Leafy Campus
Huntingdon College: program in Political Science and Public Affairs:
Freedom of Information Laws and Policies:
Research Materials.
Egan, Patrick. "Costly Monitoring:
Using Positive Theory to Analyze the Implications of the Freedom of Information Act" (APSA, 2003).
compiled by Jeremy Lewis, PhD, revised 20 June 2004.
  • Internet Effect, researchers' page
    		•
  • My Professional Writings
    		•

    1. Egan, Patrick. "Costly Monitoring: Using Positive Theory to Analyze the Implications of the Freedom of Information Act" (2003). Online. (17) Paper presented at the annual meeting of the The American Political Science Association, Philadelphia Marriott Hotel, Philadelphia, PA <.PDF>. 2004-06-18

              Publication Type: Conference Proceeding
              Review Method: Peer Reviewed
              Abstract: There is little empirical or formal research conducted by political scientists on
              freedom of information (FOI) laws. Those who have studied FOI laws have portrayed them as
              tools that help legislatures monitor bureaucracies at a lower cost than direct monitoring. In this
              paper, I formally model the effects of FOI laws to argue that such monitoring comes at a cost:
              it brings agency actions to the attention of the public, leading to a shift of policy toward the
              preferences of the median voter. In some cases, this shift is in the legislature’s interest. But in
              others, it is not. Self-interested, rational legislators therefore do not always have the incentive to
              adopt and expand FOI laws. I support this analytical finding with examples from the 40-year
              legislative history of the U.S. Freedom of Information Act, and lay out additional testable
              hypotheses that are yielded by my theory.

              Primary Author email: pjegan@socrates.berkeley.edu


    What follows is text of full paper, extracted from damaged PDF file, missing pages 10 & 11, and with tables and equations reduced to simple, unformatted text.  Horizontal lines divide the original pages.

    Costly Monitoring:
    Using Positive Theory to Analyze
    the Implications of the Freedom of Information Act
    Page 1 (title page only)


    ABSTRACT:
    There is little empirical or formal research conducted by political scientists on
    freedom of information (FOI) laws. Those who have studied FOI laws have portrayed
    them as tools that help legislatures monitor bureaucracies at a lower cost than direct
    monitoring. In this paper, I formally model the effects of FOI laws to argue that such
    monitoring comes at a cost: it brings agency actions to the attention of the public, leading
    to a shift of policy toward the preferences of the median voter. In some cases, this shift is
    in the legislature’s interest. But in others, it is not. Self-interested, rational legislators
    therefore do not always have the incentive to adopt and expand FOI laws. I support this
    analytical finding with examples from the 40-year legislative history of the U.S. Freedom
    of Information Act, and lay out additional testable hypotheses that are yielded by my
    theory.
    INTRODUCTION
    “This legislation springs from one of our most essential principles,” stated
    President Lyndon Johnson upon signing the Freedom of Information Act (FOIA) in 1966.
    “A democracy works best when the people have all the information that the security of
    the Nation permits. No one should be able to pull curtains of secrecy around decisions
    which can be revealed without injury to the public interest.” (President 1966)
    FOIA, which declares that citizens have a right to government records subject to
    certain exemptions for matters like national security, trade secrets, and law enforcement,
    is now considered a key element of the democratic process. But Johnson’s elevated
    rhetoric accompanying his signing of the FOIA elided the fact that his administration had
    never been particularly enthusiastic about the legislation (Archibald 1993). FOIA’s
    origins were in the divided government of the mid-1950s, when a committee in the
    Democratic-controlled House began an investigation by Representative John Moss (D-CA)
    of government secrecy policies in the Eisenhower administration. An 11-year
    campaign for an FOI law followed, with Moss as its unflagging advocate. It wasn’t until
    1964 that FOIA won passage in the Senate thanks to the help of Moss’ Senate allies, and
    an aggressive lobbying effort by the press. But the bill was referred to the House
    Judiciary Committee, where it died unceremoniously. A similar fate was predicted after
    Senate passage in the 1966 session—a prediction so confident that the executive branch,
    known to be wary of the bill, waged little opposition. But Moss crafted a political deal
    that landed the bill in his own subcommittee and cleared its eventual passage by the
    House. “It took the Executive Branch by surprise,” wrote Sam Archibald, who was
    Moss’ subcommittee staff director for the entire 11-year campaign. “Instead of the
    expected easy death, the Congress watchers downtown knew they had a possible
    Freedom of Information Act on their hands.” (Archibald 1993: 729) After his
    Page 2 of 17

