HISTORICAL BACKGROUND

Separation of powers in this article means the regularized participation in policy making by both a governor and an assembly, under conditions such that the governor and assembly have independent power bases. Most naturally in the present context, the power base of the governor is the Crown of England (before 1707) or Britain (after 1707). The power base of the assembly in separation of powers is some distinct subset of the colonial population (e.g., a planter elite, all white male property owners). Separation of powers does not obtain in this definition when the governor has tools to force a nominally elective assembly into submission to the governor’s plans (e.g., control over patronage appointments upon which individual assembly members depend). Separation of powers requires independence of the institutions of government from one another.Footnote ¹³

The concept of separation of powers available to the framers of the U.S. federal and state constitutions was not a novel abstraction from the minds of theoreticians such as Locke, Montesquieu, and Blackstone. In fact, separation of powers already existed in the English imperial constitution long before the 1780s (Wood Reference Wood1969; Lutz Reference Lutz1988). Without denigrating the original theoretical arguments in The Federalist (etc.), one can suspect that this colonial experience affected the framers of the U.S. Constitution. Bryce (Reference Bryce1888) put it beyond dispute: “From their colonial experience, the men of 1787 drew [the] conclusion. . .that the vesting of legislative and executive powers in different hands was the normal and natural feature of a free government. . .[and] that in order to check the head of state it was necessary. . .to destroy his opportunities of influencing the legislature” (26). And another classic study: “The Framers of the first American constitutions were impressed by the separation of powers theory only because their own experience. . .confirmed its wisdom” (Wright Reference Wright1933).

Mixed government came to America with the English. Starting with the royally approved corporate charters of Virginia (1606, 1609, 1612), the proprietor- or Crown-appointed colonial governor was always required at least to seek advice from some sort of council. However, mixed government is different from separation of powers (Vile Reference Vile1967). Some charters specifically did not provide for the council to represent the people; rather, council members could be selected on recommendation by the governor (e.g., Virginia until 1619, New York in 1664). Moreover, representative assemblies were not immune to domination by the governor through selection, bribery, and prorogation, such that assembly independence was compromised (e.g., Virginia in the 1670s).

Institutional development in English North America occurred within colonial (i.e., within colonies) and imperial (i.e., between colonies and the metropole) spheres (Greene Reference Greene1986). Institutional forms in the 17th century derived legal authority from the assent of the Crown (Keith Reference Keith1930), but strategic interaction by key actors in these spheres—the Crown, governor, and settlers—determined their structure (Greene Reference Greene1986). These institutions followed complex paths of development with numerous influences and no single, time-consistent royal plan (Greene Reference Greene1963). Nevertheless, the Crown often acted as though colonial and imperial political structure would affect the value of the colonies to the Crown and attempted to restructure colonial and imperial institutions to suit its interests (Rainbolt Reference Rainbolt1967; Stanwood Reference Stanwood2011). During the 17th and 18th centuries, colonial assemblies developed the independence necessary for separation of powers in a gradual process (Greene Reference Greene1963), punctuated by important Crown support (Taylor Reference Taylor2002).Footnote ¹⁴

These institutions were not replicas of those found in the metropole itself (Lutz Reference Lutz1988). To be sure, there is a clear homology between assemblies in English colonies and the House of Commons in England. However, the colonial assemblies, their eventual independence from colonial governors, and the resulting separation of powers were not merely imported from England or created because it was “the English way.” The English Crown took steps in support of assemblies in America that it did not take contemporaneously with Parliament in England (discussed further below after the model is presented). For example, Charles II granted several American colonial assemblies the right to schedule their own meetings, while opposing the Triennial Act guaranteeing the English Parliament the right to meet every 3 years. Some North American assemblies held powers over disbursement of funds and planning of military excursions that Parliament never claimed (Labaree Reference Labaree1930).

Beyond direct assertions of control over internal colonial structure, the English Crown supported the independence of colonial assemblies, simply by refusing to allow challenges to them. This fell under the policy of “wise and salutary neglect” of the colonies by the metropole, a policy at its high water mark under the Prime Ministers of the early Hanover kings (ca. 1721–1754). The policy implied more than simple disregard: the king-in-council passively protected colonial assemblies from metropolitan incursions. For example, on three separate occasions between 1734 and 1749, the king’s Privy Council declined to support bills in Parliament asserting the supremacy of Crown instructions over colonial law, effectively vetoing the bills (Greene Reference Greene1963).

All of this implies that to understand the genesis of separation of powers in the United States, it is not sufficient to examine the framing of the U.S. federal or state constitutions of the 1780s, nor to examine the domestic English constitution of the 17th century. Instead, it is necessary to examine the design of institutions in the British Empire (Bryce Reference Bryce1888; Wright Reference Wright1933; Lutz Reference Lutz1988).

A MODEL OF THE COLONIAL POLITICAL ECONOMY

This section describes a baseline model of a colony’s political-economic process. It is designed to capture endogenous economic development of the colony driven by settlers—a salient condition in English North America in the 17th century.Footnote ¹⁵ This baseline model is then embedded in two different models of the political process: one with hierarchical control among imperial officials, and one with separation of powers within the colony.

There are three players in the model: the Crown C, the governor G, and settlers S. All players are infinitely lived with a common stationary discount factor δ, and interested in maximizing their own (and only their own) discounted stream of economic payoffs.Footnote ¹⁶

The economy recurs in an infinite sequence of identical periods indexed by t. Colonial settlement is such that without a governor in the long run, the Crown has no legal claim on economic output at all and would probably lose its colony to some other European crown (MacMillan Reference MacMillan2006). Thus, I assume that there is a governor in office at the start of every period. The colony’s output in a given such period t is V ^t . The minimal value of V ^t is 1. S can make an investment in the colony, e ^t ∈ {0, 1}. High-value investment in the colony, e ^t = 1, adds V _H to the colony’s period t output. Low-value investment in the colony, e ^t = 0, is valued at V _L and captured entirely by settlers. It does not add to the colony’s output and cannot be expropriated by officials. Assume 0 < V _L < V _H . Either investment fully depreciates after each period.

