UNIVERSITY OF ILLINOIS LIBRARY AT URBANA CHAMPAIGN BOOKSTACKS Digitized by the Internet Archive in 2012 with funding from University of Illinois Urbana-Champaign http://www.archive.org/details/politicalappoint91129spil Faculty Working Paper 91-0129 Political Economy Series U7 330 B385 1991:129 COPY 2 Political Appointees vs. Career Civil Servants: A Multiple Principals Theory of Political Bureaucracies Pablo T. SpiUer Department of Economics Santiago Urbizto Department of Economics Bureau of Economic and Business Research College of Commerce and Business Administration University of Illinois at Urbana-Champaign Papers in the Political Economy of Institutions Series Susan I. Cohen. "Pareto Optimality and Bidding for Contracts" Working Paper # 1411 Jan K. Brueckner and Kangoh Lee. "Spatially-Limited Altruism. Mixed Clubs, and Local Income Redistribution" Working Paper #1406 George E. Monahan and vijay K. Vemuri. "Monotonicity of Second-Best Optimal Contracts" Working Paper #1417 Charles D. Kolstad. Gary v. Johnson, and Thomas S. Ulen. "Ex Post Liability for Harm vs. Ex Ante Safety Regulation: Substitutes or Complements?" Working Paper #1419 Lanny Arvan and Hadi S. Esfahani. "A Model of Efficiency Wages as a Signal of Firm Value" Working Paper #1424 Kalyan Chatterjee and Larry Samuelson. "Perfect Equilibria in Simultaneous-Offers Bargaining" Working Paper #1425 Jan K. Brueckner and Kangoh Lee. "Economies of Scope and Multiproduct Clubs" Working Paper #1428 Pablo T. Spilier. "Politicians, Interest Groups, and Regulators: A Multiple-Principals Agency Theory of Regulation (or "Let Them Be Bribed" Working Paper #1436 Bhaskar Chakravorti. "Asymmetric Information, 'Interim' Equilibrium and Mechanism Design" Working Paper #1437 Bhaskar Chakravorti. "Mechanisms with No Regret: Welfare Economics and Information Reconsidered" Working Paper #1438 Bhaskar Chakravorti. "Communication Requirements and Strateoic Mechanisms for Market Organization" Working Paper • 1439 Susan I. Cohen and Martin Loeb. "On the Optimality of Incentive Contracts in the Presence of Joint Costs" Working Paper #1445 Susan I. Cohen and Martin Loeb. "The Demand for Cost Allocations: The Case of Incentive Contracts Versus Fixed-Price Contracts" Working Paper #1455 Jan K. Brueckner and Kevin M. O'Brien. "Modeling Government Behavior in Collective Bargaining: A Test for Self- Interestea Bureaucrats" Working Paper »1481 Jan K. Brueckner. "Estimating a Bargaining Contract Curve: Prior Restrictions and Methodolocy" Working Paper #1490 Peter C. Reiss and Pablo T. Spilier. "Competiton and Entry in Small Airline Markets" Working Paper #1497 Pablo T. Spilier. "A Note on Pricing of Hub-and-Spoke Networks" Working Paper #1498 Larry DeBrock. "Joint Marketing Efforts and Pricing Behavior" Working Paper #1500 Frank A. Wolak and Charles D. Kolstad. "A Model of Homogenous Input Demand Under Price Uncertainty" Working PaDer ♦ 1502 Susan I. Cohen. "Reputation, Intertemporal Incentives and Contracting" Working Paper #1511 Lanny Arvan and Antonio Leite. "A Sequential Equilibrium Moaei of Cost Overruns in Long Term Projects" Working paper #1514 Jan K. Brueckner and Pablo T. Spilier. "Competiton and Mergers in Airline Networks" Working Paper #1523 Hadi S. Esfahani. "Reputation, Product Quality, and Production Technology in IDC Markets" Working Paper #89-1525 Hadi S. Esfahani. "Moral Hazard, Limited Entry Costs, and 'Introductory Offers'" Working Paper #89-1526 Bhaskar Chakravorti. "Mechanisms with No Regret: Welfare Economics and Information Reconsidered" Working Paper ♦89-1527 Susan I. Cohen. "Implicit Cost Allocation and Bidding for Contracts" Working Paper #89-1558 Rafael Gely and Pablo T. Spilier. "A Rational Choice Theory of the Supreme Court" Working Paper #89-1559 Rafael Gely and Pablo T. Spilier. "An Economic Theory of Supreme Court Statutory Decisions The State Farm and Grave City Cases" Working Paper #89-1560 No. 29 Rafael Gely and Pablo T. Spilier. "The Political Economy of Supreme Co'urt Constitutional Decisions: The Case of Roosevelt's Court Packing Plan" Working Paper #89-1561 No. 30 Hadi S. Esfahani. "Reputation and Product Quality Revisited." Working Paper #89-1584 No. 31 Jan K. Brueckner. "Growth Controls and Land Values in an Open City." Working Paper #89-1594 No. 32 Jan K. Brueckner. "Tastes. Skills, and Local Public Goods." Working Paper #89-1610 No. 33 Luis Cabral and Shane Greenstein. "Switching Costs and Bidding Parity in Government Procurement of Computer Systems." Working Paper #90-1628 No. 34 Charles D. Kolstad. "Hotelling Rents in Hotelling Space: Exhaustible Resource Rents with Product Differentiation." Working Paper #90-1629 No. 35 Santiago Urbiztondo. "Investment Without Reoulatory Commitment: The Case of Elastic Demand." Working Paper #90- 1634 Bhaskar Chakravorti. "Sequential Rationality. Implementation and Communication in Games." working Paper #90-1636 Pablo T. Spilier. Rafael Gely. "Congressional Control of Judicial Independence: The Determinants of US Supreme Court Labor Relations Decisions. 1949/1987." Working Paper #90-1637 Hans Brems. "Dynamic Macroeconomics: Fiscal and Monetary Policy." Working Paper #90-1640 Lanny Arvan. "Flexibility Versus Commitment in Strategic Trade Policy Under Uncertainty: A Model of Endogenous Policy Leadership." Working Paper #90-1651 David T. Scheffman. Pablo T. Spilier. "Buyers' Strategies. Entry Barriers, and Competiton." Working Paper #90-1674 Richard Arnould and Larry DeBrock. "Utilization Control in HMOs." working Paper #90-1698 Shane Greenstein. "Did Installed Base Give an Incumbent Any (Measurable) Aavantages in Federal Computer Procurement?" Working Paper #90-1718 3haskar Chakravorti and Charles M. Kahn. "Universal Coalition-Proof Equilibrium" Working Paper #91-0100 No. 1 No. 2 No. 3 NO. 4 No. S No. 6 No. 7 NO. 8 NO. 9 NO. 10 NO. 11 No. 12 NO. 13 No. 14 No. 15 NO. 16 NO. 17 No. 18 No. 19 No 20 NO. 21 No. 22 No. 22 No. 24 NO. 25 NO. 26 No. 27 No. 28 NO. 36 No. 37 NO. 38 No 39 No. 40 No . 41 No. 42 BEBR FACULTY WORKING PAPER NO. 91-0129 College of Commerce and Business Administration University of Illinois at Urbana-Champaign April 1991 Political Appointees vs. Career Civil Servants: Principals Theory of Political Bureaucracies Pablo T. Spiller* Santiago Urbiztondo* Department of Economics *William B. McKinley Professor of Economics and Public Utilities, Economics Department, and Professor of Government and Public Affairs; and Graduate Student, Department of Economics. University of Illinois, respectively. We would like to thank Lee Alston, Charlie Kahn, Jim Kuklinski, Peter Nardulli and seminar participants at the lnstituto Torquato Di Telia for helpful comments and suggestions. POLITICAL APPOINTEES VS. CAREER CIVIL SERVANTS: A MULTIPLE PRINCIPALS THEORY OF POLITICAL BUREAUCRACIES by Pablo T. Spiller and Santiago Urbiztondo March, 1991 Abstract: This paper analyzes a multiperiod delegation model with two principals and an agent. The principals have different horizons. One principal lives through the entire game and faces a series of short lived principals. The emphasis of the model is on the choice of the agent, who can be either short or long lived. We analyze the optimal agent type from the perspective of each of the principals. Since a longer horizon agent can be better controlled by the long lived principal, the latter will prefer a long lived agent. The short lived principals, on the other hand, may prefer short or long lived agents. We show that if the two types of principals have opposed preferences in relation to the outcome of the agent's action, then the short lived principals will prefer a short lived agent. Otherwise, the short lived principal alive in the first period will also prefer long lived agents. We apply this model to the determinants of civil service structure, namely, the use of political appointees (short lived agents) vs. civil servants (long lived agents). We see this process as a bargaining problem between two principals, Congress, a long lived principal, and the President, a short lived one. We use this model to explain three aspects of the U.S. Civil Service: its evolution during this century, the high proportion of political appointees in the higher civil service with respect to its counterparts in occidental democracies with parliamentary systems, and the different scope of civil service coverage in local governments in the United States. We argue that.these differences in the organization of civil services across countries and over time can be understood as the result of a game among multiple principals for the control of the bureaucracy, with the main determinants being the extent by which the legislative and executive xidies are aligned in their interests, who is politically more powerful, whether they have different political horizons, and to what extent political parties control their representatives at both the executive and legislative bodies. Wilham B. McKinley Professor of Economics and Public Utilities, Economics Department, and Professor of Government and Public Affairs; and Graduate Student, Department of Economics, University of Illinois, respectively. We would like to thank Lee Alston, Charlie Kahn, Jim Kuklinski, Peter Nardulli and seminar participants at the Instituto Torquato Di Telia for helpful comments and suggestions. "Apart from differences of constituencies and goals, there is institutional rivalry between legislative and executive branches in the United States... Men in politics want to count for something, to have power, and our government is set up in a way that tends to pit legislature against executive. Though the truth is probably not quite so simple, it usually appears that as one waxes the other wanes. In sum, seeking to