    administration had won some important concessions in the legislation, Johnson signed the
    Freedom of Information Act—without ceremony or press coverage—on July 4, 1966.
    Johnson’s wariness about FOIA was reasonable. With his signature, he enacted a
    powerful new tool with which the activities of the executive branch could be exposed to
    public scrutiny. Although he evoked high-flung principles upon its passage, it was clear
    that he—and most of the subsequent occupants of the Oval Office—believed the
    legislation to be contrary to the interests of the executive branch.
    FOIA as a monitoring instrument
    The handful of political scientists who have studied FOIA’s history largely agree
    with Johnson’s assessment. Although access to government documents is considered a
    hallmark—and indeed, a necessary element—of a functioning democracy, political
    scientists argue that at base, these laws are less about democratic principles than about a
    way for legislatures to monitor the bureaucracies of the executive branch. As in any
    principal-agent relationship, legislatures face information asymmetries in their attempts
    to monitor the bureaucracies they create, as agencies inevitably hold more information
    about their activities than do legislators.
    Legislators can tackle this problem directly by undertaking monitoring efforts
    such as legislative hearings and investigations, ordering research to be conducted by the
    General Accounting Office, and requiring agencies to report on their activities.
    McCubbins and Schwartz (1984) call these “police patrol” activities: they are
    “comparably centralized, active and direct.” (166) But these activities have obvious
    opportunity costs: the time that legislators spend directly monitoring bureaucracies means
    less time spent pursuing other policy goals and working for reelection.
    Political scientists argue that, like a profit-maximizing firm, Congress reduces
    these costs by outsourcing the activity to those who have greater incentives to pay the
    monitoring costs—most notably, interest groups. Using institutions such as public
    hearings, proposed regulations and comment periods, sunshine laws, and FOIA, these
    actors monitor the bureaucracy’s activities themselves and alert legislators when their
    interests are threatened. McCubbins and Schwartz (1984) term these “fire alarm”
    activities in contrast to the “police patrols” described above: “Congress’s role consists in
    creating and perfecting this decentralized system and, occasionally, intervening in
    response to complaints.” (166)
    McCubbins, Noll and Weingast (1987) expand upon this argument by claiming
    that Congress creates administrative procedures not only to monitor the bureaucracy, but
    to control it. The coalition that forms to create an agency—legislative committees,
    House, Senate and president—“seek[s] to ensure that the bargain struck among the
    members of the coalition does not unravel once the coalition disbands. Specifically, the
    coalition will seek to…create pressures on agencies that replicate the political pressures
    applied when the relevant legislation was enacted.” Such a coalition does not “preselect
    specific policy outcomes,” but “they will know which interests ought to influence a
    Page 3 of 17

