Displacement, Not Obstruction: Why Insecure Leaders Need Not Fear Free Media

The manipulation of information is a standard tool precarious leaders might turn to in order to solidify their rule. The ob jective of information manipulation is to influence the behaviour of citizens, dissuading them from making demands of the government. We use the term manipulation here broadly, in reference to any effort the state may exert to prevent the learning and updating necessary for citizens to hold the government accountable. The bulk of the literature on information manipulation focuses on the suppression of traditional media by autocrats looking to quell dissent and consolidate power (Chen and Xu, 2017; Guriev and Treisman, 2019; Guriev and Treisman, 2020; Little, 2017; Shadmehr and Bernhardt, 2015). This behaviour is certainly not exclusive to autocrats; the seminal work of Besley and Prat (2006) studies how a “cozy” media can help entrench corrupt or incompetent incumbents in an electoral context.

The contemporary media environment is characterised by a now dizzying array of possible information sources for the citizen-consumer. This proliferation in media sources appropriately has consequences for how political elites attempt to manipulate information. For example, movements like the Arab Spring initially brought enthusiasm for social media by liberalising information (Diamond, 2010; Howard and Hussein, 2013; Shirky, 2009; Shirky, 2011), but the literature has quickly shown autocrats to be quite savvy in manipulating social media to entrench themselves instead (see Keremoglu and Weidmann 2020; Zhuravskaya et al. 2020 for reviews).

How does the proliferation of media options for the citizen affect the state’s incentive to manipulate information? In this paper, we argue that states should not always be so quick to tighten the reins on emerging media. Instead, the presence of small, independent information sources may actually be a boon for a leader’s ambitions of controlling information. The driver of this effect becomes clear when the objective of states is understood as the “displacement” of citizen attention rather than the outright obstruction of information.

When there exists a particularly informative or easily accessible source of damaging information, states may wish to prevent citizens from paying attention to that particular media source. Entirely cutting off a source of information may be difficult or costly (Gehlbach and Sonin, 2014; Shadmehr and Bernhardt, 2015). However, we argue that states do not need to exert such an effort to succeed in their aims. States instead simply need to make a particular source of information sufficiently difficult to access that citizens will, of their own volition, choose to divert their attention and listen to alternate sources of information instead. Roberts (2018) might call this the “friction” that impedes citizens from critical information; the displacement effect thus reduces the amount of effort states must exert to succeed in generating sufficient friction. In this manner, the proliferation of media sources “subsidises” efforts to further manipulate information.

We present our arguments through a formal model where citizens have the opportunity to learn about the state of the world from multiple sources of information. The government wishes to prevent citizens from learning from these information sources whenever possible as an informed public is more likely to make costly demands of their government. Our model gives states the option of achieving this by exerting costly effort—which may represent investments into propaganda or censorship regimes, among other things—to obstruct particular sources of information. As earlier advertised, states may find this option of exerting effort to be cheaper when information alternatives are available. Illustrating this effect is the main result of this paper.

Substantively, the contribution of our paper is closest to that of Trombetta and Rosignoli (2021) who claim to be the first to argue that the diversification of media can encourage the political manipulation of information. Their mechanism is entirely different to ours, focusing instead on a provider-side explanation where a fractious media is less able to financially resist capture by politicians. We build on this finding by offering our consumer-side mechanism of displacement as another factor contributing to the same relationship.

The displacement effect emerging from our model contributes an element previously overlooked in the literature on political manipulation of information. This effect comes to light due to how we give citizens both multiple sources of information to learn from and a limited capacity to pay attention to them. Our paper is thus an application of interplay between costly information gathering and information manipulation. Such dynamics are well studied by theorists (Bizzotto et al., 2020; Kamenica and Gentzkow, 2011; Matyskova and Montes, 2023; Wei, 2021), and we here provide a simple application for the politics of information. In doing so, we hope to break new theoretical ground that is applicable for both the literatures on the manipulation of traditional media (Besley and Prat, 2006; Chen and Xu, 2017; Guriev and Treisman, 2019; Guriev and Treisman, 2020; Little, 2017; Petrova, 2008; Trombetta and Rosignoli, 2021; Shadmehr and Bernhardt, 2015) and the manipulation of social media (Deibert, 2015; Edmond, 2013; Gunitsky, 2015; Weidmann and Rød, 2019).

Without information, citizens may be unsure of when action is needed to keep elites responsive and accountable. States thus have an interest in manipulating media sources to prevent citizens from learning politically relevant information. This manipulation can take many forms—propaganda, censorship, media capture and so on. While these different tools can have specific and nuanced differences in their use, they also share the same general feature of making it difficult for citizens to grasp the true state of the world, and in turn give governments more leeway to act as they please. We review a selection of models on traditional media manipulation and how their design captures these tools.

First is censorship. Shadmehr and Bernhardt (2015) model a government’s choice to censor news. A highlight of their model is the ability for news to vary in how damaging it is to the government. Through this, they find that governments will never censor all bad news because, among other reasons, a complete void of news is clear signal to voters that censorship is occurring. Instead, the government will leave some less severe bad news uncensored to keep citizens unsure over the extent of the censorship regime. A mechanism that blocks truthful signals is thus the key characteristic of censorship.