The colony’s output in period t can thus be compactly expressed as

(1) $$\begin{equation} V^t= 1 + e^t V_H. \end{equation}$$

With the high-value investment (e ^t = 1), the economy’s output is 1 + V _H . With the low-value investment (e ^t = 0), the economy’s output is 1.

Let α ^t _i denote a share of V ^t taken by player i in period t (these shares are endogenously determined in equilibrium below). Then period t payoffs are

$$\begin{eqnarray*} u^t_S &=& \alpha ^t_S V^t + (1-e^t)V_L, \nonumber \\[6pt] u^t_G &=& \alpha ^t_G V^t, \nonumber \\[6pt] u^t_C &=& \alpha ^t_C V^t. \nonumber \end{eqnarray*}$$

Thus, settlers can generate a larger economy V ^t in a given period through high investment in that period, but some of the value may be captured by imperial officials—the governor and the Crown. So, if e = 1, settlers take α _S (1 + V _H ). High investment in synergistic activities with other economic agents, traders, shippers, and so on, builds more value, but by the same token, it is observable by many actors and so more easily within reach of the state. It can be extracted by officials and damaged or lost in the event of political disruptions or civil unrest.

On the other hand, settlers can generate a safer return for themselves through low investment in the colony: when e = 0, settlers take α _S + V _L . The settlers can lose only what they put within reach of the government. Investments outside of the economy are also more easily placed beyond the reach of the state and its agents: they are harder to tax, cannot be captured by an absconding governor, and are not destroyed or forfeited to the Crown in rebellion.Footnote ¹⁷ For example, small-scale agriculture on the frontier not only generates less value than cash crops for international trade but also does not produce surplus attainable by revenue collectors.Footnote ¹⁸

The focus on settler investment decisions reflects that white English settlers came mostly voluntarily,Footnote ¹⁹ and it had to be worth their while to labor for the development of the colony.Footnote ²⁰ This participation constraint was apparent to colonial and imperial elites at the time; for example, a British minister counseled a royal colonial governor, “In the Plantations, the Government should be as Easy and Mild as possible to invite people to Settle under it.”Footnote ²¹

POLITICAL PROCESS: HIERARCHICAL AGENCY

The political process modeled in this section is designed to highlight an agency problem between the Crown and imperial governors. It is a moral hazard model of political agency. It modifies standard models (Barro Reference Barro1973; Ferejohn Reference Ferejohn1986) by introducing successive rent extraction by two actors, of a form that appears not to be present in the political agency literature to date. The first actor (governor) extracts rents from the entire real economy, and the second (Crown) extracts rents from whatever is left after the governor’s cut. The first can be dismissed by the second if its rents are not high enough, and they both can be punished by the last mover (settlers, via rebellionFootnote ²² ) if its total rents are not high enough.Footnote ²³

Formally, the stage game in period t, which is repeated infinitely many times, consists of five steps. This structure, all parameters, and all utility functions are common knowledge.

1. 1. S chooses investment e ^t ∈ {0, 1}, which determines V ^t as per Equation (1).

2. 2. G extracts a share x ^t ∈ [0, 1] of V ^t .

3. 3. C observes (1 − x ^t )V ^t and chooses whether to retain or sack G.Footnote ²⁴

   • • If C retains G (r^t _c = 1), then play moves to step 4.

   • • If C sacks G (r^t _c = 0), the stage is over, S and C move to stage t + 1, G is replaced with an identical G in t + 1, and period t payoffs are

     • * u^t _G = σ _G V^t ,

     • * u^t _C = σ _C V^t ,

     • * u^t _S = σ _S V^t + (1 − e^t )V_L .

4. 4. C extracts a share y ^t of (1 − x ^t )V ^t .

5. 5. S observes (1 − x ^t )(1 − y ^t )V ^t and remains loyal or rebels.

   • • If S remains loyal (r^t _s = 1), the stage is over, all players continue to period t + 1, and period t payoffs are

     • * u^t _G = x^tV^t ,

     • * u^t _C = y^t (1 − x^t )V^t ,

     • * u^t _S = (1 − x^t )(1 − y^t )V^t + (1 − e^t )V_L .

   • • If S rebels (r^t _s = 0), the stage is over, all players continue to period t + 1, and period t payoffs are

     • * u^t _G = ρ _G V^t ,

     • * u^t _C = ρ _C V^t − M,

     • * u^t _S = ρ _S V^t + (1 − e^t )V_L .

The endogenous choices (lowercase letters) are e^t , x^t , y^t , r^t _c , and r^t _s . The exogenous parameters (Greek or capital letters) are V _L , V _H , δ (explained above), M, σ _i , and ρ _i (explained below). The stage game is stationary, with identical parameters recurring every period.

Each player i’s total payoff for the repeated game is the discounted sum of stage game payoffs. As is standard in stationary political agency models (e.g., Ferejohn Reference Ferejohn1986), I focus on stationary subgame perfect equilibrium (stationary SPE).Footnote ²⁵ These are based on simple retrospective judgments and do not rely on complicated rules or multigenerational commitments. Given stationarity, the time superscript t on endogenous variables will generally be suppressed below; for example, V = 1 + (1 − e)V _H simply represents (expropriable) economic output in a given period with investment e ∈ {0, 1}.

In the rest of this section, I offer substantive motivation and justification for several facets of the model’s structure, then I consider stationary SPE and key comparative statics, and finally I consider alternative model structures and the robustness of the results to them.