protect their power and to respond to their constituencies and to further the goals they value, legislature and executive so often face each other as competitors or opponents that it is easy to accept their opposition as fundamental." (Mainzer (1973), page 74.) I. Introduction. All political bureaucracies are not created equal. Some, like that of the modern US Federal system, are composed of tenured civil servants at the lower levels, and of mostly political appointees at the upper levels. On the other hand, political bureaucracies in parliamentary systems have relatively fewer political appointees at their upper levels. Similar differences can be found at other levels of government in the United States, where some political bureaucracies at the local level have relatively more political appointees than others. Finally, a main feature of the evolution over time of the civil service in the United States is the decrease in the proportion of political appointees. The main thrust of this paper is that those differences in the organization of civil services can be understood as the result of a game among multiple principals for the control of the bureaucracy. While the bureaucracy is, in principle, directed by the executive branch of government, the legislature cannot be deterred from being involved in the determination of the policies the bureaucracy ought to follow. Thus, the bureaucracy has to respond to the interests of at least two political principals. These two principals, however, may differ in their political interests as well as in their political horizons. In the United States, for example, legislators tend to last substantially longer than presidents. Also, in the United States both branches are usually not controlled by members of the same political party. This seems to be the case for most other democracies based on the presidential system with strong legislative branches. Patronage, or the use of political appointees at the discretion of the executive, provides the latter with a bureaucracy that is, to a large extent, receptive to its interests. The legislature, on the other hand, prefers a bureaucracy that is more responsive to its own interests than to those of the relatively transient member of the executive. Since legislators stay in power longer than presidents, ; they would be interested in dealing with a bureaucracy that will also outlive the current chief of the < 1 > executive branch. In that way legislators can control the bureaucracy to an extent that the executive branch may not. Legislators, then, will have an incentive to reduce the ability of the executive to appoint its bureaucrats. Civil service provisions perform just that function. 1 This basic insight is at the core of our explanation of the organization of civil services. 2 The following empirical implications arise from our framework: First, if the executive and the legislature are usually composed of individuals with relatively similar preferences (i.e., when both members of Congress and the president share the same goals in terms of industry regulation, distribution of income, etc.), then patronage does not confer the executive with any particular advantage over the legislature, as both would be better off with the use of long lived bureaucrats (i.e. civil servants). 3 If, on the other hand, the executive and the legislature are usually composed of individuals with different and opposed preferences, then patronage confers the executive an advantag' over civil service as it restricts the ability of the legislative branch to control the bureaucracy. In this case, then, we should expect patronage to be more prevalent. Second, among the political systems with separation of powers, those in which the legislators' bargaining power in determining public policy, or interest in the actual performance of the bureaucracy, is relatively small would tend to organize the bureaucracy so that it serves mostly the interests of the executive branch. Thus, patronage should also be more prevalent. These two empirical propositions can explain the three features of civil service organization discussed at the beginning of the paper. That is, the decline over time in the proportion of political appointees in the United States at the federal level, the relatively higher proportion of political appointees in the United States as compared to other occidental democracies, and finally, the different proportions of political appointees among cities of the United States. In this paper we first present some evidence concerning the organizational differences of civil services across countries, US cities and over time. We then develop a multiple principals-multiple period model from which we develop the empirical implications discussed above. < 2 > II. Some Observations on the Organization of Civil Services. Consider, first, the evolution of the civil service in the United States. The first movement away from patronage was the period following the Pendleton Act of 1883, 4 and the subsequent extension of public positions subject to the "merit system" until the turn of the century. The second movement of avil service reform occurred during the 1920s and 1930s. 5 Figure 1 summarizes this evolution. 6 7- 6 - 4 - 2- 1884 1891 1901 1910 1920 1930 1935 1940 1950 1960 1970 197E Year Figure 1 Source: US Civil Service Commission, 1978, and US Department of Commerce, 1976. There are, at least, two views on the original passage of the Pendleton Act. One sees the endleton Act as resulting from an attempt by the Republican Party to limit the benefits the •emocrats were going to enjoy from the "spoils system" after the 1884 elections, as a Democratic ictory was easily foreseeable at that time. The Democrats, it is claimed, supported the project (which ley had presented a year before), because they were interested in an article limiting public employees' pntributions to Republican campaigns (see, e.g. Van Ripen (1958)), Knott and Miller (1987)). The i jther view, recently presented by Johnson and Libecap (1990), is that the evolution of the civil service < 3 > in this period obeys two factors: first, because the federal bureaucracy concerned itself mostly with postal and customs services, big users of those services exerted pressure over their legislators to pass a civil service reform so as to improve their efficiency. 7 Second, patronage, it is claimed, provided negative marginal benefits to the President as the latter had to devote substantial amount of time to the selection process (see Johnson and Libecap (1990)). While both arguments may be correct, they fail to take into account that civil service legislation exists in all democracies, and that many of the civil service reforms across countries happened at similar historical times. Furthermore, since these explanations are based on particular historical accidents, they cannot account for differences across countries in the structure of civil service organizations. Our model explains the timing of civil service reforms in the United States as being the result of the increase in the interest of the US Congress in domestic public policy. 8 ' 9 While throughout the 1800s Congress was involved in legislation concerning commerce, banking and public lands, there were very few legislative actions dealing with Federal regulatory intervention in the economic activity (apart from banking laws). At the time of the Pendelton Act, however, several pieces of legislation were passed which provided a new role for an activist Congress. For example, the Interstate Commerce Act of 1887 was the first Federal regulation of industries (regulating interstate railroad rates); the Sherman Act of 1890 was the first piece of Federal antitrust legislation, and the Meat Inspection Act ol 1890 was the first "consumer protection" legislation. 10 This legislative movement, all happening at the time of the Pendleton Act, shows a Congress that is interested in more than just postal services, customs, banking and defense. The legislators started to care about providing goods to their constituencies. Not only did members of Congress start to be more interested in the activities of the bureaucracy at that time, but their "party lovalty" started to decline. 11 For example, Knott and Miller ■ (1987, page 81) note, that "Party-based voting in the House, which had been at 50 percent in 1897.... decreased during the reform era to 35 percent in 1905 and to 23 percent in 191 1. Legislators < 4 > I ound it necessary to vote with their interest group constituencies rather than with party leaders." : urthermore, "ticket splitting" also greatly increased. The number of counties in which the trailing Republican fell at least 5 percent behind the ticket leader increased from just over 4 percent in 1894 to ust over 40 percent in 1910." 12 As the activities of Congress started to touch individual :onstituencies differently, members of Congress started to bend their party loyalty, and to vote their )articularistic interests. A further indication that the actions of members of Congress became more important at the urn of the century is that their average period in office increased, particularly since the urn of the century. Figure 2 shows that Senators' expected length in office increased rapidly during he turn of the nineteenth century and following Word War II. 13 ' 14 17 - ft 1 16 - pflfc/ 15 - I 01 u o 14 - 13 - t c 12 - 1 III 11 - J ■o 0> u 4' a X LU 10 - 9 - 8 - 7 - 6 - 5 1 4 - r 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 II 1 1 I 1 1 1 1 II I 1 I 1 ! I 1 1 II 1 II 1 1 1 II ! 1 II 1 II 1 1 II 1 I III 1 M 11 1 Ml 1 1 1 1 1 1 1 1 11 M 1 II 1 799 1 B09 1 8 19 1829 1839 1 849 1 839 1869 1 879 1 B89 1899 1909 19 19 1S291 939 1S49 1959 1969 1979 Yeor Figure 2 Until the turn of the century, then, we can see that the legislative branch was essentially ininterested in the actions of the bureaucracy. 