Position Taking or Position Ducking? A Theory of Public Debate in Single-Party Legislatures

Abstract

Are representatives in authoritarian legislatures encouraged to take positions on salient issues? More generally, why do some autocracies allow public debate on hot topics at all? Understanding the dynamics of public legislative debate is important for the roles authoritarian legislatures are theorized to play in regime legitimation and information provision. I argue that the decision to allow public debate depends on autocratic incentives to mobilize public sentiment against the bureaucracy. While allowing debate on salient issues risks galvanizing antiregime sentiment, doing so may also mobilize public opinion against wayward government officials to improve performance and deflect blame. Therefore, I predict that autocrats will only allow public debate on issues they have delegated to the government. I test this using an automated content analysis of debate in the Vietnam National Assembly, with results showing evidence of position taking on salient issues, but only on issues the party delegates to the state.

Keywords

representation and electoral systems nondemocratic regimes legislative studies Southeast Asia

Why do some autocracies allow legislators to take public positions on certain salient issues but not on others? Furthermore, why do some autocracies allow almost no public debate at all? Anecdotally, evidence suggests variation in the degree to which legislators are encouraged to make public statements on hot button issues. In Vietnam, in 2010, after revelations that a major state-owned corporation was US$4.4 billion in debt, delegates were allowed to question the prime minister on live television. Similarly, in Iran, President Ahmadinejad endured a grilling in 2012 over poor economic growth and inflation.¹ However, in China, amid the stock market slide, the legislature was not allowed to comment on government economic reports.² Indeed, in the Chinese legislature, no public debate is allowed on government performance at all.

Variation in public responsiveness to salient issues is an important, yet understudied puzzle for our understanding of one of the most important theorized roles of authoritarian institutions—information provision. Increasingly, scholars contend that legislatures and other institutions can be useful for providing information to autocrats (Egorov, Guriev, & Sonin, 2009; Lorentzen, 2014; Manion, 2016; Truex, 2016). However, implicit in each of these theories is the notion allowing such information provision comes with a risk, which is that the information can also be used by opponents to coordinate against the regime (Truex, 2016). For this reason, a number of scholars have noted that while even the most restrictive autocrats may tolerate private grumbling, they do not allow such criticism to enter the “public transcript” (Scott, 1990; Wedeen, 2015). These concerns are likely to be especially acute when an issue is salient, as that is the time when public debate could have the greatest capacity to mobilize antiregime collective action (King, Pan, & Roberts, 2014). From this perspective, why don’t all regimes require legislators to provide information privately as is done in China, particularly when the issue is salient? This way, autocrats can gain information on public dissatisfaction without risking antiregime collective action.

More recent theoretical research offers a potential explanation. Formal models suggest that regimes may benefit from the public airing of grievances because it will force bureaucrats to improve their performance through public shaming (Chen & Xu, 2017). This relies on the notion that the dynamics of mass opinion is fundamental to the decision to allow such debate to take place in public. Although this theory is compelling and forms the basic foundation for the theory I propose, in its current form it does not explain the variation we see in the degree to which autocrats allow public debate across regimes or across issues within regimes. In contrast to the information theories, where we should expect only the private provision of information, the bureaucratic oversight model suggests that all autocrats should allow public debate on any issue where government performance is potentially poor.³

Theoretically, I build on the bureaucratic oversight model of public debate by arguing that public debate in an authoritarian legislature is used to mobilize public opinion to put pressure on wayward bureaucrats and to deflect blame from the autocrat. However, to explain the variation, I contend that the incentives to deflect blame and generate pressure will depend on the degree to which the autocrat, whom I define as the individual or group with de facto control over domestic security forces, has delegated authority to the state. In some contexts, autocrats delegate minimal authority to the state, and therefore maintain tight control over all portfolios. As such, autocrats in these cases will have little incentive to pressure the bureaucracy as they will essentially be pressuring themselves. In addition, because the autocracy and the government completely overlap, it will be more difficult to deflect blame for poor performance. By contrast, in other autocracies, where leaders delegate greater authority to the government in certain policy areas, the autocrat will be more incentivized to mobilize public opinion on salient issues pertaining to delegated portfolios. This is because on those delegated issues, the leadership can more plausibly deflect blame and because concerns about agency loss will be greater.