Substantive Motivation

Stage Game Sequence. The stage game allows the governor to choose his tax rate before the Crown.Footnote ²⁶ This allows the governor to determine the Crown’s tax base, (1 − x)V. The Crown could attempt to set its tax rate y in advance by statute, but its tax base will then determine its revenues, y(1 − x)V.

Any extraction by governors from the economy’s productive capacity works as a reduction in the Crown’s tax base in this way. Governors had numerous channels for such extraction from the colonial economy, through tax rates and beyond (McCusker and Menard Reference McCusker and Menard1985; Rabushka Reference Rabushka2010)—for example, by preferential access the largest and best tracts of land (even abrogating existing land titles in some cases), control of colonial credit, fees for administrative services, management of the lucrative black market international trade,Footnote ²⁷ regulation of Indian trade, and so on. If the governor takes the most productive parcels of land for his own use and evades imperial taxes by smuggling his own output to foreign shippers, the governor’s self-dealing shrinks the base to which the Crown’s tax rate y applies, as the model reflects. Colonial taxes themselves also worked in this way, due to sequential points of collection (Rabushka Reference Rabushka2010): if a share x is taken from the colonial harvest for internal management,Footnote ²⁸ then at most (1 − x) is left to ship to England, where import duties were collected upon arrival.

I return to the foundations of the stage game sequence later in this section, after the analysis is complete.

Sacking and the Role of the Governor. If the Crown is not satisfied with (1 − x)V in a given period, it can sack the governor (step 3 of the stage game). Governors were understood as agents of the Crown, particularly in the royal colonies (Keith Reference Keith1930), but eventually in all of them (Wood Reference Wood1969). Sacking or recalling the governor was the Crown’s principal remedy for unsatisfactory performance (Greene Reference Greene1898).Footnote ²⁹ Sacking G immediately ends the period—and G’s political career. S and C move on to period t + 1, but G is replaced in period t + 1 by an ex ante identical replacement, coincidentally also named G. Period t payoffs are u^t _i = σ _i V^t for i ∈ G, C, and u^t _S = σ _S V^t + (1 − e^t )V_L .

The shares σ _i ∈ (0, 1) are exogenous parameters.Footnote ³⁰ I assume that ∑ _i σ _i < 1, so sacking the governor causes short term economic disruption.Footnote ³¹ Governors provided law and order, militia defense, management of credit and currency, black market access, and land distribution for the colony (McCusker and Menard Reference McCusker and Menard1985). These services are interrupted when the governor is sacked, and thus the economy experiences a short-run disruption. I assume this disruption lasts only a single period; installation of a new governor in the next period restores these services. Low-investment returns V _L are not lost by settlers in case the governor is sacked. Investments outside of the economy do not take their value from a well functioning public sector.

Note that σ _G > 0, so the governor does not obtain zero payoff from being sacked. Governors could execute rapid transfers of land or appointments to office for themselves and close family and associates, allowing them to claim at least some rents if they expect to leave office. Moreover, their centrality in the colonial economy and administration meant that orderly transition would be in the interest of the Crown (Greene Reference Greene1898). This would be facilitated by conceding at least some rents to the governor upon termination from office.

Settler Rebellion. If the Crown retains the governor, the Crown extracts a share y from (1 − x)V. Then (step 5 of the stage game) the settlers observe (1 − x)(1 − y)V and decide whether to remain loyal or rebel. Either choice ends period t, and all players move on to period t + 1.Footnote ³² If settlers remain loyal, period t payoffs are u^t _G = x^tV^t , u^t _C = y^t (1 − x^t )V^t , and u^t _S = (1 − x^t )(1 − y^t )V^t + (1 − e^t )V_L . If settlers rebel, period t payoffs are u^t _G = ρ _G V^t , u^t _S = ρ _S V^t + (1 − e^t )V_L , and u^t _C = ρ _C V^t − M. The shares ρ _i ∈ (0, 1) are exogenous parameters. I assume that ∑ _i ρ _i < 1, so that rebellion destroys some economic output in period t.Footnote ³³

North American settler rebellions were rare before 1776, but several did occur (Virginia, 1675–1676; Maryland, 1676; Carolina, 1677; New York, 1688), and they imprinted an unpleasant memory on future monarchs. For example, to suppress Bacon’s Rebellion (Virginia, 1676; this is discussed in more detail below after the model analysis is complete), Charles II dispatched a battalion of 1,000 Redcoats and 14 war ships—an exceedingly large mission at the time for the distance involved (Webb Reference Webb1995). Thus, rebellions were costly for the Crown, represented in the model as a parameter M. Since the governor did not have a military independent of the militia, which cannot be counted on in rebellion, governors did not face costs of rebellion on the same scale. I will assume that

(2) $$\begin{equation} M \ge 1+V_H \end{equation}$$

so that the Crown would never voluntarily and knowingly permit a rebellion.Footnote ³⁴

Analysis

There is an essentially uniqueFootnote ³⁵ stationary subgame perfect equilibrium (stationary SPE) in the repeated game for a given parameter configuration. It is one of two types. In extractive equilibrium, the governor is retained in every period; in sacking equilibrium, the governor is sacked in every period. Extractive equilibria are the main focus of the analysis. Sacking equilibria imply restrictions on parameters that may not be substantively plausible, as I discuss further below, but I include them in the analysis for completeness.Footnote ³⁶ Proofs and derivations of formal statements below are in the Appendix.

Extractive Equilibria and Retention of the Governor. By definition, an extractive equilibrium is a stationary SPE of the hierarchical agency model with retention of the governor in every period. This exists when the governor and Crown both receive enough rents each period to prefer having the governor retained over being sacked, and enough of the economy is left for settlers to prefer loyalty over rebellion.

More specifically, if G is retained in the stage game, then S has a decision to remain loyal or rebel. Settlers remain loyal if their cut of economic output V in a given period is at least as great as their cut if they rebel:

(3) $$\begin{equation} \rho _S \le (1-x)(1-y). \end{equation}$$

This is S’s loyalty constraint. Footnote ³⁷ In view of the high cost of rebellion M (Equation (2)), C will sack G if the loyalty constraint cannot be satisfied given x. Thus, an extractive equilibrium requires not only retention of the governor but therefore also settler loyalty every period.