    decision and what distributive outcomes will be consistent with the original coalitional
    arrangement.” (255)
    Among the reforms achieving this effect, argue McCubbins, Noll and Weingast,
    is FOIA, which “limits the ability of an agency to impose a change in policy without
    warning by requiring that, with minor exceptions, all records be made publicly
    available.” Along with the Administrative Procedure Act and the Government in the
    Sunshine Act, FOIA “enable[s] interested parties to learn about any attempt by the
    agency to develop a new constituency or to change policy while it is still on the drawing
    board. This disadvantages the agency by making political intervention possible much
    earlier in the policymaking process.” (259)
    The conclusion drawn by these scholars, therefore, is that FOIA is a relatively
    cheap way for legislators to ensure that the views of the coalition that create an agency
    continue to influence agency policy long after the coalition has disbanded. For example,
    FOIA would ensure that the same political forces that created the Department of
    Education in 1979—that is, those represented by a Democrat-controlled Congress and a
    Democratic president—are continued to be brought to bear upon the department even
    after the legislature and the presidency (and therefore the department itself) are controlled
    by political actors with differing views than the department’s creators.
    In the view of McCubbins, Noll and Weingast (1987), then, FOIA is an institution
    that helps preserve the policy preferences of those who create an executive agency or
    department. How reasonable a claim is this, however? Rather than recreating the
    political views of an agency’s founders, a procedure like FOIA could alternately be
    conceived of as expanding the field of conflict over an agency to the entire political
    spectrum. For sure, FOIA makes it difficult for an agency to develop policy to its liking,
    away from the attention of other political actors. But FOIA makes information about
    executive branch activity available to everyone—not just those sharing the policy
    preferences of the agency’s founders. It is therefore unclear that FOIA gives those who
    hold such preferences an advantage over other political actors. As R. Douglas Arnold
    writes in a response article to McCubbins, Noll and Weingast, “It is easy to argue that
    these techniques affect administrative behavior. It is not so easy to predict exactly which
    interests in society would be advantaged by specific techniques. That is surely true for
    the broad procedural rules which merely open up the administrative process to normal
    political forces.” (Arnold 1987: 285)
    To return to my example, it is hard to make the argument that a procedure like
    FOIA helps keep policy at the Department of Education steered toward the preferences of
    the Democrats who created the agency—or their constituents, notably the teachers’
    unions, who supported its creation. Rather, a more reasonable claim is that FOIA shines
    a bright light on the department’s activity, leaving it open to the inspection of any party
    who is interested in what the department does.
    Thus a rational consequence to predict from FOIA is that, instead of preserving
    the policy preferences of the creators of executive agencies, FOIA drives agency policy
    Page 4 of 17

    toward the preferences of the median voter in the electorate. Agencies are headed by
    appointees of the president, who—following the scholarship of Downs and his
    descendants—faces electoral pressures to provide policies that meet the preferences of
    the median voter. The president can be assumed to have preferences over the policy of a
    given agency that, in the absence of public scrutiny, he would pursue unencumbered. But
    FOIA increases the likelihood that agency policies will become known to the public, and
    therefore leads agencies to think more carefully before they craft policies that will be
    unpopular.
    Take, for example, a story that appeared in the June 5, 2000 issue of USA
    TODAY. The story reported that the Justice Department had sought and received
    permission for a record 880 wiretaps in spying and terrorism investigations in the
    previous year—up from 484 in the year before President Clinton took office (Willing
    2000). This information had been obtained by a privacy advocacy organization through a
    FOIA request and shared with the newspaper. Certainly, it makes more sense to think
    that the effect of the release of this information to the public is to push the department’s
    policy on antiterrorism toward the preference of the median voter (terrorism was low on
    the list of concerns for the median voter in 2000), rather than the preferences of any other
    political actor. Similar conclusions can be drawn about other revelations of information
    under FOIA, such as that the Food and Drug Administration has not studied the effects of
    genetically modified corn (Lappe and Hayden 2000), or that the independent
    investigation of the FBI’s siege of the Branch Davidian compound near Waco, TX was
    projected to cost $11 million (Washington Post 2000). In all of these cases, we can
    presume that the revelation of information causes policymakers in the executive branch to
    think more carefully about the extent to which the median voter will approve of their
    policies.
    The implications for Congressional action
    Given that FOIA most likely drives executive branch policy toward the
    preferences of the median voter, when is it likely that Congress will push for an
    expansion of FOIA, and when will it advocate for its curtailment? I address this question
    formally below, but a brief informal analysis provides us with the intuition that Congress
    is most likely to press for the expansion of FOIA during times of divided government,
    and least likely to do so during times of unified government.
    If we presume that legislators are utility-maximizing agents, they make decisions
    by weighing the costs and benefits of each potential action. They are likely to expand the
    ability of the public to make FOIA requests when the benefits of the expansion to the
    legislature outweigh whatever costs it may incur from the expansion. I assume that
    legislators’ costs and benefits regarding FOI laws depend on three factors: (1) legislators’
    preferences regarding the agency under scrutiny; (2) the median voter’s preferences
    regarding the agency; and (3) the policy pursued by the agency itself. Depending on the
    locations of the ideal points of the legislature, the agency, and the median voter, a policy
    correction in the direction of the median voter can be either beneficial or harmful to the
    legislature.
    Page 5 of 17