Second is propaganda. Chen and Xu (2017) study the interaction of policy concessions and information manipulation. In their model, citizens vary in their individual characteristics and are unsure of the preferences of other citizens. They are thus also uncertain if action against the government should be taken. The government has two tools through which they can appease citizens and prevent action: implementing actual policy reform or fabricating information to suggest the status quo to be popular. The two choices are substitutes that achieve the same immediate result. However, a government that is known to occasionally concede reform—and thus acknowledging the unpopularity of the status quo—is more likely to be believed when it does lie, creating an incentive to combine the two tools. Authoritative media sources in Chen and Xu (2017) are directly at the whims of the government and can be used to spread misinformation at will. The ability to send false signals thus is the key characteristic of propaganda.

Third is capture. Besley and Prat (2006) study this in an electoral context. They design a model where the media can be “captured” through the use of transfers that entice them away from truthfully reporting information damaging to an incumbent’s reelection prospects. The media is hence an active “player” in Besley and Prat (2006) and must be co-opted through payments for them to be manipulated by the state. Transfers are thus the key characteristic of manipulation through capture. Other papers studying the capture mechanism explicitly include Petrova (2008) and Trombetta and Rosignoli (2021).

Given that existing works have shown the viability of different media manipulation tools to achieve similar results, we are comfortable in remaining agnostic on the exact tool in play in our model. Indeed, works such as Guriev and Treisman (2020) and Kocak and Kibris (2023) consolidate several of these tools and use them as substitutes towards the same end of obfuscation at a cost. Following these lessons, we allow the state to generically exert effort in media manipulation, and greater effort will make it increasingly difficult for the citizens to learn information from traditional media. This effort can be interpreted as the cost of generating propaganda, maintaining a censorship regime, making transfers to capture a source, or some combination of such tools.

A handful of papers attempt to capture the growth in number of media outlets in the design of their models. Edmond (2013) takes a global games design to demonstrate how increasing the available number of information sources weakens the government’s ability to hide damaging information. More recently, Kocak and Kibris (2023) focus on social media and model them not as independent sources but as a tool for traditional media to reach a greater audience.

Our main displacement result does not require any intricate design in capturing the variety of media sources available to citizens. Our baseline model simply offers citizens different sources of information to which they can pay attention. These information sources provide citizens the opportunity to learn the true state of the world, which may or may not require policy action on the part of the state.

As enacting policy change is costly, states will normally prefer to not act. However, citizens can organise to demand the state take action should they learn information that suggests action to be the correct course. In turn, the state may wish to suppress the ability of citizens to learn in order to avoid the costs of action. Our model offers the state the opportunity to exert effort in order to discourage citizens from paying attention to a media source. The greater the effort exerted, the greater the obstruction. The state is able to specify the level of effort—and hence the level of obstruction—directed at the particular media source.

Solutions to the model that we focus on suggest that states only need to exert partial effort in order to succeed in turning citizen attention away from a particular information source. This is because as a particular source becomes more difficult to learn from, citizen attention becomes “displaced” onto other sources. Achieving this displacement thus lowers the critical effort threshold for the state’s media manipulation.

Thus, the availability of alternative information sources may not actually be detrimental to a state looking to control information. The displacement effect can benefit the state as (1) it may force citizens to rely on less accurate sources for information, and (2) it achieves this at a cheaper cost (or, equivalently, with less effort) than would be necessary to entirely obstruct a source. Our main result thus suggests that a more robust media environment may actually give states more options—options that can be considered particularly less overt and more subtle—in their battle to control information.

Despite this, our model does not conclude the proliferation of media as a negative. As the models will soon show, citizens can still benefit from greater accountability mechanisms as the number of information sources grows. The punchline of the models, however, is that this proliferation of media is simultaneously not necessarily a disutility to leaders looking to entrench themselves. Thus, we unintuitively find that insecure leaders may not need fear more media.

We believe the manipulation of media is relevant to both democratic and non-democratic contexts. We hence abstract from the exact mechanism of responsiveness through which states are moved by citizen demands to capture both electoral and non-electoral pressures. Certainly, autocracies can have a greater incentive to manipulate the media; the greater scarcity of information that some societies living under autocracy suffer from makes the manipulation of remaining sources more meaningful. But this should be understood as a matter of scale. There can be specific subjects for which the public is particularly aloof to for which even democratic governments may find it possible to meaningfully bottleneck information by manipulating media sources. A key example may be foreign policy, where it is traditionally argued that the public cares little for under normal circumstances (Rosenau, 1964), but can be activated when sufficiently informed (Baum and Potter, 2008; Baum and Potter, 2015; Hyde and Saunders, 2020; Robinson, 2000; Robinson, 2002).

A Leader and n citizens play a one-shot game of accountability. We study two variants of a simple game to demonstrate what happens to the state’s incentive to manipulate information as an independent media source emerges. Let there be a one-media-world (1MW), where citizens only have one media source from which they can learn about the state of the world, and a two-media-world (2MW), where there are two. In both worlds, media source 1 is vulnerable to state interference. Media source 2, if it exists, is independent from state interference. We index media sources by j .