Retention of G in equilibrium also requires incentive constraints for both G and C: C has to want to retain G, and G has to want to be retained. A necessary condition for C to prefer retention in a given period is

(4) $$\begin{equation} y(1-x) \ge \sigma _C. \end{equation}$$

If this does not hold, C obtains greater utility from sacking G and moving to the next period. Joint satisfaction of this constraint and the loyalty constraint is necessary and sufficient for C to prefer retention.

For the governor’s part, in a stationary equilibrium with retention, G obtains xV every period, for an infinite flow of benefits with present value $\frac{xV}{1-\delta }$ . If instead G were sacked, he would obtain the one-time payoff σ _G V. G prefers to be retained if and only if

(5) $$\begin{equation} x \ge (1-\delta )\sigma _G. \end{equation}$$

A high discount factor relaxes G’s constraint, because he places high value on the flow of future payoffs he obtains from retention.

Putting constraints (3), (4), and (5) together yields

Proposition 1 An extractive equilibrium exists only if the economic cost of sacking the governor is sufficiently high, or

(6) $$\begin{equation} \sigma _C + (1-\delta )\sigma _G \le 1 - \rho _S. \end{equation}$$

Inequality (6) is the retention constraint. It requires that the share of V needed to make G and C both prefer retention does not exceed the share of V left after settlers are induced to remain loyal. If G and C have less than a share σ _C + (1 − δ)σ _G to divide between themselves, at least one of them will prefer to have G sacked. However, settler loyalty implies that at most the share 1 − ρ _S will be available for G and C to split. Retention is possible in any equilibrium (stationary or otherwise) if and only if the amount available is at least the amount required. The constraint becomes easier to satisfy as the governor and Crown incur greater economic costs from sacking the governor (σ’s are smaller), when the settlers fare poorly under rebellion (ρ _S is smaller), and when the governor places more weight on the future stream of payoffs (δ is larger).

A simple way to view the retention constraint is as a limit on the short-run economic disruption caused by sacking the governor. If σ _C + σ _G is small, this disruption is large, at least as far as imperial authorities are concerned. Large economic disruption means the imperial authorities fare poorly if they do not mollify the settlers, so it is easier to do so.

When the retention constraint is satisfied, we can turn to the distribution of surplus rents between the players, and its effect on settler investment. In an extractive equilibrium, settlers make the same investment decision e in every period. Because the Crown’s utility is increasing in its share y for given V, it prefers to concede rents ρ _S V to settlers—just enough to satisfy the loyalty constraint with equality, given G’s tax rate x.Footnote ³⁸ This implies C will choose

(7) $$\begin{equation} y^*(x) \equiv 1 - \frac{\rho _s}{(1-x)}. \end{equation}$$

This captures the strong agenda power the Crown has over settlers. The Crown moderates rent extraction just enough to induce loyalty, but no more. C must moderate its rent extraction y by S’s take in case of rebellion (ρ _S ), adjusted for the rent extraction (x) by G. As settlers do better in case of rebellion (and holding x fixed), more rents must be conceded by C. And as G takes more rent, C must moderate its rent extraction.

In extractive equilibrium, the governor has strong agenda power with respect to the Crown, just as the Crown does with respect to settlers.Footnote ³⁹ In particular, G concedes just enough to allow C to satisfy the loyalty constraint with equality and to make C indifferent between sacking and retaining. G then keeps the rest of the pie for itself. C would like to threaten G with sacking unless G concedes more rents to C, but because sacking is costly for C (due to short-run economic disruption), this threat is credible if and only if C’s retention constraint binds (this is proved formally in Remark 1 below).

As a result, the governor’s share in an extractive equilibrium is x* = 1 − ρ _S − σ _C . Substituting x* into Equation (7) yields a share $y^* = \frac{\sigma _C}{\rho _S + \sigma _C}$ to the Crown. This implies the following proposition.

Proposition 2 If the economic cost of sacking the governor is sufficiently high (the retention constraint, inequality (6), holds), the unique stationary subgame perfect equilibrium is an extractive equilibrium. In every period, settlers are indifferent between rebellion and loyalty; the Crown is indifferent between sacking and retaining the governor; and the governor captures all surplus rents (1 − ρ _S − σ _C ).

They are called “extractive” because they involve maximal rent extraction from settlers every period, given the loyalty constraint. Propositions 1 and 2 together imply that the retention constraint, 1 − ρ _S ≥ σ _C + (1 − δ)σ _G , is necessary and sufficient for existence of a unique extractive equilibrium.

Extractive equilibria have a simple retrospective character within each period. If total rent extraction was too high, the settlers rebel. If the governor’s rent extraction was too high, the Crown sacks him. “Too high” is judged by each player relative to its payoff in case punishment is inflicted. In particular, rent extraction from settlers is constrained by their “outside option” of rebellion (provided this value is not so high that the retention constraint from Proposition 1 fails). As for the Crown, we can think of σ _C as measuring the governor’s value directly to the Crown in the short run. When σ _C shrinks, the Crown does worse in the short run without the governor, and thus the short-run value of the governor to the Crown grows. In extractive equilibria, the governor exploits his value to the Crown, and the Crown’s strong desire to prevent rebellion.