    Figures 1 through 3 depict three arrangements of the ideal points of a
    hypothesized legislator, a median voter, and an agency in a unidimensional policy space.
    (Other arrangements of ideal points are possible, but they have outcomes that match the
    three described here.) The preferences regarding agency policy are denoted by ideal
    points XL for the legislator, XA for the agency, and XM for the median voter. (For our
    purposes, we can consider XL to be either the ideal point of the legislature’s median voter,
    or of the median voter in the party in control of the legislature.) In each scenario, the
    agency adopts the policy that matches its ideal point, and then a successful FOI request
    exposes agency activity that is out of sync with the preferences of the median voter,
    resulting in a policy correction that moves agency policy closer to the median voter’s
    ideal point. I make the simple assumption that each political actor’s utility is greatest if
    policy matches its ideal point and decreases as policy moves further away from the
    actor’s ideal point in either direction.
    Page 6 of 17
    Last printed 8/22/2003 10:01 AM
    XL XA XM
    Figure 1: Agency policy is located
    between preference of legislator and
    that of median voter. Policy
    correction results in loss of utility to
    legislator.
    direction of
    policy correction
    direction of
    policy correction
    XA XM XL
    Figure 2: Ideal point for median voter
    is located between legislator and
    agency. Policy correction results in
    gain of utility to legislator.
    Figure 3: Ideal point of legislator is
    located between that of the agency
    and the median voter. Legislator
    gains utility from policy correction
    as long as the distance between the
    median voter’s ideal point and the
    agency’s ideal point is no more than
    twice the distance (x) between the
    ideal points of the legislator and the
    agency.
    direction of
    policy correction
    XA XL
    x
    XM
    2x
    Page 6 of 17

    In Figure 1, agency policy is located between the ideal points of the legislator and
    that of the median voter. Any information that becomes public about the agency will, we
    assume, result a correction of policy toward the ideal point of the voter—and further
    away from the ideal point of the legislator. In scenarios like that of Figure 1, then, FOI
    laws are costly to legislators and we can expect such laws to be resisted—or curtailed, as
    was the case with amendments enacted to FOIA in 1986.
    Now consider Figure 2. In this scenario, the median voter’s ideal point is located
    between those of the legislator and the agency. Therefore, any exposure of information
    that becomes privy to the public through FOIA will, we can assume, lead to a correction
    in agency policy toward that of the median voter. Happily for the legislator, this is a
    correction toward its ideal point, as well. In cases like these, FOI laws are costless to
    legislators: citizens simply become additional monitors of errant agency behavior that
    lead to policy corrections toward the ideal point of the legislator.
    Finally, Figure 3 depicts the scenario in which the legislator’s ideal point is
    located between that of the agency and the median voter. In this case, unless the median
    voter’s ideal point is located more than twice the distance (depicted as the variable x)
    between the ideal points of the agency and the legislator, the legislator gains utility from
    the policy correction, because the resulting policy is closer to its ideal point than the
    agency’s original policy. Note that x is small when the agency and the legislator have
    similar preferences, and larger to the extent that their preferences diverge. The greater
    the distance between the ideal points of the agency and legislator, the more likely the
    legislator is to gain utility from the release of information about the agency’s activity.
    So when might we expect to see passage of—and restrictions on—FOI laws by
    legislators? When the legislature believes that there is a “runaway bureaucracy”—and
    the average voter shares its view. Conversely, we might expect that committee chairs
    with extreme views—i.e., whose ideal points are on the opposite side of agency policy
    than the median voter—might enact more restrictions on FOI laws’ coverage of the
    agencies in their jurisdiction to limit the exposure such agencies have to citizen oversight.
    All of these insights support a simple conclusion: Congress is more likely to make
    information access easier during times of divided government—when it is less likely to
    share the policy preferences of the president, and therefore the agency. Conversely, we
    should expect Congress to make access to government information more difficult during
    times of unified government, when it is likely to have preferences closer to that of the
    executive branch.
    A Formal Analysis
    To gain additional insight on the interaction between Congress and the
    bureaucracy, I now proceed with a formal analysis. I first show how adopting laws like
    FOIA reduce the monitoring costs faced by the legislature in every instance if, as
    McCubbins, Noll and Weingast assert, such procedures reinforce the policy preferences
    of the coalition that created the agency. In this setup, the legislature never has an
    incentive to make it more difficult to access information under FOIA.
    Page 7 of 17