15 Our model predicts, then, that until that time the rganization of the bureaucracy should have served the interests of the executive, that is, the < 5 > bureaucracy should be characterized by political appointees. By the turn of the century, though, the legislative branch became interested in the actions of the bureaucracy, and thus it started to restrict the ability of the executive to control it, that is, to designate political appointees. This view, though, is not foreign to students of the organization of the US government. 16 Consider, now, the differences in the composition of the higher civil service in the United States and in the parliamentary democracies of Western Europe, in particular, Great Britain, France and West Germany. Since in parliamentary democracies the executive is composed of, and needs the support from, members of the legislative branch, our model would characterize this type of political system as one where both the long lived principal (i.e. parliament) and the short lived principal (i.e. the prime minister and the cabinet) have similar preferences. The organization of the bureaucracy, then, should serve -as much as possible- their common interests, that is, should be based on long lived bureaucrats, or civil servants. Thus, parliamentary democracies should have a lower percentage of political appointees than democratic countries based on the presidential system. 17 Great Britain has had a parliamentary system of government since the end of the seventeenth century, with parliament's power becoming dominant since the middle of the eighteen century, and as footnote 8 noted, parliament's and the executive's objectives became more aligned from the 1860's on. 18 The executive power is delegated to a prime minister, who can only be installed and remain in office with the support of parliament. 19 In France, during the Third and Fourth Republics, from 1875 to 1958, Prime Ministers (and the presidents in the Fourth Republic) were elected by the Assemblee Nationale (the French parliament). Also, party discipline was weak. With the Fifth Republic, and in particular since 1962, the power of the Presidency has increased. The President is currently elected directly by voters, and has received several new attributions. This shift in the allocation of power coexists, however, with the necessity to form a coalition government that is still in charge of most domestic affairs. Finally, West Germany has gone through many different stages since 1870. From the Bismarck to the Nazi era, the different regimes were characterized by a very strong executive. Since 1949, < 6 > however, West Germany, has had a parliamentary democracy, characterized by strong legislature and a mostly ceremonial President. 20 The unifying characteristic of these countries is that during the period of formation of the current public service agencies, there was not a politically powerful institution that would, according to our model, prefer the bureaucracy to be based on political appointees. Although this situation has changed to some extent in France it is still the case that a parliamentary coalition is required to install a government, and hence at least the part of the executive branch dealing with domestic issues has to reflect Parliament's own preferences. The United States, on the other hand, has had separation of powers since its creation, and as noted above, since the beginning of this century members of Congress have not only been easily reelected, but also they have had a high degree of autonomy with respect to partisan preferences. Given the different systems of government (and indeed of power allocation) described above, it is natural that these bureaucracies are shaped differently. According to our model, the degree of politization should be relatively higher in the U.S. and relatively lower in Great Britain, France and West Germany. Comparative studies of civil service provide support to our theory. Fesler (1983, page 88), for example, provides evidence that the ratio of high-political to high-career officials in the US is 1:5, while in Great Britain and France is 1:40 and in Germany is 1:80. 21 These differences, however, have been previously explained as the result of diverse traditions (the effect of the "unwritten laws"), of different economies of scale, and of public opinion influence. 22 Finally, consider the differences in the organization of the civil services in the United States at the local level of government. The main feature to explain is the differential use of merit systems across cities. 23 When the mayor is elected directly by voters the local form of government resembles the presidential system with separation of powers, since there is no need for the mayor and the council to share the same policy goals. 24 If elected indirectly, the mayor responds to the council and therefore there is no real separation of powers, the local analog to parliamentary systems. We find, as predicted by our model, that, even considering regional differences, there is a negative correlation < 7 > between complete adoption of merit systems and mayors being elected directly across cities over 25,000 inhabitants. In Table 1, below, we estimate the determinants of complete adoption of civil service Table 1 DEPENDENT INDEPENDENT (T-statistic) cs CS CS CONSTANT -.228 (-2.84) -.219 (-2.13) .026 (-.20) MDEWCAO -.436 (-2.59) -.438 (-2.60) -.514 (-2.92) POPULATION (thousands) .00139 (3.01) .00124 d-12) .0015 (3.16) POP. SQUARED 2.08 E-7 (.14) SOUTH -.887 (-4.72) MIDWEST -.313 (-1.96) WEST .37 (1.98) LOG-LIKELIHOOD -291.99 -291.98 -270.22 % RIGHT PREDICTIONS 57 58 66 # OBSERVATIONS 442 442 442 Method: Probit. Year 1963. Variables: -CS (Civil Service) = 1 if all employees are covered by civil service legislation, otherwise, -MDEWCAO (mayors directly elected without CAO) = 1 if the city has this type of government, otherwise, -SOUTH (includes Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Texas and Virginia) = 1 if the cities belongs to the region, otherwise, -MIDWEST (Illinois, Indiana, Michigan, Ohio, Wisconsin, Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota and South Dakota) = 1 if the city belongs to the region, otherwise, and -WEST (Colorado, Idaho, Montana, Nevada, Utah, Wyoming, California, Oregon and Washington) = 1 if the city belongs to the region, otherwise. Source : Municipal Year Book . 1963, 1972 and 1976. The 1963 Municipal Year Book is the most recent to classify cities over 25,000 inhabitants according to complete, partial and nil scope of civil service coverage of municipal employees (there are 611 observations of cities); the 1972 issue has the nearest (in time) series of city governments discriminating by election system and existence of CAO, so it is used (except for a few cities in which the statistic is laking, in which cases it is completed with observations of the 1976 issue) whenever possible, resulting in 442 observations (169 were lost). legislation. We use as explanatory variables the form of government, population and the regional < 8 location of the cities. The estimated coefficient of the form of government is negative and significant, providing some support to our theory. At the mean values of the explanatory variables, a change towards directly elected mayor without CAO decreases the probability of complete adoption of merit systems by twenty percentage points. To summarize, a framework that sees in civil service regulations ways to restrict the ability of the executive to control the bureaucracy is consistent with some stylized facts of the organization of civil services across countries, local governments and over time in the U.S. The model that we develop below is rooted in the transaction cost/agency approach to institutions, where the nature of organizations is the result of the need to control. Our framework combines institutional features, like separation of powers and the need to delegate, with the modern theory of common agency in a multiperiod setup. 26 The organization of political bureaucracies, however, will differ depending on the characteristics of the principals the bureaucracy is supposed to serve. III. The Moael. In this section we develop a stylized model of delegation with two principals in a two-period setting. The principals differ in one significant way: one principal lives for two periods and faces a sequence of two short lived principals, each living for just one period. The risk-averse agent is of two types: she lives either one or two periods. This framework is chosen so as to resemble the institutional features of government: a long lived legislature, a short lived executive and an agent who can be either a political appointee (i.e., short lived) or a career civil servant (i.e., long lived). We are interested here in exploring the type of agent that each principal prefers. If both principals prefer a long lived agent, then bargaining among them for the choice of the agent should result in a long lived agent being chosen. If, however, one prefers a long lived agent and the other prefers a short lived agent, then the probability of selecting a short lived agent increases, a result having implications for our understanding of Civil Service structure. We assume that the decision < 9 > a bout which type of agent to chose results from a bargaining game between the principals alive in period one, which is left unspecified. Long lived principals should prefer long lived agents. Since a long lived principal has an instrument not available to the short lived principal -a reward paid in a later period for a good performance in the earlier period- he may be able to better control the performance of a long lived agent. Although this is well documented in the (infinitely) repeated moral hazard literature (see for instance Radner (1985) and Rubinstein and Yaari (1983)), we show that there are circumstances under which this instrument is indeed used in equilibrium when a sequence of short lived principals interact with another long lived principal and a long lived agent in a two periods model. The delegation problem is of a moral hazard nature, where agents take unobservable actions in two periods (1 and 2), with two feasible outcomes in each period, x 1 €{x5f,x}} and x^x^xj}, where the subscripts denote the period and superscripts L and h denote low and high respectively. 