The key question in this model is investment e by settlers. Settler investments are sunk in each period, so they face a hold-up problem after high investment that is mitigated only by the possibility of settler rebellion in extractive equilibrium. Thus, high investment brings returns V _H to the economy, but settlers capture only ρ _S V _H of the marginal returns. Low investment, on the other hand, brings returns V _L that are safe from imperial rent extraction. Thus, settlers invest in extractive equilibrium if and only if the investment constraint is satisfied:

(8) $$\begin{equation} \rho _S \ge \frac{V_L}{V_H}. \end{equation}$$

This constraint is easy to satisfy when returns to investment (V _H ) are very high. In that case, even a small share of a vastly enlarged pie meets the incentive constraint for settlers. In addition, settlers get more in extractive equilibrium when they do relatively well in rebellion, and so the constraint is easier to satisfy as ρ _S increases—provided it is not so large that the retention constraint (Equation (6)) fails.

Combining the retention and investment constraints, it follows that a high investment extractive equilibrium exists if and only if 1 − σ _C − (1 − δ)σ _G ≥ ρ _S ≥ ${V_{L}}/{V_{H}}$ : the economic disruption from sacking, and the returns to settler investment, are both high. But if the disruption from sacking is large, and returns to settler investment are middling or worse (V _H close enough to V _L ), then settlers do not gain enough from investment to compensate for their hold up problem. In these cases, investment is low, and rent extraction by the Crown and governor undermines efficiency in equilibrium.

This logic is summarized as follows.

Proposition 3 Assume the economic cost of sacking the governor is sufficiently high (the retention constraint, inequality (6), holds). Then there is high settler investment in the unique extractive equilibrium if and only if returns to settler investment are sufficiently high (i.e., the investment constraint, inequality (8), holds).

Sacking Equilibria. Extractive equilibria involve governor retention in each period. By Proposition 1, if the retention constraint is not satisfied, all stationary equilibria entail sacking the governor in every period. When G is sacked, the payoff to player i is σ _i V, so S invests (e = 1) in a sacking equilibrium if and only if σ _S V _H ≥ V _L .

Proposition 4 If the economic cost of sacking the governor is sufficiently low (the retention constraint, inequality (6), fails), the governor is sacked every period in a stationary subgame perfect equilibrium. Settlers invest if and only if σ_S ≥ ${V_{L}}/{V_{H}}$ .

However, sacking equilibria call into question why the colony exists at all. Intuitively, settlers must do so well by rebelling, or the governor contributes so little to the colony’s value, that the imperial authorities are better off crippling the colony and running it without a governor every period. Colonial governance is not doing much good in this case, so it is better to dissolve it. Sacking equilibria with high-value investment exist only if both ρ _S and σ _S are very high, such that settlers obtain little benefit from the law and order (etc.) provided by the colonial governor, and would do relatively well under rebellion.

Cooperative Equilibria. In some cases, there are also subgame perfect equilibria in which S’s loyalty constraint is strictly satisfied. These “cooperative equilibria” alleviate inefficiencies from low investment as well as allowing S to capture more rents. An example of this type of equilibrium is presented in the Appendix. Here I will discuss some intuition behind them and then explain why they are not deeply compelling on substantive grounds. While subgame perfect, these equilibria involve history-dependent, or nonstationary, strategies.

Cooperative equilibria require several restrictions on parameters. First, the retention constraint (Proposition 1) must be satisfied. Second, the discount factor δ must be large enough. Players need a carrot of high future payoffs to sustain cooperative equilibria. Third, the investment constraint must not be satisfied. This is essential for making credible threats that underpin these equilibria.

Credible threats to sustain cooperative equilibria can come only at the stage of settler investment or governor rent extraction.Footnote ⁴⁰ Settlers can threaten the governor and Crown with a trigger strategy: low investment forever in the future if C and G do not restrain their rent extraction today. The settlers’ threat is to play the (inefficient) extractive equilibrium in every subsequent period after the Crown or governor deviate from the cooperative equilibrium path. Since the extractive equilibrium is subgame perfect, this threat is credible by construction. Restraint by the governor and Crown to taxes below the extractive equilibrium levels supports investment in every period on the equilibrium path.

However, cooperative equilibrium arguments are not exceedingly compelling ways around inefficiency or low rents for S and C in extractive equilibria. Despite the intuition that the Crown (at least) is a “long-run player” and should place high weight on the future, it is not likely that early modern Crowns and governors had the high discount factors needed to sustain these equilibria. These Crowns were perennially cash strapped and indeed famous for trading valuable long-term assets for small but immediate gain (Pritchard Reference Pritchard2007)—the hallmark of a small discount factor. This is particularly true of the pre-revolutionary Stuarts in England, who struggled to find revenue sources in the face Parliament’s resistance at home (Rabushka Reference Rabushka2010). North and Weingast (Reference North and Weingast1989), Stasavage (Reference Stasavage2003), and Cox (Reference Cox2016) all point out these same monarchs were unable to restrain their own expropriation from investors, at major long-run cost to themselves. Folk theorem arguments imply these monarchs should have avoided this problem. Clearly, their discount factors were sufficiently low as to make their weak commitment ability legendary.

EMPOWERING SETTLERS: SEPARATION OF POWERS

This model reveals the stark agency problem that the Crown faces with the governor. We should expect rent-maximizing Crowns to devise institutions to mitigate that problem. One possibility is to delegate control of the governor to the settlers. In extractive equilibria, the Crown and the settlers have a common interest in restraining the governor’s rent extraction. The Crown can leverage this common interest by empowering an independent settler assembly to check the governor’s power, and separating it from the governor’s control. This can constrain the governor sufficiently that settlers invest—and thereby increase the Crown’s rents.

There are two essential elements of 17th- and early 18th-century elected colonial assemblies to consider in a model. First, the focus of these assemblies was overwhelmingly on fiscal matters (Greene Reference Greene1963; Elliott Reference Elliott2007).Footnote ⁴⁹ The assemblies took control of raising and appropriating public revenue for internal colonial business, and fiercely resisted attempts at direct taxation by the Crown. Governors administered the appropriated funds under close supervision by the assemblies. Thus, assembly independence from the governor and separation of powers in colonial constitutions developed hand in hand. Second, the Crown was the legal basis for colonial institutional forms (Keith Reference Keith1930). Colonial assemblies could only function in the long run with its assent and recognition.