    I then analyze the implications of my claim that FOIA actually drives policy
    toward the preferences of the median voter. Here, the analysis supports my conclusion
    that the legislature has an incentive to expand access under FOIA during times of divided
    government, and has an incentive to cut back access to FOIA during times of unified
    government.
    I first examine the implications for the interaction between Congress and the
    bureaucracy in a hypothesized world where FOIA does not exist—that is, where
    Congress monitors the bureaucracy’s activities directly, per McCubbins and Schwartz
    (1984)’s “police patrol” model. I will build upon these findings in the analysis to follow.
    I formalize the interaction in the following game, based on a popular example in the
    game-theoretic literature known as the “inspection game.” (See, e.g., Fudenberg and
    Tirole 1991, 17.) In this stylized interaction, an agency must decide whether to comply
    with the policy mandate given it by a legislator (who as before may be considered the
    median voter in a legislature, or the median voter of the party controlling the legislature),
    or to “defect”—that is, adopt a different policy closer to its preference. For its part, the
    legislator must decide whether to “inspect” the agency’s activity—conduct hearings,
    investigations, etc.—or to not inspect. Both actors determine their moves
    simultaneously, unaware of the other’s plans.
    How might we expect each actor to approach this game? For the purposes of
    formalization, we make a few standard assumptions. We assume that the legislator
    (denoted L) and the agency (A) have ideal points XL and XA respectively located along a
    unidimensional policy space. (See Figure 4.) The assumption of unidimensionality—
    while not without its critics—can arguably be said to make sense here. Because both
    bureaucracies and Congressional committees specialize on particular issues, we can
    reasonably expect the interaction to involve only one dimension at a time.
    Each actor i’s utility Ui from a given policy X is calculated with the quadratic loss
    function Ui = - (X- Xi)2 – that is, an actor’s utility decreases the farther away a policy X is
    away from the actor’s ideal point. Without loss of generality, we can assume that L’s
    ideal point XL is equal to zero, and that A’s ideal point, XA, is equal to some positive
    number, x (XA = x > 0). In addition, we assume that should L choose to investigate
    agency activity, it incurs costs c, 0 < c < XA, and that should A be caught defecting, it
    incurs “embarrassment” costs e, 0 < e < XA. The legislature gains the same amount, e, if
    it catches the agency defecting. We can consider e the political gain (for L) or loss (for
    A) that comes with the revelation that the agency is not following its policy mandate. For
    illustrative purposes, e is depicted as greater than c in Figure 4, although this need not be
    the case for the analysis here to hold.
    Figure 4: Ideal points of legislator and agency in a unidimensional policy space
    XL
    e c x
    XA
    0
    Page 8 of 17