Each citizen may choose to pay attention to however many media sources as they wish to. Paying attention is costly and each media source j will incur the citizen x_j ≥ 0 if they choose to listen to it. The cost component x¹ is endogenously decided by the Leader, but as our assumption is that source 2 is free and independent, we let x₂ = 0 in these initial set-ups.

After citizens have chosen their information source(s), Nature draws the correct policy. With probability $p∈(0,12)$ the correct policy is Act and with 1 - p it is Not Act. This restriction on p is present to make citizens typically prefer inaction absent new information as inaction should always be perceived to be the more likely correct choice before any learning occurs.

When Act is drawn as the correct policy, there is some probability that the media sources receive an always-accurate-signal that reveals Act as the correct policy. Let q_j ∈ (0, 1) represent this probability for media source j. The draw for the signal for any particular source is independent of the draw for other media sources, if any. If the signal is received by a particular media source, then all citizens who are paying attention to that same source also receive the signal. All probabilities are common knowledge.

In the two-media-world, let q₁ > q₂ . This reflects how a traditional and established media outlet is more likely to provide crucial information than a more newly established, emerging media source.

At this point, a reader may interject and disagree with our design as it appears to say that emerging media is less likely to report against government than established media. Such a design would be contrary to the typical role of new, independent media in literature where they are engines of anti-government criticism (Guriev and Treisman, 2022; Peisakhin and Rozenas, 2018).

This would be a mis-reading of our design. The parameters q_j only speak to the objective “quality” of a media source prior to any government intervention to manipulate them. That is, our parameterisation restriction (q₁ > q₂) only claims that—in the absence of government manipulation—more established media have the better resources and opportunity to uncover and report antigovernment news. Our design then endogenises the possibility that traditional media sources become effectively less capable of reaching citizens with critical information through the active choice of manipulation.

To describe this more concretely, it might be that the journalists at source 1 regularly discover government wrongdoings through their extensive resources (high quality q₁), but are prevented from publishing on the topic due to government pressures on the publication ownership (equilibrium behaviour later seen in propositions 2 and 3). Meanwhile, source 2 only occasionally comes across government wrongdoings (low quality q₂), but is not hindered in reporting what they know (also seen in later equilibria). These possibilities will partially recapture standard assumptions of the role of new, independent media endogenously through choices during the game.

After all signals have been revealed, each citizen chooses between Demand and Not Demand. This is in reference to whether the citizens pressure the Leader to Act on the particular issue. After demands are made, the Leader chooses between Act or Not Act. The leader prefers to Not Act over Act. However, choosing Not Act can potentially incur a cost if the citizenry is demanding action. The greater the number of citizens making demands, the greater this cost.

Making demands of governments is costless for citizens in our model. While this may not always be reflective of reality, it suffices for our purposes as the effect of action costs on collective action problems is straightforward and well-studied. When this cost is particularly high—regardless of uniformity or variance across i—then governments may not even need to impede information gathering, as citizens may not succeed in organising their demands. Thus, our model speaks especially to situations where either when this collective action can be solved or it is not particularly significant. The cost of Demand is thus suppressed from our model for parsimony.

Citizens want the correct policy to be implemented, but they also wish to minimise the costs they have to pay for information. Let α =1 when the correct policy is chosen and α = 0 when it is not. Let β_i,j = 1 when citizen i is listening to media source j and β_i,j = 0 when they are not. The utility function for any citizen i is then

The Leader can pay a cost to increase the barrier for citizens to obtain information from media source 1. For simplicity, this cost paid by the Leader, for any j, is simply x₁. When the Leader chooses to Act, this cost is the only component of their utility. When the Leader does Not Act, they gain 1 to represent the savings from inaction. However, they may also be penalised by some k ∈ (0, 1) for ignoring the demands of the citizens, the latter moderated by the number of citizens choosing Demand y. Let γ = 0 when the Leader Acts and γ = 1 otherwise. Then

This completes the description of the game. The order of play can be summarised as

1. The Leader chooses x₁ ≥ 0.

2. All citizens simultaneously choose to which media sources to pay attention.

3. Nature chooses either Act or Not Act as the correct policy.

4. For all j , Nature chooses whether that j receives the signal revealing the correct policy. Citizens connected to any j that receives the signal also receive the signal.

5. All citizens simultaneously choose either Demand or Not Demand.

6. The Leader chooses either Act or Not Act.

We must first establish the necessary amount of Demand from citizens before the Leader is sufficiently pressured to Act. This threshold is achieved when $𝔼ULeader(Act)=𝔼ULeader(Not Act)$⁠. revealing y^★, the minimum number of citizens who can force the State to Act through Demand, as the smallest non-negative integer greater than or equal to $1k$⁠. Thus, as the pressure from Demands k increases, the number of citizens required to force the State into acting y_★ decreases. For simplicity, everything henceforth assumes that $y⋆=1k$⁠. Note that the minimum number of citizens necessary to demand action is always at least 2 as we have restricted k to be strictly less than 1 (as y^★ > 1 is always true).

We quickly turn to describe an equilibrium where no citizens are paying attention. This equilibrium is always possible, as we have specified citizens’ prior belief to be that inaction—or maintaining the status quo—is always the more likely correct policy.