Bringing these elements together, the model of separation of powers allows settler determination of funds available to the government, but only if the Crown prefers to recognize this right. The new stage game begins with C choosing to have S play the separation of powers game or the baseline game above.Footnote ⁵⁰ In separation of powers, S chooses a public sector budget 0 ≤ P ≤ V for G and C to split. Play then proceeds as before with choice of tax rates, and possibilities of sacking and rebellion.

If settlers allocate the budget P in period t under separation of powers, and the governor is retained and settlers are loyal, then payoffs in that period are

• • u _S = (V − P) + (1 − x)(1 − y)P + (1 − e)V _L ,

• • u _G = xP,

• • u _C = y(1 − x)P.

If the settlers are not loyal or the governor is not retained under separation of powers, then payoffs are as in the baseline game. A sacked governor affects not only the government budget; it affects all of V. And the stakes of a settler rebellion are not merely its appropriation of public money; it is all of V.

Since sacking gives exactly the same utilities in the separation of powers and baseline agency games, the retention constraint is the same in each case. However, in separation of powers, settlers now have proposal power to implicitly determine precisely which incentive compatible share of the pie is taken by the governor. In extractive equilibria of the baseline game, the governor uses agenda power to choose its most preferred share that satisfies the retention constraint. That is the crucial difference between the games.Footnote ⁵¹

Suppose there is a stationary subgame perfect equilibrium such that the Crown chooses separation of powers, settlers choose a budget P, the Crown retains the governor, and settlers remains loyal in every period. The retention constraint (Proposition 1) implies the separation of powers budget constraint,

(9) $$\begin{equation} P \ge ((1-\delta )\sigma _G + \sigma _C)V. \end{equation}$$

If inequality (9) does not hold, the retention constraint cannot be satisfied, and G is sacked. If it does hold, G leaves σ _C V for C. G then keeps P − σ _C V ≥ (1 − δ)σ _G V and so prefers to be retained.

In equilibrium under separation of powers, S pays exactly enough to keep the colonial administration running: the separation of powers budget constraint holds with equality. Any additional payment would be extracted by G. Any less, and the administration would shut down, which S does not prefer as long as the retention constraint holds.Footnote ⁵² The equilibrium budget P leaves a share of the economy σ _C to C, (1 − δ)σ _G to G, and 1 − σ _C − (1 − δ)σ _G to S. Thus, empowering the assembly shifts agenda setting powers from the governor to the settlers.

Note, therefore, that conditional on the settlers’ investment e, the Crown’s payoff is σ _C V under either separation of powers or an extractive equilibrium.Footnote ⁵³ It is only the governor’s share that is restrained by separation of powers in equilibrium, and the rents are transferred to settlers. In this way, separation of powers clearly restrains the Crown’s agent.Footnote ⁵⁴

From the Crown’s point of view, this restraint on the governor is beneficial because it affects settlers’ incentives to invest. Settlers will invest (e = 1) under separation of powers if and only if

(10) $$\begin{equation} 1-\sigma _C-(1-\delta )\sigma _G \ge {{\raise3pt\hbox{${V_{L}}$}/\raise-3pt\hbox{$V_{H}$}}}. \end{equation}$$

In contrast, in an extractive equilibrium of the hierarchical agency game, investment occurs if and only if 1 − σ _C − (1 − δ)σ _G ≥ ρ _S ≥ ${V_{L}}/{V_{H}}$ (inequalities (6) and (8)). When investment returns are very high ( ${V_{L}}/{V_{H}}$ is very low), inequalities (6), (8), and (10) are all satisfied, and high-value investment would occur in both separation of powers and the hierarchical agency game. But for moderate investment returns, inequality (10) is satisfied, while inequality (8) is not. Then high-value investment occurs only under separation of powers.

This is summarized in the following proposition, and illustrated in Figure 1 below.

FIGURE 1.

Investment and retention in extractive equilibria (Regions I, II, and III) and sacking equilibria (Regions IV and V). Returns to investment shrink as ${V_{L}}/{V_{H}}$ grows. Separation of powers causes a change from low to high investment in region II

When returns are moderate, separation of powers leads to an efficiency gain—higher settler investment—that benefits the Crown. The Crown has a strict incentive in this case to empower settlers to restrain governors.Footnote ⁵⁵ With low returns on investment, we would expect the Crown to be relatively disinterested in restraining the governor through colonial institutions. With high investment returns, we would expect the Crown to be satisfied with extractive equilibria (though separation of powers also supports high investment in these cases). But with intermediate returns, it is reasonable to expect the Crown’s active support (and per force, toleration) of independent colonial assemblies with power of the purse—a fundamental component of separation of powers. This is interesting because the Crown is assumed to pursue only its own economic rents; in this case, they are bolstered by protecting liberal institutions of government, with an ensuing gain in efficiency.Footnote ⁵⁶

In the next section, I interpret historical patterns of New World institutional development in light of this result.

INSTITUTIONAL DEVELOPMENT IN NEW WORLD EMPIRES

Virginia, 1676: Bacon’s Rebellion. The logic of the model is aptly illustrated by Charles II’s response to Bacon’s rebellion in Virginia in 1676. The colony had been governed by William Berkeley, nominally as an agent of the Crown since the Restoration of the English monarchy in 1660. Governor Berkeley prevented new assembly elections from 1661 to 1676, and he used his power to fill up vacancies to stack the Burgesses with his allies (Wertenbaker Reference Wertenbaker1914). Berkeley also controlled assignment of valuable public executive offices (which sitting Burgesses were not prohibited from holding at the time), and granted enormous tracts of the best land to favored Burgesses (Elliott Reference Elliott2007). Berkeley himself sat ex officio in the assembly, facilitating coordination with Burgesses and monitoring deals (Webb Reference Webb1979). In this manner, the governor obtained the assembly’s complicity in governing the colony as a large rent extraction scheme.