27 Let 4> h (e) be the probability that the random variable x equals x h when the level of effort is e, where <£ h '(e)>0, <£ h '(e)<0 and h (e) represents the probability of a low outcome. 28 Note that, since the equilibrium in period two can be a function of what happened in period one, we need to distinguish the variables in period two according to whether the outcome in period one was high or low. Therefore, e\ will denote effort level e in period two after a high outcome in period one. We can express, then, the expected utility functions of the short lived principal alive in period 1 (called Sj), of the short lived principal alive in period 2 (called S>\ if the outcome in period one was high and S 2 if th( outcome in period one was low) and of the long lived principal (called L) in the following way: EU sl = He x )[ax\-T^ - ^{e^ax^-T^] (D EU ] s2 = <^{e[)[ax\-T^} ♦ h (efylax* -T*] for j = L,h, ( 2 ) EU L = £ 4keffl>x[-Tl L -Rl+ £ ^(e^bx^-T*)], (3) j=h,L k-h,L where a is the marginal utility S a , S 2 and S2 receive from x and b is the marginal utility L receives < 10 > from x; T^CT^T^) is the transfer Sj offers to the agent when the outcome in the first period is x L and x h respectively; T^=(T^,T^) and T^Tft.TiJ) are similarly defined for S* and S^; 29 T n = ( T Li' T Li) ( T L2=( T L2' T L2) and T L2=( T L2' T L2)) is the transf er L off ers in the f irst (second) period depending on the observed outcome x a (x 2 after x\ and x 2 after x\), and R=(R L ,R h ) is the reward L offers in the second period to the agent according to the outcome that occurred in period one. We let b be positive, but a can be either positive or negative (i.e., it could be the case that the short lived principals prefer a low level of effort.) Furthermore, the agent's utility at each period is given by and I ^A2 -V{T^T^R)-e 2i where V(0)=0, V'(T)>0, and V"(T)<0 (i.e., the agent is averse to risk). Thus, the expected utility of an agent in period one is given by eu a^^i) « E U Al * PEU^, (4) where £ represents the discount factor and expectations are taken with respect to (e). /S takes values zero and one: /?=0 corresponds to a short lived agent (called A s ) and £=1 to a long lived agent (called Al). Transfers are assumed to be nonnegative for both x L and x h , so that the individual rationality constraints have to be satisfied in an ex-ante sense. Finally, we assume that the agent does not have access to the credit market. 30 Note that since the transfers and the rewards have to be nonnegative and b>0, the long lived principal will effect zero transfers when x=x L (i.e., T^^Tl^T^^R 1 -^) unless tne agent's individual rationality constraint is binding. Also, we assume that L can commit in period 1 to a reward in period 2 according to the outcome that occurred in period 1. Otherwise R h would be expected to be zero by the agent and it would not affect the development of the game in the second period. The timing of the model is as follows: ■ < ii > 1) L and S : make simultaneous offers of T L1 and R and T sl respectively; 2) A chooses e^ 3) Xj results and transfers are effected; 4) L and S 2 make simultaneous offers of T L2 and T s2 respectively; 5) A chooses e 2 ; 6) x 2 results and transfers are effected. Note that the principals in each period play a Nash game between them. As equations (1) to (3) suggest, the reward R h that L commits to give A L in the second period as a result of a high first period outcome makes e 2 , T L2 and T s2 depend on x : as it is explained later in the paper. Denote e 2 * (e 2 *) the equilibrium level of effort in period two if the outcome in period one was high (low). Therefore, the solution of the model is given by the 10 th -tuple {ej, e 2 \ e 2 \ T* 1? T^, T|? 2 , T^, T{^, T^ 2 , R*}, where e 2 " = argmax EU A2 given T J s2 , T^ 2 and RJ* for j=h,L; ,h* a L m __, • e 1 = argmax EU A given e 2 , e£ , T sl , T L1 , T s2 , T L2 and R ; '.32 (A 2 's problem) (Aj's problem) • _ • T*j = argmax EU sl subject to ej, e^*, e^*, T Ll , T s2 , T L2 , R", EU A (e; e ;) > U A1 +/3U A2 , 33 and Ae^EU^e^+^e^EU^eL) > TJ A2 . (Sj's problem) (IRC A1 ) (Expected IRC^) * <_. i» T J s2 = argmax EU J s2 subject to e 2 , T^ 2 , R J and EU^ej" ) > U A2 , for j=h,L. (S 2 's problem) (IRCJla) Tu. R = argmax EU L subject to ej, e$*, e 2 *, T* ls T^ 2 , 7' L2 , EU A (e;,eJ) > U A1 +£U A2 , and Ac^EU^cSj+rtc^EU^cJ) > U A2 . (L's problem) (IRC A1 ) (Expected IRC^) T{j2 = argmax EU^xj) subject to e 2 ", T J s2 , RJ* and EU^Cej*) £ U^, for j=h,L. (L 2 's problem) (IRC A2 ) < 12 > Thus, in equilibrium, the agent, in both periods, maximizes her utility given the transfers offered and also, in period one, given the equilibrium evolution of the game. The short lived principals maximize their utilities subject to the optimal reaction of the agent in their own period. Finally, the long lived principal maximizes his utility subject not only to the optimal reaction of the agent in the first period, but he also considers in period one his own equilibrium reaction in period two as well as that of the second period agent and the short lived principal. In other words, in period one, the short lived principal alive in period one, the long lived principal and the agent (if she is also long lived) maximize their utilities subject to the (expected) equilibrium evolution of the game in period two. The game in period two is played conditional on what happened in period one. Given our assumptions about the stochastic production function 0(e), we can safely use the First Order Approach to solve this model. 34 The agents' effort levels, e x , e^ and c\ are then determined by the following first order conditions, ^'(e^ViT^+T^-nT^+fiAEU^R*)}-! = 0, (ICC^,) and j (e 1 ){/(4)F(7'S +T&+R1) -V(Tg +* j )]-l) = for j=h,L, (ICCj^,) where LEU^R h ) = 4>\el)V{T£+R h ) + \e*)V(T^ + T^+R h ) - L {e\)V{T^)-\e\)V{T^ + T^) -{ e *-e\)>0 is the change in the agent's expected utility in period two following a high outcome in period one. 35 Note that, since h '(.) is positive and V(.) is concave, an interior solution develops only if , T^+T^ > T& T* h 2+ T*S > T JL ail d T^+Tj, > I*. 96 IV. A Multiperiod Single Principal-Single Agent Model. Before solving the model, let us consider as a benchmark the case in which there is only one long lived principal (with expected utility EU ) and one long lived agent. The purpose of this exercise is to show that there are conditions in which a long lived principal will prefer to deal with a long lived agent (i.e., that it is optimal for the principal to offer a second period reward). Here, the principal's < 13 > problem is to maximize his profits subject to the agent's incentive compatibility constraints in periods one and two. Assuming that the individual rationality constraints are not binding, the problem is given by max EU p (T5,R h ,T5 h ,T^ h ) subject to ICC^, ICC& and ICC^ with respect to T* and R h , where T* h and T 2 come from the principal's second period optimization. Since a positive reward affects the development of the game in the second period, the problem has to be solved backwards. 37 Assuming an internal solution, the first order conditions with respect to e\, l\ h and \\ (the Lagrange multiplier associated with ICC^) characterize the equilibrium in the second period: e\: ^\e\)\Kx\-x\)-Tf\ + \\^\e\\V{Tf + R h )-V(R b )] = 0, if: -\e h 2 ) + \\*\el)V\Tf + R*) = 0, \\: h \e*)[V(T* h + R h )-V(R h )] -1=0. Complete differentiation of this system of equations allows us to obtain the reaction of &\ and T% h as R h changes. We show in the appendix, section 1, that for utility functions displaying nonincreasing risk-aversion the equilibrium level of effort in period two is a decreasing function of R h , whereas T^ increases with R h . We now proceed to solve the problem of the principal in period one. Let e5'=g(R h ) and T; hh =f(R h ), with g'(R h )<0 and f '(R h )>0 for R h >0. Then the problem becomes max EU p (T;,R h ,f(R h ),g(R h )) with respect to e lf T* and R h subject to ICC^, ICC^ and ^C^- 38 Replacing \^ and AJJ (the Lagrange multipliers associated with ICC^ and ICC^) from the first order conditions in the first and second periods respectively, and using the envelope theorem, the first order condition with respect to R h becomes < 14 > t h V'(**\ h h V f (T^*R h ) V ' (R h\ „ h h V\Tf+R h ) . . t\eJ[-l+P*HebL^+t\e*)(fi ' 2 - ' +1- igJ )]^ h (^ h (^ 2 V 1— — /'(* h ) =0. F / (r h } ^/^hj K / (r 2 hh +/? h ) K^rj 1 ) We can now state the following Lemma. Lemma 1: A sufficient condition for equilibrium R h to be positive is that <£ h (e)\e){l+/3)] > l!®L[p-\e)(U/3)] + fiAe)f'{0). V'(T h ) Let £=1. 39 Since V'(0)/V '(T h )>l for T h >0, this condition is satisfied only if h (e)>l/2. 