Proposition 1 is a typical form of equilibrium found in collective action problems. Here, so long as it takes more than one citizen to pressure the Leader into action (y^★ > 1) as we have specified, no citizen has any incentive to invest in collecting information when no other citizen intends to do so either. This allows the Leader to never need to Act without any sort of investment on its part (x₁ = 0).

We turn our focus to a specific equilibrium that demonstrates our main result: the power of displacement. All following solutions assume that $n≥1k$⁠, or that there are indeed enough citizens in the game to collectively hold the government accountable.

Citizens choose whether or not to listen to a particular source j based on how costly x_j it is to pay attention to it. For source 1—the media source vulnerable to government manipulation—this will depend on the Leader’s effort x₁. It follows that the relevant choices of x₁ for the State are either x₁ = 0, where the State spends nothing, or $x1=x1⋆$⁠, where $x1⋆$ is the value of x₁ that is just enough to discourage citizens from listening to source 1. The value of $x1⋆$ can be found when a citizen is indifferent between paying attention to 1 and not paying attention to 1.

For clarity, we describe the strategy played by citizens in our main equilibria of interest. Let j^† be an arbitrary media source.

By the above definition, Dutiful Citizens will pay attention to the best quality media source that they are not priced out of listening to. They will listen to source 1 so long as the Leader does not sufficiently obstruct them. If there is sufficient obstruction, then they will not listen to any source in the one-media-world and they will listen to source 2 in the two-media-world. Consider the following equilibrium where accountability is in the hands of Dutiful Citizens.

In this equilibrium, the State exerts effort and successfully directs citizen attention away from the best source of information 1. To illustrate the displacement effect, we will demonstrate that the effort required to achieve this $(x1=x1⋆)$ is smaller in the two-media-world than in the one-media-world.

As should be apparent, $x1⋆(2MW)<x1⋆(1MW)$⁠. This observation is the formal manifestation of how the displacement mechanism is a “boon” for leaders. That is, the effort needed for displacement $theeffortneededfordisplacemen$ is smaller than the effort needed for obstruction $(x1⋆(1MW))$⁠. The Leader will thus find it cheaper to divert attention away from source 1 when source 2 is present than when it is absent. These savings can be quantified as $x1⋆(1MW)−x1⋆(2MW)=pq2.$ Therefore, when moving from the one-media-world to the two-media-world, leaders can use these savings as a discount on their media manipulation efforts.

For completeness, we provide the equilibrium pay-offs for the Leader in both set-ups

It can be seen that the Leader’s pay-off when obstructing source 1 is the same in both set-ups. Putting the displacement effect to use thus does not deprive the state of any welfare relative to when such an effect cannot be present.

Given that the displacement mechanism introduced savings on media manipulation, some utility must therefore be lost by the Leader elsewhere to balance the gain from the lower equilibrium effort x₁. This loss can be found by comparing the equilibrium level of accountability across the two set-ups. Citizens of the 1MW get no information in equilibrium as their equilibrium behaviour is to listen to no sources. However, citizens of the 2MW do get some information as they do listen to source 2 in equilibrium. As such, there is accountability in the Dutiful Citizen Equilibrium in the 2MW —even if it is not an ideal level from a citizen perspective. This nominal accountability is a disutility for the Leader as the Leader will on expectation still occasionally be forced to Act in the 2MW but never in the 1MW . This added accountability rebalances the pay-offs.

Despite this, the comparison we would like to invoke in interpreting these results is to the common sense assumption that an increase in media sources in society weakens the position of the leader. Such common sense is also the result of existing formal models such as Edmond (2013). Our results thus show that, under the conditions specified in some of our models, these preconceptions are not always true. Equilibrium accountability has indeed increased. However, by using the savings introduced by the displacement mechanism, Leaders can offset the disutility of greater accountability and be not particularly worse off than before.

From this perspective, media proliferation may be a (weakly) win-win situation for citizens and leaders. Formally speaking, Dutiful Citizen utility in the two set-ups at equilibrium are

Citizens thus see a gain of q₂α when moving from the 1MW to the 2MW, due to the chance of occasionally holding the Leader accountable through their information from source 2. Collectively then, citizens gain accountability in equilibrium, whilst Leaders are no worse off than before. A robust media environment may therefore still have positives for society, but is simultaneously something insecure leaders need not fear.

How does displacement interact with media independence (or lack thereof)? Consider a third variant of the game we call the dominant-leader-world (DLW). In this third set-up, both x₁ and x₂ can be decided by the Leader in Step 1 of the game. We demonstrate the equivalent to proposition 2 in this set-up.

This variant of the Dutiful Citizen equilibrium is largely identical to that of the two-media-world. Thus, on the equilibrium path for proposition 3: Leaders will only manipulate source 1, and Dutiful Citizens listen only to source 2 as they are priced out of listening to source 1.

Our interest here in proposition 3 is both in what happens in equilibrium and what potential alternatives it rules out. Namely, we would like to highlight that Leaders will not be tempted into deviating to manipulate both sources.