In terms of testing the theory, no existing empirical work examines the relationship between public issue salience and legislative debate in an authoritarian setting. Existing work on legislative behavior in authoritarian regimes focuses on who participates and what types of issues legislators raise (Malesky & Schuler, 2010; Truex, 2016). Although this research shows evidence of some critical behavior and congruence with constituent interests, we do not know whether the opinions expressed were expressed publicly and if they were expressed about issues considered timely either by the public or the media, as is the norm in democratic legislatures (Mayhew, 1974; Proksch & Slapin, 2015; Vliegenthart & Walgrave, 2011). If debate occurs in private or focuses only on issues when the public is not focused on them, this constitutes an important limitation in the degree to which legislators represent constituent interests, are able to provide information on pertinent issues, and can pressure government bureaucrats on issues that matter. Timeliness and transparency are critical features of responsiveness.⁴

With regard to testing the theory, this article meets this empirical challenge by creating an original dataset of public opinion and legislative behavior in the case of Vietnam. The single-case design is necessary for several reasons. First, while some other countries, such as China, provide a useful point of comparison, precisely because of the theory laid out in this article, there are no data to analyze in those countries because delegates are not allowed to speak publicly on the legislative floor. The second reason for limiting the research to a single case is the considerable challenge in operationalizing the three core concepts of the theory—public interest in different issues, party control over those issues, and the legislative agenda. Constructing such measures presents a challenge even in democracies, where existing research on public issue salience and legislative attention relies on massive amounts of news stories or answers to country-specific survey questions (Soroka, 2002). While these measures are difficult to obtain in democracies, they are more difficult to acquire in autocracies, where survey data may not be available and neither the heavily censored state-run media nor the foreign media may accurately reflect or drive the concerns of the public.

In this article, I conduct the analysis in the following manner. To identify salient issues, I scraped more than 50,000 stories from two Vietnamese news media outlets—one state-owned and one foreign-owned—between 2007 and 2013. To identify important issues, I run a fine-grained topic model on these articles to identify spikes of interest in different events. Then, to quantify whether these issues resonated with the public, I use Google Trends search data of terms related to issues in those stories. I then match these topics with topics discussed in the legislature.

To measure legislative attention to different topics, I rely on 1,935 speeches made in oversight hearings on the floor of the Vietnam National Assembly (VNA) between 2007 and 2013 processed from hundreds of Microsoft Word files scraped from the legislature’s website. After cleaning and processing the data, I apply an automated topic model to the speeches and match the topics discussed in the legislature to the most closely related topics gleaned from the news data. After matching the legislative topics to the topics discussed in the public, I divide the issues into two groups—state controlled and party controlled based on whether or not Vietnam Communist Party (VCP) maintains an explicit party institution versus those issues where the party does not have a dedicated institution. Matching the events with the related topics in the VNA leads to a panel dataset of salient issues for 12 VNA sessions between 2007 and 2013.

Fixed-effect ordinary least squares (OLS) and fractional logit regressions confirm that the VCP uses agenda control to allow the public to weakly influence the legislative agenda only on issues it has delegated to the government. The validation section, which addresses the potential feedback effects between debate and public salience, further confirms that the VCP structures the legislative agenda to allow oversight hearings to follow public attention on some delegated issues. With that said, the correspondence even on these issues is weaker than in democratic contexts, partly due to party control and partly because the VNA, like most authoritarian legislatures, meets more infrequently than democratic legislatures.

In an extension, I also consider who takes the opportunity to raise salient issues. Although autocrats may allow legislators to debate salient issues, delegates may avoid discussing such topics for fear that government officials, some of whom may rise up within the party, could punish them later. Using data on delegate backgrounds, regression analysis shows that consistent with previous work, the delegates most likely to take the opportunity to raise salient issues are those with more legislative experience, such as incumbents and those who sit in the legislature on a full-time basis.

The findings and theory provide the first assessment of how publicly responsive authoritarian legislatures are with regard to issue salience. In doing so, it presents an answer to the puzzle of why some authoritarian regimes require legislators to provide information only in private, whereas others allow legislative position taking. Taken together, these findings paint a complex picture for theories suggesting that authoritarian legislatures generate representation (Manion, 2014, 2017; Truex, 2016). It suggests that where public legislative oversight is allowed, debate may not actually be targeted at the autocrat but rather toward the government at the autocrat’s behest. Second, there may be limits as to how well legislatures can perform their safety valve function. While there is evidence for some responsiveness, if a legislature ducks important issues, the lack of debate may signal that the institution is irrelevant and thus diminish its ability to confer legitimacy upon the regime. Finally, these findings will also be of interest to scholars of Vietnam’s political institutions. In particular, it will help explain why Vietnam, which is also ruled by a communist regime, has steadily expanded the ability of VNA delegates to take positions on salient issues, whereas Chinese National People’s Congress (NPC) delegates are forced to duck them.

Theory and Hypotheses

The theory presented above is simple and intuitive. Where autocrats delegate more authority to the government, they should allow public debate on those issues that are delegated. However, this theory raises two questions. The first concerns delegation. Is there variation in the degree to which autocrats delegate authority in practice and what explains this variation? Second, my theory rests on the notion that autocrats can structure the legislative agenda. This section addresses both of these issues before detailing the hypotheses.