    The strategic form of this game appears in Figure 5. If A defects and L does not
    investigate, A adopts policy equal to its ideal point XA = x and enjoys utility UA = (x –
    XA)2 = zero, which by construction is the maximum value of the quadratic loss function.
    L’s resulting utility UL equals - (XL – x)2 = - (0 – x)2 = -x 2 . If A defects and L investigates
    at a cost of c, Congress discovers the defection, sets policy to its ideal point of zero, and
    A suffers a loss of e while L enjoys a gain of e. If A complies with L’s policy preferences,
    its adopts policy XL = zero and its resulting utility is - (0 – XA)2 = -x 2 . If L does not
    investigate in this case, its utility is zero. If L unnecessarily investigates, it incurs costs c.
    Figure 5.
    Game I: The Investigation Game
    Legislator
    investigate don’t investigate
    defect (-x 2 - e, -c + e) (0, -x 2 )
    Agency
    comply (-x 2 , -c) (-x 2 , 0)
    To explore plausible outcomes in an interaction like this, we identify Nash
    equilibria—that is, sets of strategies in which neither player has an incentive to
    unilaterally change its strategy given the strategy of the other player. For example, one
    set of strategies we could examine would be (Comply, Don’t Investigate)—where A’s
    strategy is listed first, followed by L’s strategy. But we can easily see that this pair of
    strategies is not a Nash equilibrium. If A knows that L will not investigate, it has the
    incentive to change its strategy to Defect. Subsequently, if L knows A will Defect, then L
    wants to change its strategy to Investigate, and so forth. A glance at Figure 5 reveals that
    there are no Nash equilibria in this game in pure strategies—that is, if a player’s strategy
    is limited to choosing one move and “sticking with it,” no Nash equilibria exist.
    In situations like these, game theorists look for mixed strategy Nash equilibria. A
    mixed strategy is a probability distribution over a player’s pure strategies. For example, a
    mixed strategy for L would be to investigate ¼ of the time and not investigate ¾ of the
    time. In mixed-strategy Nash equilibria, the players’ strategies are best replies to each
    other because each strategy makes the other indifferent between its pure strategies,
    meaning that the actors have no incentive to unilaterally change their course of action.
    In the mixed-strategy Nash equilibrium of this game, then, L will investigate just
    often enough such that A obtains the same expected payoff regardless of whether it
    chooses to comply or defect. Formally, L makes A indifferent between its two strategies
    by investigating with probability p such that:
    Page 9 of 17

    pages 10-11 damaged in PDF file; continue with page 12.

    In the previous game, I assumed that if A was caught defecting, policy reverted to
    L’s ideal point XL. Now, I assume that regardless of the policy chosen by A, an
    inspection by L or a successful revelation of information through FOI results in policy
    being set equal to M, the ideal point of the hypothesized median voter, XM. For this
    reason, I assign A’s “comply” policy to be M, as opposed to XL = zero as I did in the
    previous game.
    I examine two games that incorporate these assumptions. In the first game, as
    before, only “police patrol” monitoring exists; in the second game, A is subject to
    monitoring by an outside agent as happens under a law like FOIA. I am interested in
    calculating the expected value L can obtain from both games, and determining under
    what conditions L is likely to favor one game over the other. In both games, A decides
    whether to “defect” by adopting policy XA or to comply, by adopting policy M.
    Simultaneously, L decides whether or not to investigate A’s behavior.
    The standard forms of the two games are depicted in Figures 6 and 7. Figure 6
    depicts the payoffs in the “police patrol” monitoring scenario. Note that the payoffs are
    similar to those shown in Figure 5, except that M has been substituted as the value to
    which policy reverts should there be an investigation, rather than XL = zero as was the
    case in the first game.
    Figure 6.
    Game II. The Inspection Game with a Median Voter
    Legislator
    investigate don’t investigate
    XA -(x-M)2 -e, -M 2 -c + e 0, -x 2
    Agency
    M -(x-M)2 , -M 2 -c -(x-M)2 , -M 2
    A pure Nash equilibrium exists in this game, but only for certain values of the
    parameters. (XA, Don’t Investigate) is a Nash equilibrium if -M 2 -c + e < -x 2 , as neither
    player has the incentive to change strategies unilaterally given the other player’s strategy.
    We would expect this inequality to hold in situations where x, the distance between XL
    and XA, is relatively small compared to the distance between XL and XM –in other words,
    where the legislature and the agency hold relatively similar policy preferences compared
    to that of the median voter. One case where we might expect to see such an arrangement
    of ideal points would be where the legislature and the presidency were controlled by the
    same party.
    For all other values of the parameters, pure Nash equilibria do not exist. As
    before, if L chooses the pure strategy Investigate, A then wants to “comply”—that is,
    choose M. This leads L to want to forgo investigating, and so forth. A Nash equilibrium
    Page 12 of 17