Consider the equilibrium in proposition 3. Here, the equilibrium effort towards media manipulation is $x1⋆(DLW)=p(q1−q2)$ and $x2⋆(DLW)=0.$ Leader expected utility in equilibrium is

Note that these are the same as their equivalents in proposition 2 for the 2MW set-up.

Contrast this with a potential deviation to manipulating both sources. The critical level of effort required to obstruct both media sources is found when the Dutiful Citizens are simultaneously indifferent between listening to source 1, listening to source 2 and not listening to any sources. These are simply $xj⋆(manip. both)=pqj$ for j ∈ 1, 2. The total cost the Leader pays for effort in this potential deviation is thus p(q₁ + q₂), and their expected utility from this deviation is

It can be seen that manipulating both sources gives the Leader a lesser expected pay-off than manipulating just source 1. This is driven by two components: (1) manipulating two sources is obviously more costly than manipulating one source, and (2) manipulating source 1 in the DLW becomes particularly expensive compared to the 2MW because the Leader becomes unable to take advantage of the displacement mechanism. That is, if both sources are to be blocked, then the Leader becomes unable to use the discount that comes with moving citizen attention from the better source to the weaker source. These increases in cost are high enough that they overpower any utility gain from completely shutting out citizen accountability by manipulating both sources. To sum up

A key assumption that the model makes is that the alternative media source 2 has no significant barriers to access (apart from the potential endogenous cost in the dominant-leader-world). While this assumption is both strong and necessary for the displacement result, it also seems mostly palatable. If the case of interest is the emergence of new and independent sources of information, then it seems fair to say that many of these sources are indeed very easily accessible to consumers. The new medias themselves have an interest in being broadly read and may typically emerge online and on social platforms.

A further key assumption is that the alternative media is of lower quality— that is, less likely to signal a need for action when appropriate. This, too, seems mostly defensible. Whilst new media need to have its competitive niche, it seems fair to say that, on certain topics of political import, they would not have the resources and access that more established outlets may boast. The later case studies provide some examples of this. Instead, the selling point of these alternative media should be their accessibility, as aforementioned.

The strategy profile we attribute to citizens is also important for our results. If citizens are conditioning their attention to media sources based on the availability of other sources, then the State has the opportunity to use the displacement phenomenon to “subsidise” the effort necessary to manipulate a certain source. The collective action nature of our design also emphasises that, in reality, a group of citizens would need to behave in this way for the displacement result to emerge. We believe this to be a reasonably realistic portrayal of real-life citizens who do indeed have limited capacity to pay attention to politics and may be “shopping around” for convenient sources of information.

A comparison of propositions 2 and 3 reveals that the displacement result can emerge regardless of whether the alternative media is or is not vulnerable to potential government intervention. Such a result is not trivial. If media source 2 is considered as a new and emerging source of information, then one would surmise that it would not immediately be vulnerable to government manipulation. This would, in turn, imply that, if a larger game was to be added where it is to be determined whether the set-up is the two-media-world or the dominant-leader-world, then the two-media-world would be the status quo. The Leader might then have the opportunity to expend effort in order to change the set-up to the dominant-leader-world. A comparison of the Leader’s expected utilities across the two set-ups in propositions 2 and 3 reveals that the Leader should never expend this effort as it brings them no greater utility at the game’s end in equilibrium.

Thus, a conclusion that might be taken from our model is that governments interested in self-preservation, under some circumstances, should not expend effort to gain control over media. Specifically, such circumstances are when (1) these new media are fairly accessible to citizens but (2) simultaneously not as well positioned to report on issues the government is vulnerable to compared to media already under the government’s thumb. This is driven by the displacement mechanism where the government may find it more profitable to redirect citizen attention to a lesser source of information rather than obstructing learning entirely.

It is worth noting that the displacement mechanism may be insufficient to discourage states from wanting to control new media when long-term dynamics are accounted for. This will be true if new media are expected to improve in quality over time. Under such an assumption, a government that cares enough about its long-term prospects (that is, a high δ in a hypothetical extension) may find it profitable to invest effort to take control of new media sources today to prevent them from becoming a threat in some far-off tomorrow.

We proceed to discuss two cases to further illuminate the mechanisms highlighted by our models. To conform with our theory, we focus on three criteria to show how our model is reflected in empirical reality: 1) Efforts to control new media should lag behind efforts to control traditional media, 2) controlling traditional media should require less effort following a surge in new media usage, and 3) these efforts should promote leader welfare. Since reality is not as sanitised as our models, we are not able to decisively demonstrate these implications due to myriad other confounding political factors simultaneously at play. After all, a “perfect example might suggest that further analysis (i.e., social science) is unnecessary” (Dryzek, 2025, p. 459). Still, we hope to convince that these cases are at least suggestive of the displacement effect at play.

Malaysia’s Printing Presses and Publications Act of 1984 gave the Ministry of Home Affairs absolute discretion—without judicial oversight—over the renewal of media licenses and the power to shutdown any outlets that were deemed to offend national development or sensitivities. Such a policy is in line with the “developmental journalism” attributed to developing nations where the primary purpose of media was seen as nurturing national unity towards achieving development goals (Xu, 2009). The resultant highly selective distribution of licenses to publish meant that all major media outlets in Malaysia were then under the control of government allies, leading to predictable biases in their reporting (Abbott, 2011). This likely contributed to the continuous single-party rule of UMNO, spanning from independence in 1957 all the way to 2018.