The resulting colonial taxes (principally a poll tax) were oppressive and highly regressive; middle and lower class households paid upwards of half their annual harvest in taxes (Taylor Reference Taylor2002). The governor poured the money into such public goods as high salaries for council members (about 250 pounds of tobacco per Burgess per day the assembly was in session),Footnote ⁵⁷ bills for their arduously long meetings at Richmond taverns, and a system of earthen fortifications that, while largely useless from a martial point of view, were conveniently located on the estates of political elites, which thus required large public subsidies for property improvement (Webb Reference Webb1995).

Middle-class planters responded to sharply limited quantities of suitable land to cultivate crops to meet the high tax burden by building small farms on the fringes of Indian country. This was an attempt to bring new land into cultivation, thereby raising output and productivity to meet the extraordinary demands of the colony, and also of course to put more production beyond the immediate view of colonial authorities. The Indians, harassed by the settlers, complained to colony officials for redress. The governor maintained and captured rents from lucrative Indian trade and did not want it disturbed, so the assembly attempted to tamp down on frontier settlements (Webb Reference Webb1995).

The settlers, led by one Nathaniel Bacon, a well-born Englishman lately in America to make his fortune, rebelled against the colonial government in 1675-76. Bacon’s rebellion took up apparently most of the young, middle-class men of the colony. It culminated in the destruction by the rebels of the capital in Richmond and many surrounding estates of elites, including that of Governor Berkeley (Webb Reference Webb1995). The rebellion disrupted all economic activity in the colony for months and threatened the king’s £100,000+ revenue from Virginia tobacco (against annual royal treasury revenue of about £1,800,000), but lost steam when Bacon precipitously died of illness in 1676.

Deeply alarmed by the loss of official control and threat to revenue,Footnote ⁵⁸ Charles II dispatched 1,000 Redcoats, 14 war ships, and a retinue of advisors to restore order and institute reforms. This was easily the most extensive and costly English or British military police action in the New World before the Revolutionary War. The royal commissioners took a statement of grievances from the people of each county—a striking acknowledgement of the legitimacy of a popular insurrection by an absolutist-leaning Stuart Crown. The commissioners ascribed the rebellion to extreme rent extraction and political corruption by the governor and assembly. Aside from the obvious costs of military instability, this undermined the interests of the Crown in a large economy to provide abundant revenue. The king’s commissioners wrote of the common planters,

Needless to say, corn in tobacco country was not very useful to the king’s treasury.

The king’s commissioners recommended reforms to tie the assembly more closely to the common planters and break its complicity with the governor (Elliott Reference Elliott2007). These included requiring a new election every time the Burgesses were to meet (which increased turnover and reduced the governor’s ability to trade favors), an end to the governor’s participation in the assembly, and an end to executive office holding by Burgesses (Rainbolt Reference Rainbolt1967; Webb Reference Webb1979). The executive and legislative functions of colonial government were thus to be separated (Webb Reference Webb1979; Taylor Reference Taylor2002), and the burgesses tied more closely to the common planters (Rainbolt Reference Rainbolt1970). As if acknowledging the legitimacy of grievances behind a mass rebellion were not enough, here was the spectacle of a Stuart king moving for more frequent popular elections.

After the Crown’s intervention, the governor was no longer able to govern in a cartel with the Burgesses. Virginia electoral politics, therefore, took a more popular cast in the 1680s and 1690s (Rainbolt Reference Rainbolt1970). Burgesses became more attentive to the desires and interests of common people, competing for office by proclaiming their opposition to taxes (Taylor Reference Taylor2002). Within 15 years of Bacon’s rebellion, the poll tax enacted by the Burgesses fell from about 100 pounds of tobacco per tithable person to less than 20 pounds (Rabushka Reference Rabushka2010).Footnote ⁶⁰

In short, from the Crown’s point of view, extreme rent extraction by the colonial government threatened the economic output of the colony. This inhibited the flow of revenues from the colony to the Crown. The Crown, therefore, instituted reforms to restrain rent extraction in the colony that put the assembly on the path to developing an independent power base from the governor, and allowed it to check rent extraction from common planters.

English Imperial Governance in the 17th and 18th Centuries. Because of the need for strong governors and high cost of royal intervention, the English Crown faced agency problems with colonial governors across all its New World possessions. Yet the specific agency problem differed across colonies with different economic endowments. This in turn engendered a different strategic Crown response. Thus, the English Crown supported the development of separation of powers in North America, and not elsewhere in the New World (Kammen Reference Kammen1969; North Reference North1990). The model in this article organizes and interprets this historical pattern. This pattern is captured by the relationship between returns to settler investment and Crown support for independent assemblies—the backbone of separation of powers. Roughly speaking, this maps the Caribbean, future United States, and British Canada to Regions I, II, and III in Figure 1, respectively.

First consider returns to settler investment. These were very high in the English Caribbean and West Indies.Footnote ⁶¹ Their economies were dominated by sugar monoculture, with a small number of extremely wealthy planters.Footnote ⁶² Sugar was arguably the most capital-intensive consumer product in the world at the time, so scale economies were high. On the other hand, in the future U.S. states, returns to settler investment were more moderate.Footnote ⁶³ The bulk of export value was from Southern tobacco, indigo, and rice, but Northern timber and fish were significant as well. Scale economies were modest, and white labor generated most of the value through the 17th century (Taylor Reference Taylor2002). Finally, returns to settler investment were relatively low in British Canada (1763–1867).Footnote ⁶⁴ Exports were small, and 90% were from fish, whales, and pelts—lightly capitalized, itinerant enterprises with fewer holdup problems.