Starting with the question of delegation, does such variation in delegation occur in practice? Although cross-national data are not available on this, qualitative evidence suggests that it does. In the Soviet Union, Gill (1994) suggests that the party was involved in the formulation, implementation, and oversight of policy on all issues. In China too, under Mao, policy making was centralized within a small coterie of party leaders (Miller, 2008). However, in other contexts, there is greater separation between unelected autocratic institutions and the state across different issue areas. Vietnam responded to criticism that excessive ideological control was contributing to ruinous economic policy by cutting the number of party policy committees nearly in half in 1989 (Thayer, 1991, p. 7). In Iran as well, the Supreme Leader of the Guardian Council has the power to appoint the head of the judiciary, military, and intelligence ministries while the President is primarily in charge of economic issues.⁵

Why does delegation take place in some contexts and not in others? Egorov et al. (2009) note that authoritarian regimes face a trade-off between political control and bureaucratic autonomy. While increased autonomy for the bureaucracy may improve efficiency, it also means that the autocrats are less able to manage threats against their rule and control rents within those ministries. To explain why regimes will allow greater delegation, the most common explanation centers on the ability of the regime to access rents easily, either through natural resources or foreign aid (Egorov et al., 2009). This is because when the regime can access such rents more easily, effective economic management becomes less important. Additional possible reasons for delegation could be to encourage risk taking by government bureaucrats (Shirk, 1993).

If delegation occurs, the types of portfolios delegated are not likely to be random. Truex (2016) argues that autocrats may seek out information but only on “non-sensitive” issues that are not vital for authoritarian control. The specific ministries he cites include defense, public security, and information. Although he focuses only on China, these ministries are likely to be deemed sensitive in other contexts as well. Indeed, in Burma, as per the 2008 Constitution, the military retains control over the ministries of border affairs, defense, and home affairs (Ostwald & Schuler, 2015). In this article, I assume that the types of ministries that are likely to be delegated to the government should broadly follow issue areas Truex (2016) find to be sensitive. At the same time, the case of China also shows why it is important to consider the degree of delegation for an issue in addition to sensitivity when considering public debate. As the cases of Vietnam and China show, while the sensitivity can explain variation in delegation, it cannot explain why public debate on nonsensitive issues is not allowed in China but is in Vietnam, and even there only in the past 25 years.

I now turn to the assumption that autocratic leaders can more tightly control the legislative agenda than in democracies, which, as stated above, is a crucial assumption in my theory. Party control is a crucial component of legislative control even in democratic legislatures. However, in democracies this theory is applied mostly to roll call votes and the legislative calendar (Cox & McCubbins, 1993), and not other types of hearings, where individual legislators can raise any subject they wish (Maltzman & Sigelman, 1996; Vliegenthart & Walgrave, 2011). Looking within the largely unexplored world of micro-institutions within authoritarian legislatures (Schuler & Malesky, 2015), evidence suggests autocratic parties retain strong control over legislative institutions through parallel party institutions. In communist countries, the legislative leaders are generally co-opted into the ruling party institutions such as the politburo or the central committee (White & Nelson, 1982). The constitution also generally does not require the legislature to allow legislators the freedom to raise any item they wish, as is the case in democratic settings (Proksch & Slapin, 2015). As such, autocratic leaders have greater discretion in their ability to raise or suppress issues as they see fit.

With this discussion in mind, I can state the primary hypothesis of this article. If autocrats vary in the degree to which they delegate authority across different portfolios and they control the legislative agenda, this leads to the following predictions:

Hypothesis 1 (H1): In oversight hearings, legislators will publicly debate salient issues on portfolios delegated to an executive.

Hypothesis 2 (H2): In oversight hearings, legislators will not publicly debate salient issues on portfolios the autocrat controls.

An important auxiliary question is who will stick their neck out and raise salient issues in an authoritarian context. Evidence shows that autocrats claiming to want an active legislature are often frustrated by the reluctance of many delegates to criticize the government even when offered the chance. Joseph Nyerere, ruler of then-single-party Tanzania in the 1960s lamented that legislators evinced a “failure to make more than minimal use of their prerogative to criticize in the Assembly” (Hopkins, 1970, p. 763). How can autocrats resolve this problem? For space concerns, I cannot flesh out a full theoretical model of who will speak. However, previous work covers some of this territory, suggesting that professionalism, electoral competitiveness, and nomination procedures structure who is willing to criticize government ministers in a legislature (Malesky & Schuler, 2010). Delegates who are full-time members of a legislature, face greater competition for their seats, and are nominated by local rather than national party institutions are more likely to represent constituent interests even if it conflicts with the government. In the final empirical section, I will take up this issue to assess whether the delegate backgrounds that predict speeches in other work transfers to the willingness to publicly debate salient issues.