In 1997, then-Malaysian supremo Mahathir Mohamad pledged a no-internetcensorship policy in a speech courting high-tech investors in California. This policy became enshrined in the Multimedia Bill of Guarantees the following year, marking the beginning of Malaysia’s rather permissive stance towards the Internet. Some commentators deemed this a mistake (Chin, 2003; Chin, 2011), signing away the regime’s ability to control online media unwitting of its potential future influence. The contrast is particularly stark to the severe restrictions placed upon traditional mainstream media. Wittingly or otherwise, this policy choice reflects criterion (1) above and suggests more depth behind the decision.

Mahathir’s pledge of online freedoms was quickly followed (speaking temporally, not causally) by a wave of pro-democratic sentiment known as reformasi (borrowed from the roughly contemporaneous democratisation movement in neighbouring Indonesia). At the forefront of this movement was the birth of new online medias that flourished under the offered freedoms, giving citizens greater exposure to opposition viewpoints and government criticism. The most successful and well-known of these outlets is certainly Malaysiakini, though numerous other online newspapers and blogs also thrived.

Yet, there are of course also difficulties. As these new online outlets do not hold an official media license, they are unable to attend government briefings. Funding is often an issue; even the pioneering Malaysiakini operates under significant budget constraints (Steele, 2009; Tapsell, 2013). Access and finances thus give these new medias significant quality disadvantages relative to established mainstream media.

The target and reach of these newer media focused on the politically-savvy urban elite (Tapsell, 2013), a demographic that might represent the Dutiful Citizen in our model. The first wave of new media in the late 90s is estimated to have caused a roughly 30 per cent decline in the readership of major Malaysian newspapers (Chin, 2003). Given the aforementioned significant restrictions on traditional media, we might conclude that it was indeed the actions of the then-government that led Dutiful Citizens to turn their attentions to online medias after Internet liberties were established. Whilst these liberties were about attracting Western investment and not primarily concerned with media implications, we argue that it would be naive to assume that the Mahathir government was unaware of such downstream effects. Such effects should have been particularly salient given the heavy-handed Internet regulations used by other contemporaneous Southeast Asian autocracies (Leong, 2016; Weiss, 2014). Whilst far from unlimited (Tapsell, 2013), Malaysian internet is and was starkly freer by comparison.

Following the growth of these new media, there was a simultaneous up tick in the repression of traditional media. Tapsell (2013) has interviews with many journalists from major mainstream publications such as New Straits Times and The Star who were ousted at the time due to their reformism and activism. Major publishers with still some semblance of independence, such as Nanyang Press in 2001, were coercively bought up by investment arms of government parties. We thus speculate that perhaps, with Dutiful Citizens now primarily engaged with new medias, the government felt the opportunity ripe to cull dissenters within traditional media because backlash from such repression would be dampened by the lack of attention by and tangible alternatives for the critical sections of the public. Whilst we of course cannot causally prove this, we hope its plausibility alongside the logic of the model is somewhat persuasive towards fulfilling criterion (2) above: new media make control over the old increasingly easy due to reduced scrutiny.

Whilst these new media freedoms were often celebrated (Chin, 2011; Gong, 2011; Steele, 2009; Willnat et al., 2013), the UMNO government remained significantly resilient. It was not until the 2008 general elections—ten years after the wave of Internet freedoms—that they began to show cracks; and even then, they retained power despite reduced parliamentary seat totals. Whilst then-Prime Minister Abdullah Badawi did acknowledge an underestimation of the Internet, it is difficult to decisively attribute new media as the reason for the government’s 2008 losses—especially as Abdullah himself was often criticised as a lame duck executive. It would be another ten years before the country finally delivered a general election where power changed hands—ironically to an opposition led by Mahathir himself, the original architect returning to benefit (Tapsell, 2018) from the online freedoms he enshrined after defecting from his old party.

It might be prudent to emphasise that our model focuses on governments that are modelled to be under pressure. It presumes political leadership that is on occasion not congruent with its selectorate’s preferences. Thus, the displacement effect is posited as a strategy to stave off pressure by decreasing the likelihood that the government comes under significant scrutiny. Regime change is, of course, a complex and intertwined process from which the true impact of singular causes is unlikely to be distilled. That said, given the twenty-year span that separates the enshrinement of online freedoms and a change in ruling party—a change that was reversed within two years—we believe there is room to argue for criterion (3) here: that online media freedoms did not significantly harm the then-ruling UMNO regime.

A major contributor to the eventual downfall of the UMNO government was the corruption scandal of Na jib Abdul Razak, the Prime Minister going into the 2018 general elections. Na jib is accused of embezzling roughly USD 700 million from the 1 Malaysia Development Berhad (1MDB) sovereign wealth fund that he helmed in his concurrent post as Finance Minister. This would become known as the 1MDB crisis, see Jones (2020) for details. Corruption in Malaysia is no novel thing (Jones, 2022), but has rarely previously presented a grave threat to the ruling government. This is arguably a direct consequence of the government’s stranglehold over traditional media, preventing any significant investigation and exposure of such issues.