40 Thus, since R h =0 implies that e is determined in a one period principal agent problem, situations in which the solution to the one period problem involve the implementation of a level of effort that results in h (e)0). For simplicity we also assume that the agents' individual rationality constraints are not binding in either period. This would, for example, be the case where both the President and Congress want the regulatory agency to put more effort in price regulation or pollution control, or when the cabinet is composed of members of the legislature. The central aspect of the problem is to determine if, in equilibrium, a positive reward is offered to be paid in the second period following a high outcome in the first. If a reward is offered the t implementability of a given level of effort in the second period is more difficult as the agent is < 15 > wealthier and, because of the concavity of her utility function, has a lower marginal utility of money. On the other hand, the reward also diminishes the cost of implementing an action in the first period. This being the case, the short lived principal alive in period one is benefitted by the use of the reward by the long lived principal, whereas the short lived principal alive in the second period is harmed. Accordingly, S 1 prefers a long lived agent while S 2 would prefer a short lived agent. Note that for a>0, as long as the individual rationality constraints are not binding, T^T^O. Thus, the first order condition to S^s problem (i.e., choosing its first period transfer), results in 1 mX * h W (6') V'iT^T^) Sl \e x ) ' where A sl is the Lagrange multiplier of the agent's incentive compatibility constraint faced by the short lived principal in the first period and is given by *sl ^(e^ajx.-x,)-^] (6//) 4> h (eJiViT^+T^) +pAEU^(R h )] with AEU^(R h ) = 4>\el)V{R h )+4>\el)V{T*l+T^ +R h ) -\e%)V(T$ + T£)-(e%-efe 0. As the above equation shows, d\ sl /dp is negative, meaning that if the agent is long lived (i.e., if /3=1), a given level of effort can be implemented with a lower total transfer (T^+TJ^). That is, S x receives a positive externality from a positive reward R h . As we explained in the previous section, the long lived principal's problem in period one takes into account the evolution of the game in period two after a reward was granted. We solve the model backwards. The equilibrium in period two (after x?J occurred in period one) is characterized by the following system of equations -

b Ae h 2 ) 2 [a(x$-xi)-T£]V<(T£+T£+R*) h/ \e*)[V(Tl*+T™+R h ) - V(R h )] (8) h/ (e%)[V'(T£ + Tf%+R h ) - V(R h )] -1=0, (9) where the first equation results from L^'s problem (combining the first order conditions with respect to Tl2 and e^), the second equation is similarly obtained from S^'s problem and the last one is the (common) first order condition with respect to the Lagrange multiplier (i.e., ICC^)- Here again, as in the case of one single principal, we can obtain the reaction of TJ^, T^ and e^ to changes in R h . In the appendix, section 2, we show that the equilibrium level of effort in period two is a decreasing function of the reward, whereas both T^ and T^, increase with R h . 43 These reactions are considered by the long lived principal, who in period one maximizes ElI^T^R 11 ) subject to ICC^, ICC^, and the optimal responses of S$ and U. As before, let T^=f(R h ) and T^=h(R h ), with f '(R h )>0 and h'(R h )>0 for R h >0. Then the long lived principal problem in the first period becomes max EU L (T},R h ,f(R h ),h(R h )) with respect to e lf T} and R h subject to ICC^, ICCj^ and ICC£ 2 . Since both f '(R h ) and h'(R h ) are positive for R h >0, to prove that there are cases for which this solution is interior requires exactly the same argument provided in Lemma 1 above and we omit it here. Once again, a sufficient condition for R h >0 is that <£ h (e) Proposition 1. The equilibrium to the game described in Section III when L, S 1 and S 2 prefer a high outcome and the agents' individual rationality constraints are not binding involves a positive reward given by L, which in turns leaves L and Sj both preferring to deal with A L and S 2 preferring to deal with A s . VI. Characterization of the solution when the principals' preferences are opposed. In this section we analyze the solution when the principals' preferences are opposed under the assumption that the individual rationality constraints are still not binding. This case is meant to capture the Presidential system with division of powers, where the same party does not have full control over both branches of government. We capture the principals' opposed preferences by letting a be negative and b positive. 44 Note that now T^T^Tl^Tl^R^O. ~^ ne equilibrium in the second period (after a high outcome in period one) is then characterized by the following system of equations -4> (e 2 ) ~ — F/ = ( 10) i> h (e2)[V(T^+R h ) - K(r s h 2 L +# h )] - (e 2 ) - = (n) 4> h (el)[V(T^+R h ) - V(T* 2 L +R h )] principal, who in period one maximizes ElI^T^R 11 ) subject to ICC^, ICC^* and the optimal responses of Sjj and Lj. As before, let T^=f(R h ) and T^=h(R h ), with f '(R h )>0 and h'(R h )<0 for R^O. 46 Combining L's first order conditions, and recalling that f(R h ) and h(R h ) are T^ and T s2 respectively, a corner solution (i.e., R h =0) requires K ; (r L ) v'vr^) [1-A*)(l+0)1 > s —[P- h (e)(Up)] * fit\e)f'{0) * ±Afi*He)-4>\e)yi\0). V{T£) V'{T*) As in the previous cases, h for the agents. Consider first the case when the preferences of both principals are aligned (i.e. a>0). Suppose that U A1 is exactly equal to the expected utility received with the above contract (i.e., IRC A1 is just binding). The question is whether the contract of section V will still be an equilibrium? The answer is yes, as both principals prefer a high action and therefore want the agent to participate in the game. If the individual rationality constraints were not satisfied by the contract in section V, though, the new (Nash) equilibrium will require increasing the equilibrium transfers, and a solution involving zero rents for the agent will develop. Thus, when the objectives of the two principals are aligned, relaxing the assumption that the individual rationality constraints for the agents are not binding can only lead to minor changes in the equilibrium, which would now be further characterized by zero rents being captured by the agent. Now consider a<0 and suppose that IRC A1 just binds when the contract offered is the one given in section VI. It is easy to see that this contract (as long as the solution is interior) cannot be an equilibrium, since Sj (and S 2 in period two) can withdraw his transfer to induce agent nonparticipation, the cheapest way to obtain his favorite outcome (i.e. x=0). Therefore, L has to satisfy the individual rationality constraint with his offer as if S T and S 2 's offers were zero. Since Sj and S 2 offer positive transfers, the agent captures rents when the principals objectives are opposed. 48 VII. Conclusions. This, paper presents a model that can be used to explain the evolution of the U.S. Federal Civil Service, the relatively high number of political appointees in the U.S. compared with other western democracies, and the different proportion of political appointees in different U.S. cities. The key factors are the separation of legislative and executive powers, the effective residence of that power, the expected life in office for Congressmen, and party discipline. We argue that during the nineteen century members of Congress remained in office for short periods of time, and were not particularly interested in Federal regulatory intervention in the < 20 > economic activity (except for banking issues). This being the case, neither of the two principals competing for public service positions were particularly interested in long term agents. As a result, the number of political appointees was relatively high. On the other hand, participation of the Federal State in the economy greatly increased towards the end of the nineteenth century. Similarly, the expected life of members of the U.S. Congress increased since the turn of the century. Thus, the composition of the civil service's labor force can be understood as the outcome of a game played by one short and one long lived principals (the president and Congress respectively), which in turn lead to the extension of civil service legislation over this century. Finally, our model predicts that the stronger Congress is (or alternatively, the more aligned Congress and the Executive are), the fewer political appointees will be in the civil service, a characteristic observed across countries and U.S. cities as well. < 21 > References Adrian, Charles R.: "Forms of City Government in American History," in Municipal Year Book . Washington D.C., 1988. Bernhein, Douglas B. and Michael D. Whinston: "Common Agency," Econometrica, Vol. 54, No.4, 1986, 923-942. Biographical Directory of the United States Congress: 1774-1989. 