    exists only in mixed strategies, where both players choose probability distributions over
    their pure strategies to make the other indifferent between their pure strategies. In
    equilibrium, L plays Investigate with probability (x-M)2 / [(x-M)2 + e], while A plays XA
    with probability c / (e + x 2 – M 2 ). (See the Appendix for these calculations.) These
    produce the expected values for the game for A and L:
    EGAMEII(UA) = -(x-M)2 and EGAMEII(UL) = [ce / (e + x 2 –M 2 )] – M 2 – c
    We are particularly interested in the value of this game for L, because we want to
    examine if L can improve upon this value by adopting an FOI law. Note that L’s utility is
    the utility it gains from the median voter’s ideal point ( -M 2 ) minus the costs of direct
    monitoring (c), which is offset by a fraction ce / (e + x 2 –M 2 ) that is a function of all the
    parameters in the game. Notably, amount of the offset rises, ceterus paribus, as the
    distance between XA and M decreases. We will return to this observation shortly.
    We now turn to Game III, which incorporates an FOI law. In this game, it is
    assumed (unrealistically, but for illustrative purposes) that an FOI law leads to the
    disclosure of A’s policy to the public in each and every case. A glance at Figure 7 shows
    that, in this case, L has no incentive to ever investigate A: regardless of whether L does
    so, A’s behavior will become known to the public and policy will be corrected to the ideal
    point of the median voter. L therefore avoids incurring unnecessary monitoring costs,
    and Don’t Investigate is a dominant strategy for L. For its part, A never has the incentive
    to adopt its ideal point by playing XA, for policy will simply revert to M regardless.
    Therefore, this game has a Nash equilibrium in pure strategies: (M, Don’t Investigate).
    The expected value of the game is simply the payoffs for both players in the lower right
    hand corner of Figure 7: -(x-M)2 for A and -M 2 for L.
    Figure 7.
    Game III: The Inspection Game with a Median Voter and an FOI Law
    Legislator
    investigate don’t investigate
    XA -(x-M)2 –e, -M 2 -c + e -(x-M)2 -e, -M 2 + e
    Agency
    M -(x-M)2 , -M 2 -c -(x-M)2 , -M 2
    We now turn to the question: when does L prefer Game III (which I will call the
    “FOI game”) to Game II (the “direct monitoring” game)? That is, under what conditions
    is L likely to favor the adoption of an FOI law? A reasonable conclusion is that L favors
    the game from which it gains the greater expected value. Hence, L prefers the FOI game
    to the direct monitoring game if:
    E(UL[FOI game]) > E(UL[direct monitoring game])
    Page 13 of 17

    -M 2 > [ce / (e + x 2 –M 2 )] – M 2 – c
    0 > [ce / (e + x 2 –M 2 )] – c
    c > ce / (e + x 2 –M 2 )
    1 > e / (e + x 2 –M 2 )
    x 2 –M 2 > 0
    x 2 > M 2
    The calculations show that L prefers an FOI law to direct monitoring when x 2 –
    the squared distance between its ideal point and A’s ideal point—is greater than M 2 —the
    squared distance between its ideal point and that of the median voter. In other words, the
    greater the agreement L is with A on the agency’s policy compared to L’s distance from
    the median voter, the less likely L is to want an FOI law. This is most likely in cases
    where the legislature and the agency (and thus the presidency) are controlled by the same
    political party.
    Figure 8. Circumstances under which
    a legislature prefers direct monitoring versus an FOI law
    If XM is located here, L prefers direct monitoring
    -x
    XA XL
    x 0
    If XM is located here, L prefers an FOI law
    Thus, if we make the reasonable assumption that the introduction of an FOI law
    leads policy to be corrected toward the ideal point of the median voter, we see that such a
    law is not costless a the legislature. In particular, to the extent that the legislature and the
    agency share similar preferences over the direction of the agency’s policy, the legislature
    has less incentive to adopt or expand FOI laws. In these cases, the legislature wants to
    avoid a revelation of agency policy to the public. Such a disclosure leads to policy
    Page 14 of 17