The 1MDB crisis was a corruption scandal large enough to capture the public imagination despite the Najib government’s best efforts to keep it out of the spotlight. At the forefront of the exposure in May 2014 was the Sarawak Report, an online publication primarily known for championing indigenous rights in the sub-national state of Sarawak. Traditional media only began to report on the issue in earnest the following year. The combination of this corruption scandal interacting with new media’s ability to disseminate the issue amongst the voting public is credited as a key factor in UMNO’s 2018 downfall (Tapsell, 2018).

Recall our definition of quality for our models. Quality simply means that a source is more likely to receive the signal that the Leader should be made to Act. Although Sarawak Report was among the first to break the scandal, it is entirely possible that journalists at bigger outlets were also aware of the news—but pressured into not reporting through the above described strong controls on traditional media. Whilst this is purely speculation on our part, we would like to emphasise that, just because a smaller outlet broke a particular news before a large outlet, does not mean that the smaller outlet had more quality (in our specific definition) than the larger outlets.

Thus, whilst the government faltered in the end, it might be argued that the Internet freedoms prolonged its lifespan. As the politically conscious focused their attention on new online media, the government had an easier time in controlling traditional media and keeping its faults away from the limelight. It was not until a corruption scandal large enough to activate the new media that pressure on the government became significant enough to topple it in 2018. From this perspective, the Malaysian experience might fit the mechanisms described by our model.

In May 2013, plans to redevelop Taksim Gezi Park in istanbul into a shopping complex drew sit-in demonstrations. These demonstrations subsequently snowballed into large scale unrest throughout the country, expressing general opposition to the rule and policies of Turkish supremo Recep Tayyip Erdoğan. Mainstream Turkish media at the time took the approach of entirely blanking out reports of the unrest, with CNN Türk infamously preferring to air a documentary on penguins.

It should have been apparent, however, that news of the protests would not be confined to traditional news channels. Yeşil (2018, p. 249) points to these protests leading to a “notable increase in the use of digital outlets for purposes of sharing news and information and/or expressing critical opinions.” How did the Erdoğan government respond to this new wave of online opposition? Certainly there were initiatives taken to target new media; Burak (2021) provides comprehensive examples. Yet, online media in Türkiye is still considered relatively robust and the Turkish public increasingly favours them over traditional sources (Koger and Bozdağ, 2020). By contrast, both the lead up to and the aftermath of the Gezi Park protests were marked by significant Erdoğan efforts to control traditional media.

Erdoğan used the Tasarruf Mevduati Sigorta Fonu (TMSF), a government agency under the Prime Minister’s Office, to expropriate, restructure, and auction off unfriendly media agencies into the hands of government allies. This practice began when Erdoğan’s party took power in 2002 and continues throughout his reign. We focus on efforts in the aftermath of the Gezi Park protests to demonstrate that Erdoğan’s efforts against traditional media persisted post-2013; for more thorough treatments see Akser and Baybars (2023), Over (2021), Pukallus et al. (2020), Yeşil (2016), and Yeşil (2018).

In 2013, TMSF took control of various newspapers, television channels, and radio stations controlled by Çukurova Media Group, citing the debts of the parent company. These outlets were then sold to Ethem Sancak, a Turkish mogul and Erdoğan ally, who filled their managerial and editorial offices with pro-government voices. In December 2014, media outlets owned by Feza Publications—associated with the Fethullah Gülen movement that had recently fallen out with the Erdoğan government—were raided under accusations of various crimes and conspiracies. All outlets under Feza Publications were eventually seized by the government before being closed entirely in 2016. Many more media outlets were seized, sold, or silenced in similar ways.

Aside from forceful takeovers, Erdoğan’s regime also used less overtly coercive measures to encourage media to toe the line. Advertising revenue from government or state-affiliated firms (such as Turkish Airlines and Turk Telekom) can form a significant portion of revenue for media operations. By selectively channelling these funds conditional on their publication record, Erdoğan effectively buys the loyalty of the media to make it more difficult for citizens to receive signals from the sources. This particular anecdote of media capture is thus both reflective of our story of displacement and Trombetta and Rosignoli (2021) story of how a fractured media environment is less able to resist the government purse.

At individual levels, critical journalists and other commentators suffer personal attacks, politically-motivated dismissals, and imprisonment at great scale under Erdoğan. The government’s coercive actions also drive the media into self-censorship. Beyond the infamous penguins, what coverage of the Gezi Park protests that did exist on traditional airwaves either selectively hid anti-government sentiment or outright fabricated negative anecdotes against protesters (Over, 2021).

Thus, whilst new media in Turkiye were certainly not free of interference (Burak, 2021; Saka, 2014), the examples above appear to suggest that they pale in the face of efforts Erdoğan made towards traditional media and reflects criterion (1). Criterion (2) is more difficult to evaluate in the Turkish case. Erdoğan has made a continuous effort to undermine traditional media since his party’s installment in 2002 and this continues to the present. Yeşil (2018) argues that the media strategy of the Erdoğan government should be considered a roughly consistent policy, but does note an uptick in intensity following the Gezi Park protests. For our purposes, it is unclear whether this uptick represents a straightforward increase in effort or is instead actually reflective of how traditional media have become easier to control as new media consumption surges. Regardless, the rise of new media has, at the very least, not dampened Erdoğan’s enthusiasm for controlling traditional media—contrary to what a competing theory might expect.