1989, Washington, D.C.: U.S. Government Printing Office. Cox, Gary W.: The Efficient Secret . Cambridge University Press, Cambridge, 1987. Deutsch, Karl W., Jorge I. Dominguez and Hugh Heclo: Comparative Government: Politics of Industrialized and Developing Nations. Houghton Mifflin Company, 1981. Fainstein, Norman I. and Susan S. Fainstein: "Urban Political Movements," in Urban Politics and Public Policy , Second Edition, ed. Stephen M. David and Paul E. Peterson, Praeger Publishers, 1976. Fesler, James W.: "The Higher Civil Service in Europe and the United States," in The Higher Civil Service of Europe and Canada. Lessons for the United States , ed. by Bruce L. R. Smith, The Brookings Institution, 1983. Fiorina, Morris P.: "Congress and Bureaucracy: A Profitable Partnership," in Bureaucratic Power in National Policy Making . Francis E. Rourke, editor, Little, Brown and Company, Fourth edition, 1986. Heclo, Hugh: A Government of Strangers . The Brookings Institution, Washington, D.C., 1977. Heclo, Hugh: "A comment on the Future of the U.S. Civil Service," in The Higher Civil Service of Europe and Canada. Lessons for the United States , ed. by Bruce L. R. Smith, The Brookings Institution, 1983. Holmstrom, Bengt: "Moral Hazard and Observability," The Bell Journal of Economics, Vol.10, No.l (1979), pp. 74-91. International City Managers' Association: The Municipal Year Book . Chicago, 1963. International City Managers' Association: The Municipal Year Book . Washington D.C., 1972 and 1976. Johnson, Ronald N., and Gary D. Libecap: "Patronage to Merit: Political Change in the Federal Government Labor Force", mimeo, August 1990. Knott, Jack and Gary J. Miller: Reforming Bureaucracy: The Politics of Institutional Choice. Englewood Cliffs: Prentice Hall, 1987. Laffont, Jean-Jacques and Jean Tirole: "The politics of Government Decision-Making: A Theory of Regulatory Capture," Working Paper #506, M.I.T., 1988. Libecap, Gary D.: "The First Consumer Quality Guarantees and the Origins of Antitrust," mimeo, 1990. Maimer, Lewis C: Political Bureaucracy . Scott, Foresman and Company, Glenview, Illinois, 1973. < 22 > Moe, Terry ML: "The Politics of Bureaucratic Structure," in Can the Government Govern?, ed. by John E. Chubb and Paul E. Peterson, The Brookings Institution, 1989. Pfiffner, James P.: "Political Appointees and Career Executives: The Democracy-Bureaucracy Nexus in the Third Century," Public Administration Review, January/February 1987, pp.57-65. Radner, Roy: "Repeated Principal-Agent Games With Discounting," Econometrica, Vol.53, No.5 (1985), pp. 1173-1198. Roberts, Geoff ery and Jill Lovecy: West European Politics Today . Manchester University Press, 1984. Rogerson, William P.: "Repeated Moral Hazard," Econometrica, Vol.53, No.l (1985a), pp. 69-76. Rogerson, William P.: "The First-Order Approach to Principal-Agent Problems," Econometrica, Vol.53, No.6 (1985b), 1357-1367. Rubinstein, Ariel and Manahem Yaari: "Repeated Insurance Contracts and Moral Hazard," Journal of Economic Theory, 30 (1983), 74-97. Spiller, Pablo T.: "Politicians, Interest Groups, and Regulators: A Multiple-Principals Agency Theory of Regulation (or 'Let Them be Bribed')," Journal of Law and Economics, Vol. XXXIII (1), (1990), 65-101. Stahl, O. Glenn: Public Personnel Administration. 6 th Edition, Harper & Row, Publishers, 1971. Stanley, David T.: "Lessons for the United States," in The Higher Civil Service of Europe and Canada. Lessons for the U.S.. ed. by Bruce L. R. Smith, The Brookings Institution, 1983. Stone, Clarence N., Robert K. Whelan and William J. Murin: Urban Policy and Politics . Prentice-Hall, Inc., Englewood Cliffs, 1979. Urbiztondo, Santiago: "Multiple Principals, Cost of Switching Actions and Commitment: Human Capital Investment in the Civil Service," mimeo, 1990. U.S. Civil Service Commission, various issues, Annual Reports . Washington, D.C.: GPO. U.S. Department of Commerce, 1976, Historical Statistics of the United States . Washington, D.C.: GPO. U.S. Department of Commerce, Bureau of Census, 1988, County and Citv Data Book . Washington, D.C. Van Ripen, Paul P.: History of the United States Civil Service . Row, Peterson and Company, Evanston, 1958. Wilson, James Q.: Bureaucracy: What Government Agencies Do and Whv Thev Do It . Basic Books, Inc., Publishers, New York, 1989.2) < 23 > Appendix In what follows we make the following assumptions: Assumption 1: h '(e)/0 L (e), i.e., the hazard rate, increases with e, i.e., -0 h '(e)^ L (e)«?!i h '(e) 2 . 49 Assumption 2: - h '(e)/ h '(e) decreases with e, but not "too fast", i.e., 2>(^ h '(e)^ h '"(e)/<£ h '(e) 2 )>l. 50 Assumption 3: V(.) displays nonincreasing aversion towards risk, i.e., [V(T+R)-V(R)]V"(T+R) > [V'(T+R) -V'(R)]V'{T+R) and [V(T+R)-V{R)]V"{R) < [V / (T+R)-V , (R)]V / (R). 1) Equilibrium in period 2: one principal-one agent- We now start with the problem of the single principal. The problem is given by 51 max Z = {4>\eJ[bx\ + $\e\)bx\ + \e2){bx\-T); h )} + \ed[bx\-T\-R h +\el)bx%+4>\el){bxl-Tl)) * A 1 {0 h '(e 1 )[K(r?) +J SA££/ i J 2 (/? h )] -1 } + 4>\ ei ) \\{4> h '{e\)V{T\ h )-\ } + \ ei ) \\{ h '{e h 2 )[V{Tf +R h ) -V(R h )] -1 }, with respect to e 1 ,T^,R h ,A 1 ,e],T^ h and \[ for j=h,L. We solve the problem backwards. The system of first order conditions for P^ is given in the main part of the paper. Totally differentiating that system, and after replacing [bCx^-x^-T^] and A^ from these first order conditions we have 52 / v - V <^" LVde\ + <^'v'dff = - h/ AV'dR h . The determinant (A) of the system above reduces to 112 V 1 A = h &VV'(3 h - h

h 4> h LW' - 1 I , which is positive by assumption 2. Also, using Cramer's Rule, we obtain 24 > ■tfVV // AKK / (/? h )K // (r? h +/l h ) .3 . , hh h de\ V\Tf+R h ) _i . < 0, dR h dT hh h h " LVLV' * i t AV [AVV // -AV / V / ] + LV LV' 4>\4> h ' h '" -2 h ' ) 2 — = - > 0, dR h A where the last sign results from assumptions 2 and 3. Once the responses in the second period have been determined, the problem in period one can be written as with respect to e 1 ,T5,R h and A T , where LEU^(R h ) = \g(R h ))V(f(R h ) +R h ) -Ae2)V(T$) -(e* -eh > 0. Thus, differentiation with respect to e l5 Tj, R h and A t results in ey h \e{){[b{x\-x\)-T\-R h ]*{EU^-EU^} * A 1 h % 1 )[F(rf) +) SA£:t/JJ 2 (/? h )] = 0. T\: -AeJ * X^ieJV'U*) = 0, * h : -0Vi)[l - h \e*)[b(x* -x^-T^g'iR*) ^He*)/'^)] ♦ \P h \e{){4>\el)V\R*) + 4>\el)V'(Tf+R h ) +4>\el)V'{Tf +R*)f'{R h )) * \\i\ej {^'{el)[V'{T h 2 ^) -V'{R h )} + 4> h \e^)V\T^*R h )AR h ) + h \e^[V(T^*R h )-V(R h )]g / (R h )) = 0, \f. 4>*\e l )[V(T*)+fi*EU^iR h )]-l = 0, where {[b(x5-x5)-Tj-R h ]+(EUj 2 -EUp 2 )} is assumed to be strictly positive (i.e., the principal prefers a high outcome even net of transfers). < 25 > 2) Equilibrium in period two with two principals: aligned interests. Totally differentiating equations (7), (8) and (9), and after replacing [b(x§-x^)-T^] and [a(x$-x^)-T^] from the Ljj and S^'s first order conditions we obtain the following system: $' LVde\ + (jp'v'dT^ + (jp'v'dT^ = -0, AV '=[V '(Tj^+T^+R^-V '(R h )]<0, and V '=V '(TJ^+T^+R 11 ). Noting that the first term on the right hand side of the first two equations is negative from assumptions 1 and 2, it is easy to check that the determinant is negative. Also, 3e^/BR h <0 and 3T^/5R h =aT^/3R h >0 follow immediately. Finally, the long lived principal's problem in period one can be written as max Z = {<^i)[6*i L + <^ 2 L )^ x 2 L + ^ h (*2 L )(^^^ + Ae2)(bx*-f(R h )]} + X u {^\ ei )[V(T^*T^) *fi^EU^iR h )]-l) + 4>\e{) A^l/c^V^ +T^)}-\ } + AejXu^'ie^VifiR*) + h(R h ) + R h ) -V(R h )] -1 }, with respect to e^T^R and A L1 . Differentiation with respect to e l5 T^, R h and A L1 results in e{ ^(e^Kxi-x^-T^-R^^EU^-EU^) * X u * *"(«!)[ Wu+7j) ^AEU^(R h )) = 0, T^: -rtei) + ^u^W^cri^ri) = o, R h : -Ae^l +t\e%)AR h )] - \ u p h \e 1 ){\e%)V\Tu + T s h 2 h + * h )(/'(/* h ) + /i / (/? h ))} + A^^Vi) h ^(eJ)[^ / (7*2 +T s h 2 h + fl h ) -K'CR h )] + K / (r£J + r£ h +/? h )(/ / (/? h ) +*»'(* h ))> = o, A u : * h/ (#( 7 'i^i) + ^^ h )]- 1 =0 - < 26 > Replacing A L1 and a£ 2 from L's first order conditions in the first and second periods respectively, the first order condition with respect to R h becomes V'lTJh + r sl) V ( r Ll + r sl) V ( r L2 + r s2 + R ) Agi)Ae 2 V h S2 h V ^V^W) + rte^rte^CH 11 ) =0. 3) Equilibrium in period two with two principals: opposed interests. Totally differentiating equations (10), (11) and (12), and after replacing [b(xJ-xV)-T^] and [a(x£-x§)-T^] from L^ and S$'s first order conditions respectively we obtain the system below: / a*a*" aJaJ" . -^V'iT^ +R) ^luh" U V"(T^+R) hh 6 h V'(Tl^+R) hT [-/+ * * (2-^ * )}de*+{— K -B l[t * -i]^ h u lwr L h ? + i IB _ h " LVdel + 4> h 'v'(Tl 2 +R)dTl 2 - h ' ' V'(T^ +R)dT^ = -4> h ' LV'dR, Note that assumptions 1 and 2 assure that the first term of the first (second) equation above is negative (positive). Then, it is easy to check that the determinant is negative. It is also straightforward to show that de\/dR b and 3T^/3R h are negative. Finally, 5T^/5R h is positive if and only if 4 b 'v\T s2 +R) A** jW\T s2 +R) / yy^ //'^jyV'^/oy/'' Assumptions 1 and 2 imply that the first two terms are negative and the last one is positive. Consider the two extreme cases. a) h '4> hm /ct) h " = \. Then aT^/3R h is unambiguously positive. ) b) ^ h '<£ h 7tf> h ' 2 =2. Then, since V"(Tj 2 '+R)AV=V '(T^+R^aV ' because we are considering a utility function with constant risk-aversion, 3T^/3R h >0 if and only if IV, or - •hh L2> < 27 > Since V ' (T£5+R)\e l )[bx\ + $\e\)bx\ + $\e\)(bx\ -T^)\*^\e^\bx\ -T^ -R h + \e 2 )bx\ + 4>\e{) A^{/(e 2 L )[F(r L L 2 h ) -K(r s L 2 L )]-l ) + 4>\e x )\^'{e h 2 ){V{f{R h ) +R h )-V(h(R h ) +R h )] -1 }, h r>h with respect to e^T^.R and A L1 , where LEUb(R h )=He%)V{h(R h )+R h )+\e%)y(f(R h )+R h )-^ Thus, differentiation with respect to e l5 T^, R h and X^ results in e : : 4> h \e l )\b{x\-x\)-T\-R h ],{Eul 1 -EU^ 2 ))-\'^\e{)[V{T^ R h : -4>\ ei )[l + h (e 2 V(* h )] * A^^^^^V^^^V^V^^J^V \el)V\T^ +R h )f / (R li ) +4>\el)V'{Ta+R h )h'(R h )) + *u A*i) ^\el){[V\T^ +R h )-V'(T* 2 h +R h ))+V'{T^ +R h )f'(R h )-V l {T^' + R )n )h'{R ]n )}. A L1 : * h \edlV(xlj-V(T%)+pLEU)dR Y, )-\-\ =0. < 28 > ) Footnotes 1. Civil service provisions also have implications for incentives to invest in specific human capital. While these are important considerations, they are not useful in explaining the organizational differences discussed in the text. Thus, we do not consider them here. A model capturing those incentives is developed in Urbiztondo (1990). 2. This view is also held by Wilson (1989), who notes that the US Congress was willing to give up the power to choose or replace public employees, among other reasons, because it was useful to restrict the president's power. 