    correction toward the preferences of the median voter, and thus a relative loss of utility
    for the legislature.
    Testable Hypotheses
    My theory yields the testable hypothesis that, all things being equal, Congress
    should expand FOIA access during times of divided government and curtail it during
    times of unified government. Notwithstanding the adoption of FOI in 1966, the major
    statutory expansions and restrictions to FOIA have followed this pattern: in 1974, the
    Democratic-controlled Congress expanded FOIA access over a veto by Republican
    President Gerald Ford. In 1986, leaders in the Republican-controlled Senate successfully
    spearheaded Republican President Ronald Reagan’s efforts to make access to information
    under FOIA more difficult. And in 1996, a Republican-controlled Congress substantially
    expanded information access under FOIA under Democratic President Bill Clinton.
    Recently, the Republican-controlled Congress and the administration of Republican
    President George W. Bush have moved to curtail access to information under FOIA.
    I should note that I am not positing that this is the only mechanism governing the
    expansion and contraction of FOI access: for example, Clinton ran on an explicit promise
    to make government more open and transparent, and regardless of whether it was in his
    interest once he was elected, he carried out that pledge while in office.
    To thoroughly test my hypothesis, in the future I plan to examine the expansions
    and restrictions put on FOIA in each Congressional session. (FOIA provisions are
    usually included in a half-dozen or so enacted bills per session.) Again, I expect
    expansions to take place under divided government and restrictions under unified
    government. A more fine-grained test of my hypothesis would be to examine expansions
    and curtailments at the committee level. Regardless of partisan control, I would expect
    areas of policy in which the administration and the relevant Congressional committees
    are in relative agreement to exhibit less FOIA access than for policies where there is
    substantial disagreement.
    Page 15 of 17
    REFERENCES (page 17)
    Archibald, Sam. 1993. “The Early Years of the Freedom of Information Act—1955 to
    1974.” PS: Political Science and Politics 26 (December): 726-731.
    Arnold, R. Douglas. 1987. “Political Control of Administrative Officials.” Journal of
    Law, Economics and Organization 3 (Fall): 279-286.
    Fudenberg, Drew and Jean Tirole. 1991. Game Theory. Cambridge, MA: MIT Press.
    Lappe, Marc and Tom Hayden. 2000. “Label genetically altered food.” Los Angeles
    Times April 9. Home Edition, p. M-5.
    McCubbins, Mathew D. and Thomas Schwartz. 1984. “Congressional Oversight
    Overlooked: Police Patrols versus Fire Alarms.” American Journal of Political
    Science 28 (February): 165-179.
    McCubbins, Mathew D., Roger G. Noll, and Barry R. Weingast. 1987. “Administrative
    Procedures as Instruments of Political Control.” Journal of Law, Economics and
    Organization 3 (Fall): 243-277.
    President 1966. “Statement Upon Signing Public Law 89-487 on July 4, 1966.” cited in
    “Attorney General’s Memorandum on the Public Information Section of the
    Administrative Procedure Act.” available online:
    http://www.justice.gov/04foia/67agmemo.htm
    Washington Post. 2000. “Washington in Brief.” March 15. Final edition, p. A20.
    Willing, Richard. 2000. “Wiretaps sought in record numbers,” USA TODAY, June 5.
    Final edition, p. 1A.
    Page 17 of 17