Finally, with regards to criterion (3), we will note that Erdoğan remains sufficiently popular over a decade after the Gezi Park protests. Again, the fate of a regime is much too complex to be attributed to any single phenomenon. It can only be said that there may be an association between Erdoğan’s media policies and the continuity of his reign.

Using a series of simple games, we have formally demonstrated that the effort necessary to manipulate information for states decreases as the number of sources increases. This effect we dub “displacement”, where citizens can be more easily discouraged from paying attention to particular information when alternatives are available. As a consequence of the savings brought about by displacement, a proliferation in the number of media sources that might be critical of the government does not necessarily reduce the government’s welfare. We use these results to argue that leaders looking to ensure their own survival might, under some circumstances, find it rational to keep their hands off of emergent independent media.

Proposition 2 (Dutiful Citizen equilibrium I). There exists a sub-game perfect Nash equilibrium to 1MWand 2MWwhere

2athe Leader plays$x1=x1⋆,$

2bexactlyy^★citizens are Dutiful Citizens,

2c n - y^★citizens do not pay attention to any media source,

2dall citizens play Demand if they receive the signal and Not Demand otherwise, and

2ethe Leader plays Act ify ≥ y^★and plays Not Act otherwise.

Proof. 2a We can reduce the Leader’s rationalisable set of choices to either x₁ = 0, the minimum effort cost, or $x1=x1⋆,$ the smallest effort that can change citizen behaviour. It must thus be that playing $x1=x1⋆,$ is weakly better for the Leader than playing x₁ = 0. The expected utility for playing $x1=x1⋆,$ was discussed in the main text and is $𝔼ULeader(x1=x1⋆)=1−pq1$ for both 1MW and 2MW . When playing x₁ = 0, the Leader gets positive utility except when Nature draws Act as the correct policy (p) and draws media source 1 to receive the signal (q₁). Thus, $𝔼ULeader(x1=0)=1−pq1$⁠. The leader is thus indifferent between these two choices.

2b Dutiful Citizens are always at equilibrium so long as there are exactly y^★ Dutiful Citizens. As they condition paying attention to media sources on $xj⋆$⁠, Dutiful Citizens will always only be paying attention when doing so is weakly profitable for them. They will also never be interested in paying attention to sources outside of what is dictated by their strategy profile due to the collective action nature of making Demands of government always requiring more than 1 citizen to make a difference.

2c-e Straightforward.

Proposition 3 (Dutiful Citizen equilibrium II). There exists a sub-game perfect Nash equilibrium to the DLW where

3athe Leader plays$(x1,x2)=(x1⋆,0)$⁠,

3bexactlyy^★citizens are Dutiful Citizens,

3cn – y^★citizens do not pay attention to any media source,

3dall citizens play Demand if they receive the signal and Not Demand otherwise, and

3ethe Leader plays Act ify ≥ y^★and plays Not Act otherwise.

Proof. 3a We can reduce the Leader’s rationalisable set of choices to $(x1,x2)∈{(0,0),(x1⋆,0),(x1⋆,x2⋆)}$⁠. Deviation to (x1, x2) = (0, 0) plays out in the exact same way as 2a. Deviations to $(x1,x2)=(x1⋆,x2⋆)$ is discussed in the main text, where it is described to always be less profitable than the equilibrium choice $(x1,x2)=(x1⋆,0)$⁠.

3b Same as 2b.

3c-e Straightforward.

Proposition 4. Assume that citizens play the Dutiful Citizens strategy profile. Then, even when manipulating both sources is possible, the Leader will prefer to only manipulate source 1 rather than manipulate both sources.

Proof. Take the equilibrium described in proposition 3.

The cost to manipulate only source 1 is

A potential deviation the Leader may consider is to manipulate both sources. The cost for this would be

Thus, the increase in cost from this potential deviation is

Recognise that there are two components to this increase in cost. First, the cost goes up by pq₂ because of the additional need to manipulate source 2. Second, the cost goes up by an additional pq₂ because the displacement mechanism, that previously discounted the cost to manipulate source 1, is now no longer applicable when attempting to manipulate both sources.

The increase in cost associated with this potential deviation is an attempt to gain utility by reducing accountability outcomes. In the equilibrium of proposition 3, Dutiful Citizens only pay attention to source 2. Thus Leader equilibrium expected utility is

and the deviation utility to manipulating both sources for the Leader is

The deviation is profitable when $𝔼ULeader(manip. both)>𝔼ULeader(manip. 1 only)$⁠, or when 2p > 1. It is given in the design that $p∈(0,12)$⁠. This assumption is what makes it always more likely the Leader needs to Act than not—a reasonable assumption given our interest in an accountability problem where citizens are concerned the Leader may not Act. Thus, 2p > 1 is never true and thus no deviation.