Autocratic Delegation and Legislative Debate in Vietnam

To test the delegation theory, this article examines the relationship between autocratic delegation from the VCP to the government and debate in oversight hearings in the VNA. Vietnam is a compelling case for several reasons. First, the role of the VNA has clearly changed over time. Prior to Vietnam’s economic and political reforms, which lasted from the beginning of the economic opening in 1986 until the constitutional revision in 1992, Vietnam’s legislature looked similar to China’s with regard to the inability of its legislators to engage in public representation. However, since that period, the legislature has accelerated in terms of activity. Figure 1 shows the number of days the VNA has met relative to China’s NPC. As it indicates, whereas the VNA has dramatically increased its level of legislative activity since 1989, leveling off at an average of 72 days per year since 2000, China’s legislature has remained relatively inactive with an average of about 10 days per year throughout the time period.

Figure 1.

Activity in the Vietnamese National Assembly versus the Chinese National People’s Congress.

Vietnam is also compelling because this increased activity corresponds with the introduction of publicly observable government oversight by the legislature. Vietnamese law, even prior to 1989, provided the opportunity for delegates to query government officials. However, the length and visibility of these sessions are entirely determined by the VNA Standing Committee, which is in charge of setting the agenda for legislative session. This meant that in practice, prior to 1989 even if the legislature conducted such oversight hearings, the public was largely unaware. A major change to these sessions was introduced in 1994, when members of the Office of the VNA, the legislature’s administrative staff, proposed that the query sessions be institutionalized and televised. The Politburo ultimately agreed to this change and since 1994, the query sessions have occurred from 6 to 8 days a year on live national television (Duc, 2012). No comparable institution exists in China, where delegates mostly listen to speeches by government officials and vote on legislation, but are not allowed the opportunity to publicly criticize government ministers.

Survey data suggest that the public is aware of these oversight hearings in Vietnam. Data from Vietnam’s Provincial Government and Public Administration Performance Index show that at least 40% of citizens watched at least some of these hearings in 2016 (United Nations Development Programme, 2017). Particularly noteworthy comments receive widespread media attention. Indeed, delegate Nguyen Minh Thuyet received widespread national and international attention in 2010 when he proposed a vote of no confidence in Vietnam’s Prime Minister Nguyen Tan Dung during an oversight hearing on the government budget.⁶

Despite the legislature’s increased activity, the party maintains strong mechanisms of agenda control, particularly within the oversight hearings. Vietnam’s legislature is extremely hierarchical. While all legislatures feature some degree of hierarchy to complete business (Cox, 2006), hierarchy is particularly rigid in the VNA (Malesky & Schuler, 2009). An 18-member standing committee wields veto power over the legislative agenda. Individual delegates may suggest legislation or request oversight hearings, but the standing committee must ultimately approve the requests. Furthermore, unlike in most democracies, the legislative rank and file cannot select their legislative leadership, which is determined by the outgoing legislative leadership with the approval of the VCP Central Committee and Politburo. As such, the parallel party institutions of the Central Committee and Politburo control those who wield veto power over the legislative agenda. Therefore, the party, through the Central Committee and the Politburo, has the ability to strongly determine which issues are addressed on the VNA floor.

Measuring the Concepts

Turning to the empirical tests, testing the theory requires measurement of three concepts—delegation, public attention, and VNA attention. This section details how I measure each concept. First, I discuss how I measure variation in delegation of authority to the state across portfolios in Vietnam. Then, I explain how I construct a time series dataset of issue salience, which is the key independent variable that will be used to predict legislative behavior. I also discuss how I create a measure of VNA attention to different issues in a way that can map on to the public attention data. Finally, after detailing the data collection process, I discuss the empirical strategy used to assess whether public attention predicts VNA attention and on which issues.

Measuring Delegation

In terms of party control of the state bureaucracy, since Vietnam’s economic opening in 1986 the party has increasingly pared back its control of the government. In Vietnam, as in China, the primary way that the party exercises control over policy is through policy committees accountable to the party. Responding to the criticism that excessive ideological control was contributing to poor policy performance, in 1989 the VCP decreased the number of party committees accountable to the VCP Central Committee (Thayer, 1991). While information is not available on which committees were cut, the VCP publicizes which committees remain. As of 2007, the party maintained committees that explicitly paralleled seven out of 20 government ministries. Table 1 shows these ministries, which include defense, foreign affairs, information, and internal affairs. At the other end of the spectrum, the public service-related ministries such as education and health have no explicit party body shadowing them. The only institutions available to make policy in those areas are the Politburo or the VCP Central Committee.⁷

Table 1.

Party Control Over Ministries.