Now consider Crown support for independent, representative assemblies. One indication of this is their pattern of creation. In the 17th century, England established 21 separate colonies in North America (14) and the Caribbean/West Indies (7), all with representative assemblies of varying strength. They were created throughout the century as specified in Table 1.Footnote ⁶⁵

TABLE 1.

Assembly Creation, 17th-Century English Colonies

In particular, almost all North American assemblies were created under royal authority. This was not due to any liberal tendency or exportation of existing English institutions: during the 11 years of Charles I’s personal rule without any parliament at all in England (1629–1640), he oversaw the creation of five new assemblies in North AmericaFootnote ⁶⁶ and directed his governor to recognize an existing one (Virginia). Charles II, despite his father’s execution at the hands of Parliament, oversaw the creation of seven new colonial assembliesFootnote ⁶⁷ after the Restoration (1660–1685). Assembly rights in North America were almost always codified in charters, whereas in the Caribbean their recognition was principally by a more legally tenuous royal edict (Kammen Reference Kammen1969). North American colonies used power of the purse to discipline royal governors, and by the 18th century, their primacy was unquestioned (Greene Reference Greene1898; Greene Reference Greene1963). The Crown always asserted its firm control of the imperial hierarchy (Greene Reference Greene1986) but overall allowed independent assemblies to flourish in North America.Footnote ⁶⁸

On the other hand, most assemblies in the Caribbean were created during the Interregnum.Footnote ⁶⁹ The significance of this is that most of these assemblies were apparently emboldened to assert authority by the absence of a king (Kammen Reference Kammen1969); thus, their creation came with a marked lack of royal support. Moreover, most assemblies in the Caribbean were undermined by the Crown and domesticated by the governor after the Restoration, with their powers eroded further in the 18th century (Spurdle Reference Spurdle1963).Footnote ⁷⁰ Royal review of colonial assembly legislation was applied more stringently to the Caribbean than North America (Russell Reference Russell1915). Caribbean governors had a direct hand in drafting colonial tax legislation (Spurdle Reference Spurdle1963) and invoked martial law to govern unilaterally when assemblies failed to heed the Crown’s instructions unquestioningly (Webb Reference Webb1979)—both in marked contrast to North American practice. In addition, the 18th- and 19th-century British acquisitions in the Windward IslandsFootnote ⁷¹ were generally not even extended rights to organize assemblies (Murray Reference Murray1965).

In Canada, British support for assemblies was among the weakest in any New World colony. From conquest (1763) to the Constitutional Act of 1791, Britain did not provide for Canadian assemblies at all; they were specifically excluded from the Quebec Act in 1774. After 1791, the elected assemblies were weak and there was no separation of powers. Laws were passed by the Crown-appointed governor-general and council. The elected assembly’s assent, much less initiation, was not required until the advent of “responsible government” in 1848, at the tail end of British possession (Keith Reference Keith1912).

Thus, Crown support for independent assemblies exhibits a clear but non-monotonic relationship with investment returns—as implied by the model. This pattern cannot be explained by other natural accounts such as colonial settler population or cultural/religious heritage. For example, the white population of the English Caribbean was very high in the 17th century—in 1650, 44,000 whites on half a dozen tiny islands outnumbered whites in all of English North America combined (McCusker and Menard Reference McCusker and Menard1985).Footnote ⁷² Clearly, white settler habitation was not sufficient for Crown support of separation of powers or independent assemblies. As for culture and religion, Upper Canada (Anglican British loyalists) and Lower Canada (French Catholics) were treated approximately the same in the Quebec Act and Constitutional Act (Keith Reference Keith1912). But Maryland, with a large plurality of Catholics in prime tobacco country, had assembly rights guaranteed in its charter, and its assembly convened in 1635.

Other New World Empires. The English Crown was not unique in facing agency problems with its colonial governors, but it was unique in empowering independent, broad-based settler assemblies in part of its New World empire (North Reference North1990). While a thorough comparative analysis of New World imperial governance is beyond the scope of this article, a brief comparison to other notable cases is informative.

The largest and oldest French New World possession was Canada (New France). Its principal export was beaver pelts; agricultural exports to France were negligible (Pritchard Reference Pritchard2007). Pelts were procured by trade with Native Americans. French trading companies with monopoly rights required a small labor force for this purpose. This fact, combined with low agricultural productivity in the St. Lawrence valley as compared to France itself, kept the population small (Pritchard Reference Pritchard2007).Footnote ⁷³ Rent extraction from settlers, therefore, did not undermine the Crown’s interest in the colony. Correspondingly, at no time were independent settler assemblies contemplated; New France was governed by French courtiers and bureaucrats (Eccles Reference Eccles2010). Despite a change in governing personnel, the British did not institute significant changes in institutions upon conquest.

The Spanish New World Empire was larger, older, and more variegated, but the relevant regularities for this article are simple: its value to the Crown did not derive from widespread settler investment; therefore, it was not compromised by excessive rent extraction from settlers. For its part, the Crown did not support independent settler assemblies (Borah Reference Borah1956). The predominant resource in the Spanish New World was silver, extracted with forced indigenous labor (Taylor Reference Taylor2002). The obvious agency problem with governors was to induce truthful reporting of the quantity of silver available and to maximize the amount sent to the Crown.Footnote ⁷⁴ Independent settler assemblies could not address this problem, but mutual monitoring among elites could. This is one way to interpret the high level of elite conflict built into imperial institutions, with the audiencia and viceroy reporting on each other directly to the Crown (Elliott Reference Elliott2007).Footnote ⁷⁵

In short, agency problems with governors were endemic in New World colonies, and we should think of imperial governance institutions as designed in part to mitigate those agency problems. But neither the French nor Spanish New World colonies depended on widespread investment and cultivation by European settlers in the same way that English North America did. This investment is one of the crucial components of the model in this, so it is not surprising, in light of this model, that English/British imperial institutions differed from Spanish and French.