3. Since a long lived legislature will be able to control better a long lived bureaucrat, the executive will benefit from such increased control. 4. The Pendleton Act was the first legal piece protecting public employees from partisan discretion, and reached, initially, about ten percent of the federal employees. See Van Ripen (1958). 5. For example, the Rogers Act of 1924 extended the merit principle to the State Department, and the Hatch Act of 1939 prohibited federal employees from participating in electoral activities. See Van Ripen (1958) and Knott and Miller (1987) for a detailed discussion of the evolution of the U.S. civil service. 6. Note the decline of merit appointments during the New Deal, where the Executive temporarily gained power over Congress. 7. This issue is also recognized by Van Ripen (1958) and Knot and Miller (1987), who note the influence of the "Progressive Reformist Movement," composed of professional middle-class taxpayers, urban merchants, rural populists and northeastern dogooders, who demanded a change towards a more efficient, nonpartisan and professional public service. 8. Similar timing seems to explain the introduction of civil service regulations (i.e. meritocracy) in Great Britain. While patronage through the executive, with the tacit approval of the King, was the norm through the 18th century, meritocracy started to be implemented in the 1 850- 1 860s following the Northcote-Trevelyan Report. This is precisely the time at which the Cabinet and the Parliament became more aligned. See Cox (1987), in particular page 51. See more on this below. 9. From the passage of the Pendleton Act until 1892 each election brought a different party to the White House. The outgoing President, though, used the discretion provided by the Pendleton Act to extend the scope of civil service, in principle "protecting his appointees against the incoming president's appointment power. By 1900, almost half of the federal employees were covered by civil service." ( Knott and Miller (1987), page 44.). This, however, raises the question of why the original legislation delegated this power to the President. We see, however, this delegation as inconsequential, as once the appointees became permanent civil servants, they were subject to substantial control by the legislative branch, independently of who appointed them originally. 10. See Libecap (1990) for a critical view of this legislation. 11. Note that if this were not the case we could not really argue that Congress and the president are two different principals, for in fact the political parties could have been dictating their behavior. This is particularly the case when the party in the White House had also a majority in Congress, something quite common during the nineteenth century. < 29 > 12. This, however, may be the result of the introduction of the Australian ballot electoral system. 13. The data on which Figure 2 is based are constructed in the following way: 1) From the Biographical Directory of the United States Congress, we calculate the number of personnel changes in each Senate and divide that number by the number of changes due every period (one third of the members). Call that ratio the probability of losing the seat (Pl); 2) The expected life in office for a given period is then 6*(l+P w +P w 2 +..)=6/(l-P w ), where P W =1-P L . 3) Taking moving averages of five previous, the current and four posterior periods, we call the resulting number "Expected life in office". The computation of a similar statistic for the House of Representatives would be extremely cumbersome, as the number of representatives from each state varies over time. 14. The reduction in rotation could be the result, apart from improvements in life expectancy, of the passing of a set of rules allowing the development of the seniority system in 1910. 15. That does not mean that they were not interested in the patronage system, as it was a way to compensate political supporters. 16. For example, Mainzer (1973, page 99) says "Patronage gives the chief executive a means of shaping the public service. When the governmental system itself is uncertain or threatened, as under Washington or Lincoln, patronage may be used to recruit men loyal to the regime... When merit service is routine and rigid, patronage (whether or not the appointees are party loyalists) may be used to recruit men that are vigorous, imaginative and committed to administration programs." Moe (1989) argues that interest groups influence Congressmen to limit presidential appointments in his attempts to obtain effective control of the bureaucracy. Fiorina (1986) also argues that Congressmen and bureaucrats enjoy a mutually beneficial relationship throughout time. Finally, Wilson (1989) argues that the notion that the president should administer and hold accountable to him the executive branch agencies was only established at the beginning of this century. Since the services provided at the time did not require highly specialized labor (nor did they provide substantial "on the job learning") and congressmen remained for a short time in office, patronage produced considerably more benefits than costs to legislators. Wilson's argument, however, does not violate our model, as civil service legislation would have been the result of both congressmen's increased length in office and their reaction to limit the control of the bureaucracy by politically more powerful presidents. 17. Note that according to our model this empirical implication is irrespective of which branch is formally (or informally) in control of the bureaucracy. 18. Interestingly, it is in this period that major civil service reforms were introduced in Britain. Cox (1987), page 133, notes that "the local Civil Service patronage of which the member of Parliament disposed declined throughout the nineteenth century. Positions in the revenue, postal, and other geographically dispersed branches of the Civil Service were regularly referred to the recommendation of the local member in the early part of the century. However, patronage in the revenue departments began to decline with Lord Liverpool's renunciation in 1820 of the direct appointment of superior offices in the Customs Service; and the reforms initiated by the Northcote-Trevelyan Report (1853), and furthered by Gladstone's Order in Council establishing open competition (1870), cut back the patronage throughout the Civil Service." See footnote 8. 19. See Roberts and Lovecy (1984) and Deutsch, Dominguez and Heclo (1981) for a detailed analysis of the functioning of political systems in the countries discussed here. 20. Interestingly, the Bundestag has a high proportion of former upper career civil servants. For instance, Roberts and Lovecy (1984), page 134, report that civil servants and permanent officials of parties and interest groups comprise sixty per cent of the Bundestag. < 30 > 21. With respect to the typical european agency, Fesler (1983, page 90) goes on to say that "...it presumably takes into account more long-range and general-interest considerations than does an American Department with its short lived political appointees." Also, Heclo (1977, page 107) notes that, compared to the U.S., he "[does] not know of any other developed Nation in which the executive management of the government's senior civil service is left to the tender mercies of temporary political appointees". Furthermore, "...political appointees are simply uninterested. Effective management of the higher civil service is of immense long-term importance. But nothing will have very much effect during an appointee's own short time in office. Most appointees behave exactly as you and I would if faced with the same incentives: they exploit careerists and do little to build career institutions." 22. For example, reviewing the possible determining factors that influence the different shape of higher civil services, Stanley (1983), page 93, notes that "Great Britain, Canada, France, Germany, and the United States differ in political history, structure, and culture. All but the United States have a parliamentary form of government, with its intertwining of legislative and executive work. Three of the five countries are federal; two are unitary. Transitions among the ruling parties vary in their frequency and in their effects on the different bureaucracies, as does the degree of popular and political regard for the public service.... Finally, the bureaucracies themselves differ from one country to another in their composition (social, ethnic, educational), in their traditional behavior and attitudes, and in their political strength." Pfiffner (1987), on the other hand, while in principle focusing the attention on the same issues as we do here, i.e. the separation of powers, conjectures that lack of party control over both branches of government implies that the civil service will have a higher percentage of political appointees. Wilson's (1989) explanation overlaps with ours, as it stresses the importance of different political systems, although it overlooks the relevance of different time horizons. In addressing the question of different use of political appointments between Great Britain and the United States, Wilson (1989), pages 257-258, argues that "In a word, the answer is the Constitution. That document makes the president and Congress rivals for control of the American administrative system...In Great Britain, the prime minister and parliament are not rivals because the House of Commons has no significant power to supervise, investigate, intervene in, or even obtain answers from the bureaucracy." 