Ministry	Party Committee (2007-2012)	Party control	Times queried^a
Defense	Central Military Comm.	Yes	0
Internal Affairs	Central Org. Comm.	Yes	2
Foreign Affairs	Foreign Affairs Comm.	Yes	0
Information	Propaganda Comm.	Yes	1
Justice	Internal Affairs Comm.	Yes	2
State Inspectorate	Central Inspection Comm.	Yes	0
Public Security	Central Military Comm.	Yes	1
Labor, Veterans, Soc. Affairs	None^a	No	3
Agriculture and Rural Dev.	None^a	No	4
Finance	None^a	No	6
Industry	None^a	No	6
Planning and Investment	None^a	No	3
Construction	None	No	0
Culture, Sports, Tourism	None	No	2
Education	None	No	5
Health	None	No	3
Natural Resources	None	No	3
Science and Technology	None	No	0
State Bank	None	No	4
Transportation	None	No	3

This refers to how many times the minister was required to take questions before the VNA. VNA = Vietnam National Assembly.

To be sure, the lack of a parallel party institution does not mean the party has no influence over these other ministries. Indeed, through the party’s role in appointments and the presence of party cells within all state bodies, the VCP can still provide information on general party policies and sack particularly problematic officials. However, the party cells generally exist within state bodies to provide training on the party line and do not offer independent, portfolio-specific policies. Furthermore, using nomenklatura power to enforce party discipline requires vigilant oversight of state activities to identify those in need of replacing. This task is made more difficult when the party lacks independent policy-specific institutions. For the purposes of this study, the goal is not to show that the VCP has no control over these ministries, but only that it has less control than over ministries where it has independent policy making bodies.

With this in mind, I use the presence of parallel party institutions as a proxy for party control. To show that the committees are a useful proxy for party control, Table 1 also shows some other indicators of party control such as whether the party controls policy debate on a given issue area and whether or not the leader of the ministry is also a member of the Politburo. On both of these measures, those ministries with parallel party bodies also have party bodies controlling policy journals in those issue areas. For the ministries of defense and public security, the ministers are actually Politburo members. This provides a further indication that the party exercises more explicit policy making power in these issue areas.

Measuring Legislative Attention

Moving to the measure of legislative attention, the goal is to create a set of topics the legislature considers and to match these topics to public interest. To create a measure of which topics the legislature considers, I first gathered all the debates in oversight hearings between 2007 and 2013. Oversight speeches cover two types of debate—debates on the government’s socioeconomic report and query sessions of government ministers. Critically, for query sessions, only ministers approved by the standing committee appear for questioning. In addition to these four to five ministers, the prime minister or deputy prime minister also appear at each session. In all, this amounted to 1,935 speeches, all of which were available on the VNA website and were carried on live television in Vietnam. Putting this number in perspective, this is more speeches than many autocracies, where such publicly observable speech is barred. However, the number is far less than in democracies such as the United States or the United Kingdom, where legislatures meet more regularly and are allowed greater access to the floor (Maltzman & Sigelman, 1996).

To generate a measure of topics that vary over time, I conduct a mixed-membership topic model, which assumes that each speech is a mixture of multiple topics. Of these models, the latent Dirichlet allocation (LDA) model (Blei, Ng, & Jordan, 2003) with Gibbs sampling is the most common (Grimmer & Stewart, 2013).⁸ The basic logic of the LDA model is that all corpora are made up of a latent set of topics, with each document the sum of all topics and each word in the document drawn from each of those topics.⁹ The model produces two values of interest. First, it assigns a value for the proportion of the individual document about each given topic. Second, it produces the words that are most associated with a given topic. The words help identify what the topics are about, while the proportions, when aggregated to the session level, provide a measure of how much a session is focused on a particular topic. The choice of a mixed-membership model was preferable to a single-topic model, such as k-means, a mixture model, because a single speech may raise several questions on different topics.

To conduct the automated analysis, several processing steps are required. First, because Vietnamese, like Chinese, is monosyllabic, automated analysis requires a “tokenizer” to attach different syllables together. Therefore, “gia đình,” which means family, must become “gia_đình.” To tokenize the words, I employed a Vietnamese company that tracks Internet interest in Vietnam. This company has proprietary software that automatically links syllables that form words together. To improve the meaning of the discovered topics, scholars typically drop words that do not contain substantive meaning, such as “if,” “and,” or “but.” For English, such lists come with most natural language processing packages such as nltk for Python. At the time of writing, no such lists were available for Vietnamese. Therefore, I created a list based on the English list, adding some words such as “được” and “bị,” which are specific to Vietnamese. Finally, I modified the speeches to ensure that the topics generated were based on subject of the speeches and not the identity of the speaker. One modification was to change all provincial names to the word “province” to avoid having the topics cluster on the delegate province rather than the topic of their speech. Through this process, I create the dependent variable, which is the proportion of session k about topic i.

