review interval is proposed, with the applicability check written as a standing section of the project rather than left to occur to someone.
6. Feasibility
D4, D6, D7, and D8 are executable from published literature and require no archival access; they are also, not coincidentally, the designs that carry the most discriminating weight. D5 requires assembling transport and technique series but is tractable. D1 and D2 require specialist collaboration — experimental archaeology for the former, and reading knowledge of the relevant administrative languages for the latter — and should be framed as invitations to collaborators rather than held as personal deliverables. A project that begins with D4 and D6 will know within a reasonable span whether the companion paper’s central claim survives, and that ordering should govern.
Notes
1. The asymmetry in Section 1 — uniform versus varying constraint — is the whole discriminating engine, and it fails if the institutional variables are themselves determined by the technical ones. If regimes with better storage technique systematically also had better custodial institutions, D3 and D7 lose their leverage. This should be checked directly before those designs are run, not assumed away.
2. D4’s coding categories are not fully exclusive in practice: a stock sold cheaply to a favored buyer after being written down as spoiled is simultaneously write-off and diversion. The protocol should permit a dual code with the primary category recorded, and the rate of dual-coding reported, since a high rate is itself informative about custodial opacity.
3. The valuation term in D6 is the model’s weakest joint. Placing a value on averted production collapse requires assumptions about the mortality-to-output relationship that are contested. The correct handling is to report the break-even surface across a wide valuation range and let the reader locate their own, rather than defending a point estimate.
4. On pre-registration: the practice was developed for experimental settings and transfers imperfectly to historical work, where the analyst has usually read some of the evidence before specifying the test (Nosek et al., 2018). The honest form is a dated statement of predictions and disconfirming conditions together with a disclosure of what was already known at the time of writing. That disclosure should accompany any executed design here.
5. The likeliest real outcome of this program is neither confirmation nor refutation but a revised object, in the pattern that a well-run audit usually produces: not “was it spoilage or governance” but “which storage regimes made the political choice available at all.” That is a better question than the one this paper was written to settle, and finding it would count as the program working rather than failing.
References
Alfani, G., & Ó Gráda, C. (Eds.). (2017). Famine in European history. Cambridge University Press.
Boxall, R. A. (2001). Post-harvest losses to insects: A world overview. International Biodeterioration & Biodegradation, 48(1–4), 137–152.
Büntgen, U., Myglan, V. S., Ljungqvist, F. C., McCormick, M., Di Cosmo, N., Sigl, M., Jungclaus, J., Wagner, S., Krusic, P. J., Esper, J., Kaplan, J. O., de Vaan, M. A. C., Luterbacher, J., Wacker, L., Tegel, W., & Kirdyanov, A. V. (2016). Cooling and societal change during the Late Antique Little Ice Age from 536 to around 660 AD. Nature Geoscience, 9(3), 231–236.
Gerring, J. (2007). Case study research: Principles and practices. Cambridge University Press.
Gilman, G. A., & Boxall, R. A. (1974). The storage of food grains in traditional underground pits. Tropical Stored Products Information, 28, 19–38.
Gouel, C. (2014). Food price volatility and domestic stabilization policies in developing countries. In J.-P. Chavas, D. Hummels, & B. D. Wright (Eds.), The economics of food price volatility (pp. 261–306). University of Chicago Press.
The Holy Bible: New King James Version. (1982). Thomas Nelson.
Kaplan, S. L. (1976). Bread, politics and political economy in the reign of Louis XV. Martinus Nijhoff.
Kemp, B. J. (2006). Ancient Egypt: Anatomy of a civilisation (2nd ed.). Routledge.
King, G., Keohane, R. O., & Verba, S. (1994). Designing social inquiry: Scientific inference in qualitative research. Princeton University Press.
Mahoney, J., & Goertz, G. (2004). The possibility principle: Choosing negative cases in comparative research. American Political Science Review, 98(4), 653–669.
McConnell, J. R., Sigl, M., Plunkett, G., Burke, A., Kim, W. M., Raible, C. C., Wilson, A. I., Manning, J. G., Ludlow, F., Chellman, N. J., Innes, H. M., Yang, Z., Larsen, J. F., Schaefer, J. R., Kipfstuhl, S., Mojtabavi, S., Wilhelms, F., Opel, T., Meyer, H., & Steffensen, J. P. (2020). Extreme climate after massive eruption of Alaska’s Okmok volcano in 43 BCE and effects on the late Roman Republic and Ptolemaic Kingdom. Proceedings of the National Academy of Sciences, 117(27), 15443–15449.
Nosek, B. A., Ebersole, C. R., DeHaven, A. C., & Mellor, D. T. (2018). The preregistration revolution. Proceedings of the National Academy of Sciences, 115(11), 2600–2606.
Ó Gráda, C. (2009). Famine: A short history. Princeton University Press.
Persson, K. G. (1999). Grain markets in Europe, 1500–1900: Integration and deregulation. Cambridge University Press.
Reynolds, P. J. (1974). Experimental Iron Age storage pits: An interim report. Proceedings of the Prehistoric Society, 40, 118–131.
Sen, A. (1981). Poverty and famines: An essay on entitlement and deprivation. Clarendon Press.
Sigaut, F. (1988). A method for identifying grain storage techniques and its application for European agricultural history. Tools & Tillage, 6(1), 3–32.
Sigl, M., Winstrup, M., McConnell, J. R., Welten, K. C., Plunkett, G., Ludlow, F., Büntgen, U., Caffee, M., Chellman, N., Dahl-Jensen, D., Fischer, H., Kipfstuhl, S., Kostick, C., Maselli, O. J., Mekhaldi, F., Mulvaney, R., Muscheler, R., Pasteris, D. R., Pilcher, J. R., … Woodruff, T. E. (2015). Timing and climate forcing of volcanic eruptions for the past 2,500 years. Nature, 523(7562), 543–549.
Will, P.-É., & Wong, R. B. (1991). Nourish the people: The state civilian granary system in China, 1650–1850. Center for Chinese Studies, University of Michigan.
Williams, J. C., & Wright, B. D. (1991). Storage and commodity markets. Cambridge University Press.