23. Although studies of the different systems of local government in the U.S. usually focus on whether those are mayor-council or council-manager ones (where the latter system avoids the separation of powers) we prefer to consider whether the mayor (without a chief administrative officer -CAO-, who is a professional administrator responsible for the daily management of the city, and whose nomination sometimes requires council approval) is elected directly by voters or not (in which case the Mayor is generally selected by the council among its members) as the real measure of whether there is separation of powers or not. Since almost all the cities with mayor without CAO (MWCAO) (78 out of 79 in our sample) have direct elections, the results reported below remain almost intact when we just consider all cities with MWCAO governments. 24. This is especially true for mayors without CAOs, as these are professional. Adrian (1988, page 9), for example, notes that "With the success and continuing expansion of the office of CAO and the ever- wider use of the merit system, differences in professionalism between the two forms continue to diminish." 25. There is at least another feature that can be explained by our model, named that the proportion of political appointees is a negative function of the population of the city. For example, Stahl (1971) reports that "all cities of 250,000 or over (inhabitants), except one, have total or almost total merit- system coverage for their civil services. ... Among cities with 50,000 to 250,000 population, only about 12 percent are without merit systems. ... About 22 percent of those in the 25,000 to 50,000 class are so handicapped." There are two feasible reasons for this correlation. First, excluding the five largest cities in the US, the probability of a direct system of election for the mayor without CAO, i.e., the local < 31 > parallel of presidential systems, decreases with city size. This is supported by the following estimated probit model: Prob(MDEWCAO=l) = $(-.68 -.00504 Pop +.0000074 Pop 2 ) (t-stat) (-5.23) (-2.76) (2.88) Log-Likelihood=-192.64, % Right Predictions=83, 437 Observations, with the critical value of population at which its effect becomes positive being 340.540, well above the mean size in the sample (94 percent of the cities are below this value). Second, the average length in office for councilmen increases with city size: 50 percent of the council members of cities over 500,000 inhabitants remain in office five years or more; for cities in the range 250,000-500,000, 100,000- 250,000, 25,000-100,000, the corresponding percentages are 48, 40 and 38 percent respectively. (Source: Municipal Year Book, 1972.) Our model, then, would suggest that Civil Service provisions should be more prevalent in larger cities, as seems to be the case from Table 1. 26. For models of common agency, see Bernhein and Whinston (1986). See also Spiller (1990) and Laffont and Tirole (1988) for applications. For agency models with multiple periods, see Radner (1985) and Rubinstein and Yaari (1983). 27. Throughout the rest of the paper we refer to actions, transfers and preferences in a very abstract way. We have in mind the following concepts: the action is regulatory effort (which affects the outcome of the regulatory activity over which Congress -L- and successive presidents's -Sj and S 2 - preferences are defined); Congress's transfers (T L1 , T L2 and R) are legislation affecting the agency (especially the Budget, ie., maintaining or increasing the budget of the agency, which translates into a reward), appropriations, and painful hearings and investigations; finally, President's transfers (T sl and T s2 ) are recommendations of pay raises and promotions, defense of the agencies programs, endowing the agency with autonomy, kindly treatment, etc. 28. Unless indicated otherwise, superscripts ' (") denote the first (second) derivative with respect to the argument that follows. 29. T^ 2 , for example, represents S 2 's transfer when the outcome in period two is high. 30. This assumption is admittedly strong and to a certain extent drives the results of the model. It is, however, commonly used in agency models (see Rogerson (1985a)). 31. Think of L as giving A L a check at the end of period 1 that can only be cashed at the end of period 2 as long as A is still working for L. The reward R, then, performs a similar role as a vesting pension plan. 32. The dependence of TJ 2 , T L2 and R* on x : is omitted to simplify notation. 33. In what follows, ICC A1 and IRC A1 (ICC^ and IRC^) denote Incentive Compatibility Constraint and Individual Rationality Constraint of the agent in period one (two) respectively. U A1 (U A2 ) is the reservation value of the agent in period one (two), i.e., the utility she could obtain in case the contract is not signed. 34. See Rogerson (1985b). 35. This change in utility cannot be negative, for otherwise the reward would have the effect of a punishment, and it would therefore be zero in equilibrium. < 32 > 36. This is assumed to be the case throughout the paper. For h '(0) approaching infinity the first two inequalities are necessary and sufficient conditions, whereas the last one is sufficient but not necessary. 37. Note that the equilibrium in period two depends on R h only if x : =x \. Therefore, we concentrate on the equilibrium in period two given that a high outcome occurred in period one. 38. See the appendix for a more detailed statement of the problem. 39. Note that this equation is unambiguously satisfied as /S approaches zero. 40. This is a necessary, but not sufficient, condition due to the fact that f '(0)>0. 41. To interpret this condition, note that a sufficient condition for a marginal R h , evaluated at R h =0, to increase e 2 by more than it reduces the t\ (in expected sense), ceteris paribus (i.e., holding the transfers constant), is that <£ h (e) h (e)=2-e-\ ee(l/2,l) (so that h (e)e(O,l)) and V(T+R)=-c _d(T+R) (where c and d are two positive constants, and d is the Arrow-Pratt measure of absolute risk-aversion). It can then be easily checked that, for instance, letting c=2, d=b=l, and x h -x L =l, the equilibrium level of effort is e=.62 (and then <£ h (e)=.387), which results in a contradiction to R h =0. 43. It becomes obvious here that S 2 prefers a short lived agent, since as R h increases he pays a higher transfer for a lower action. 44. Qualitatively similar results follow if we let a>0 and b<0. In this case, though, none of the short lived principals would prefer a long lived agent. 45. Using assumptions 1 and 2 in the appendix, it can also be shown that the second order condition for an interior maximum is satisfied for the principals wanting a low action as well. The proof is available from the authors upon request. 46. The appendix shows, in section 3, the precise statement of the problem and the derivation o( the first order conditions. 47. Let /3=1. Providing a contradiction becomes more difficult now, since for h <\/2, 4>H '(0)+[V'(T^)/V'(T^)l(l-2^> h )h'(0) can be either positive or negative. Nevertheless, we can show that 0>h'(R h )>[V'(T£)-V'(T^)]/V'(T£) f or all R h >0. (This requires a very tedious procedure, the full derivation of which is available from the authors upon request. The steps to follow are the following: a) Write down the expression for 8T h s \/dR h =h '(R h ) from the system of equations in the appendix; b) Compare h'(R h ) with [V '(T£)-V '(T^)]/V '(T£); and c) Cancel out terms from both sides whenever possible until one side has only negative terms and the other side has only positive terms.) We need only, then, consider two extreme cases: i) h'(0)=0. Then, the equation above reduces to the same one we found in the previous sections. Once again, a contradiction with R h =0 is obtained if 4> h <\/2; ii) h'(0)=[V'(T^)-V'(T^)]/V'(T^). Then, the equation above reduces to 0>A '(0), which is impossible for all h , resulting in R h >0. 48. This result is obtained in a more general framework by Bernhein and Whinston (1986). See Spiller (1990) for a similar treatment of the individual rationality constraints. < 33 > 49. This assumption is standard in the information literature. It basically says that given that a high outcome has not occurred, increasing effort makes its appearance each time more likely. It is respected by many distributions, in particular by ^ h (e)=2-e" 1 (ee(l/2,l)). 50. This assumption is satisfied by the density function referred to in the previous footnote. 51. Note that we already set transfers and reward equal to zero when the outcome is low, that is, R L =Tj"=T2=0. Also, note the use of the first-order approach to the principal-agent problem. 52. For ease of notation we suppress the arguments t\ from <£ h (e), ^'(e^), ^'(e^), and ^"(ej,). Similarly, V'=V'(T$ h +R h ), V"="V(T$ h +R h ), AV=[V(T$ h +R)-V(R)]>0 and AV '=[V '(T5} h +R)-V'(R)]<0. ' 53. We consider a utility function with constant absolute risk-aversion for computational simplicity. 54. If h (e)=2-e" 1 (and ee(l/2,l)) for instance, the condition is only satisfied for low <£ h (e), with the range increasing as the utility function becomes more risk-averse. 55. The upperscripts (') on the lagrange multipliers do not denote derivatives. They just differentiate them from those in the previous section. < 34 > no . 44 Pablo T. Spiller. "A Rational Choice Theory of Certiorari: Hierarchy. Strategy and Decision Cost3 at the Courts" working Paper #91-0110 No . 45 Pablo T. Spiller. "Agency Discretion Under Judicial Review" Working Paper #91-0111 No . 46 Charles D. Kolstad and Michelle H.L. Turnovsky. "Production with Quality Differentiated Inputs" Working Paper #91- 0118 HECKMAN BINDERY INC. JUN95 Bound -To -PU-: N MANCHESTER, iMniANA AR962