In terms of validation, one of the perils of text analysis is that text models are inherently “wrong” (Grimmer & Stewart, 2013). While models such as LDA provide useful ways of categorizing the data, they do not reflect the data generating process involved in creating text. Therefore, whatever method is used, it is important to validate the interpretation of the measures. Online Appendix 1 addresses the concern that the topic model incorrectly identifies the topics. In particular, I hired a team of coders to hand code a subset of the speeches to ensure that the topics generated by the automated procedure corresponded to a human evaluation of the speeches. This analysis shows that in the cases where the VNA in fact discussed an issue salient to the public, the categories generated by the LDA model correspond to coder evaluations of the topics.

Measuring Public Attention

In measuring public interest, research on democracies use different approaches, some analyzing the correspondence between polling data on the most important issue facing the country and debate (Soroka, 2002). Some also use content analyses of media reports, which of course measure a distinct phenomenon that may not directly relate to actual public opinion (Vliegenthart & Walgrave, 2011; Walgrave, Soroka, & Nuytemans, 2008). More recently, scholars have suggested that Internet search data demonstrate convergent validity with these measures (Mellon, 2014; Ripberger, 2011). In the context of Vietnam, Internet searches are the best available approach because of the lack of polling data.

To generate a dataset of public interest in different issues and match this to legislative debate, I use the three-step process shown in Figure 2. First, I identify key events through a fine-grained LDA topic model of domestic and foreign media outlets. For the domestic media, I rely on Tuoi Tre, which is one of Vietnam’s most widely read online and print publications. It is also one of Vietnam’s more open-minded state-run outlets. I also chose Tuoi Tre because it regularly reports on political events and because its online archives are comprehensive. By scraping its online archives of stories from the domestic news section, I collected 46,603 stories for an average of 25 per day. For the foreign media, I scraped 6,980 articles from BBC Vietnam, which publishes about two stories per day on Vietnamese politics. The articles are processed the same as the VNA speeches.

Figure 2.

How the panel dataset was constructed.

Using each of these media outlets, I identify events by looking for spikes of interest in specific topics then reading the news stories associated with those bursts. To identify a spike, I look for months where the interest in a given topic was more than three standard deviations above the mean level of interest in that topic. Where the issue did not concern political issues, such as the release of a new movie, or if it explicitly dealt with the VNA, I discarded the event. For the remaining events, I enter in search terms related to the event into Google Trends, which measures the number of times Internet users search for a given term on a weekly basis.¹⁰ To ensure that the searches are valid, I only include those searches where the timing of the spike corresponds with the timing of the events identified in the media reports. Through this process, I identified 140 unique spikes related to politics during the period under consideration. These spikes in Online Appendix 2 range from Chinese activity in the South China Sea to the 2009 swine flu scare.

Combining the Data

The final step in creating a panel dataset of public interest in issues is matching the public interest measure with the topics discussed in the VNA. This requires some degree of interpretation. To facilitate the mapping, I use an LDA model with 150 topics on the VNA speeches. This fine-grained level of disaggregation facilitates matching on specific issues by breaking out subtopics. For instance, although debates on Vietnam’s economic stimulus and taxes may be grouped under macroeconomics in a low dimensional model, these topics form distinct topics when more topics are allowed. Because choices on topic numbers can generate idiosyncratic results, I repeat the analysis with 100 and 200 topics (Wilkerson & Casas, 2017).

With this more fine-grained level of aggregation, some topics in the VNA may nonetheless be broader than the events gleaned from the newspapers. For example, while Online Appendix 2 shows nine events pertaining to the environment, the VNA debate only produces two topics related to the environment. For this reason, the Google searches for these events were added together to match the categories debated in the VNA. Online Appendices 2 and 3 show the categories and how they were matched. In all, this process resulted in 38 topic pairs.

Before conducting the analysis, three final steps are necessary. First, measuring whether interest in the issue corresponds to debate in the VNA requires a time series dataset where VNA debate follows public interest rather than preceding it. Because the VNA meets once every 6 months, the measure of public interest is the average Google Search measure in the period between the previous VNA session and prior to the opening of the current VNA session. To ensure that the Internet interest is not driven by debate within the VNA, I did not include measures of public interest during the month the VNA sits. However, because it is possible that even using this lagged structure that anticipation of debate could spark public interest, in the validation section I look at the specific issues to see whether they are exogenous to the VNA calendar.

Second, because Internet interest is highly skewed, with some issues receiving tremendous attention and others less so, to account for the diminishing marginal returns of Internet interest, I use the natural log of the Google Search measure as the key independent variable. Finally, to differentiate between issues controlled by the party and those that are delegated to the government, I use the measure of party control from Table 1 to match specific issues with these ministries. Online Appendices 2 and 3 indicate the relevant ministries for each matched topic.

Empirical Test

I now turn to the empirical test. The central hypothesis of this article is that public attention should positively predict VNA attention but only within the subset of issues where the party delegates authority to the state. There should be no relationship between public interest and VNA attention within the nondelegated issues. This prediction follows from my theory that the party will mobilize public attention to salient issues to pressure the government and deflect blame from the party when the government controls the issue. However, the party will seek to dampen public attention when the public is focused on an issue the party controls directly.

Figure 3 provides some initial visual evidence of a relationship between Internet interest and VNA attention. For state-owned enterprises (SOEs) and land, issues that are delegated to the government, spikes in interest were followed by increased attention in the VNA. For SOEs, the notable spike occurred prior to the seventh session when Vinashin, a major SOE, was found to be US$4.4 billion in debt. During the seventh session, the minister of transportation and the prime minister were grilled on their responsibility for the issue. Similarly, for land, the VNA considered this issue more extensively after an incident in February 2012 involving a farmer firing on security forces attempting seize his land. Contrast these two issues with party-controlled issues related to justice and national security. As the third and fourth panels of Figure 3 show, VNA attention to justice or defense was not connected to spikes in interest in the arrest of prominent dissidents or the South China Sea. What attention is paid to these issues relates to oversight of lower level court officials and budget resources for defense.

Figure 3.

Change in VNA oversight attention/Internet interest on selected issues.

Before moving on, it is important to note that Figure 3 also shows two features of the data critical for the analysis. First, there is a great deal of unexplained variation in the VNA. Other factors that this model cannot pick up also influence the oversight speeches such as party priorities, delayed attention to past issues, or local issues. For example, land spiked as an important topic in the VNA during the sixth session despite relatively low levels of public interest in the topic. This may occur either because the government is responding to calls from the VNA to more fully consider such issues months after an incident or because it is responding to events not picked up by the procedures used here. Such unexplained variation should bias all the results toward a null finding.

The second point is that VNA attention has some time dependence. For oversight hearings, although there is no formal rule, it appears that the VNA focuses on different topics in different sessions. Therefore, if an issue is considered in a given session, it is less likely to be discussed in a subsequent session. To account for this, the models use one-session lags for VNA debate. For the tests, I use two models. Because the dependent variable is a fraction, the main model is a fractional probit regression:

\begin{array}{l} \ln (L) = \sum_{i = 1}^{N} y_{i, t} \ln Φ (β_{1} {x^{'}}_{i, t} + β_{2} {y^{'}}_{t - 1} + α_{i}) \\ (1 - y_{i, t}) \ln Φ (β_{1} {x^{'}}_{i, t} + β_{2} {y^{'}}_{t - 1} + α_{i}), \end{array}

where $y_{i, t}$ is the level of attention paid to topic i during session t and $β_{1} {x^{'}}_{i, t}$ is the level of Internet interest in topic i immediately preceding session t. The fixed effect for VNA interest in topic i is α_i. The lag of VNA interest is $β_{2} {y^{'}}_{t - 1}$ . Because there may be concerns about incidental parameters with a fixed effect in a probit model, given the low number of time periods, I also estimate an OLS regression with fixed effects. For the fractional probit and OLS models, I run them separately on delegated and nondelegated issues.

Table 2 shows the results of the fractional probit and OLS regressions on delegated and nondelegated issues. Models 1 and 2 show that public attention positively predicts legislative debate on delegated issues, while Models 3 and 4 show no effect on the nondelegated issues. The findings hold using fixed effects OLS. However, even within the delegated issues, the relationship is not particularly strong and a lot of variance is unexplained. As I discuss in the next section, this is likely due to the infrequency of the oversight sessions, which cannot address all of the topics that accrue during the intervening periods. It is also likely that a number of issues raised are of local concern, and thus not appearing in the national media. Finally, it is possible that some speeches are “planted” to serve regime goals that we cannot observe such as advancing a particular policy agenda.

Table 2.

Impact of Interest in Topics on Oversight Hearings.

Model	Delegated issues		Nondelegated issues
	Fractional probit	OLS	Fractional probit	OLS
	1	2	3	4
Internet interest	0.130***	0.00462**	−0.0566	−0.00239
Internet interest	(0.0460)	(0.00220)	(0.109)	(0.00506)
Lagged VNA attention (1 session)	−0.773	−0.0445	−4.210**	−0.156***
Lagged VNA attention (1 session)	(1.346)	(0.0855)	(1.717)	(0.0423)
Fixed effects	Yes	Yes	Yes	Yes
Constant	−2.504***	0.0117***	−1.658***	0.0166*
Constant	(0.0925)	(0.00247)	(0.413)	(0.00769)
Observations	348	348	108	108
R²/pseudo R²	.0632	.018	.0385	.024

Robust standard errors in parentheses. Models 1 and 3 are fractional probits; 2 and 4 are OLS. OLS = ordinary least squares; VNA = Vietnam National Assembly.

p < .1. **p < .05. ***p < .01.

A final point to note is the lack of autoregressive effects, which are typical in these types of studies. In the context of this study, the lack of effects is quite understandable because the sessions are separated by more than 5 months and feature different ministers. As such, it is more likely that a given session will feature less relationship than would be the case in a setting where sessions occur regularly.

One concern with these findings is that high leverage observations may drive the finding. Online Appendix 4 shows a leverage plot of the OLS regression indicating that the “inflation” topic has a high degree of leverage on the delegated issues. The results also hold when dropping inflation from the analysis. Finally, reflecting concerns about topic instability, I reran the models using 100 and 200 topics for the VNA speeches. The fractional probit results were significant in both models. The OLS results were stable, but were only significant at the 85% level. Overall, this suggests that there is a link between public interest and VNA debate, but that it is not as strong as in democratic contexts.

Validity Checks, Alternative Explanations, and Robustness

Although the preceding analysis aligns with my predictions, some concerns remain. First, does VNA attention result from intentional party agenda control? Second, are the events used to generate the Internet search data actually discussed in the periods suggested by the LDA method? Third, is Internet interest endogenous to anticipation of VNA debate? Fourth, an alternative explanation for the findings could be that debate is tolerated in certain areas simply because those issues are not sensitive rather than as part of a desire to ensure government oversight. Finally, there is a question of what to make of the weakness of relationship, even on the issues delegated to the state. This section addresses each of these issues.

Starting with the first concern, to assess whether party agenda control played a role, I examine one particular feature of the VNA’s agenda setting institutions, which is the power of the standing committee to choose which ministers are queried. If the party is strategically suppressing certain issues, it should prevent party-controlled ministers from appearing before the VNA. To look at this, I conducted a t test on the number of times a delegated ministry was queried versus a nondelegated ministry (see Table 1 for a list of the ministries). The average number of query sessions for a nondelegated ministry was less than one for the 6-year period compared with 3.2 for delegated ministries. This difference is significant at the 99% level. Confirming the statistical results, VNA officials candidly admit to using this strategy to keep certain issues off the table. Deputy chairman of the VNA Office Nguyen Sy Dung said, “All countries have two criteria for not holding public debate. . . . The first is whether not the issue is a difficult one. . . . The second is whether the issues are defense issues.”¹¹ Therefore, he continued, the ministers of foreign affairs and defense should not appear before the VNA. As to Dung’s assertion that foreign affairs are not discussed in other contexts, evidence from democracies suggests that debate does occur over foreign affairs and defense (Quinn, Monroe, Colaresi, Crespin, & Radev, 2010; Vliegenthart & Walgrave, 2011). Indeed, the United States and the United Kingdom held robust legislative debates on whether or not to enter the Iraq War in 2002 and 2003.¹² However, even if foreign affairs are somehow different, the analysis shows that other issues, like information policy, justice, and public security are also excluded.

Addressing the second and third concerns, to ensure that the issues driving the positive result for delegated issues were discussed and were not endogenous to anticipation of VNA debate, I examine issue spikes that preceded spikes in VNA attention. I then looked within those sessions to see if the issue was on the agenda. Similarly, for the nondelegated issues, I looked within the sessions following large spikes in public interest to see if those issues were in fact avoided. Because I am most concerned with the issues driving the positive result in the model, I focus on the issues where the T-score of Internet interest on legislative debate is greater than 1. Table 3 shows that for seven out of the eight issues there is clear evidence that the VNA debate reflected the issue generating public concern. For example, on the issues of the stimulus package, land compensation, and electricity shortages, the VNA leadership instructed delegates to focus on these issues because of public interest. The only issue where there was not a clear link was war history, where the data show a spurious link between interest resulting from an Agent Orange trial in the United States and coincidental VNA debate of veterans’ issues during the next session.

Table 3.

Discussion of Key Issues in VNA.

Issue	Timing of spike	Focus of discussion	Notes
Delegated issues
Stimulus	January-October 2009	Yes	On June 12, 2009, Ministry of Planning and Investment queried. VNA Chairman Nguyen Phu Trong tells delegates to “focus on . . . use of the stimulus package”
Natural disasters	October 2013	Yes	Cyclone Nari hits central VN raising the issue of when to release floodgates. Deputy VNA chair Son introduces Minister of Agriculture saying delegates should raise the issue of “flooding and releasing the floodgates.”
State-owned enterprises	July-October 2012	Yes	Vinashin bankruptcy builds interest in July 2010. In opening address to the fall 2010 VNA session, VNA Chair Nguyen Phu Trong explicitly raises Vinashin as an issue to focus on
Land compensation	February 2012	Yes	Farmer Doan Van Vuon attacks police attempting to seize land. On June 13, 2012, VNA chair Hung calls on the Minister of Natural Resources to answer questions directly related to land compensation
Taxes	January 2009	Yes	Individual income tax comes into effect along with immediate reduction in rate for businesses and individuals as part of stimulus. Minister of Planning and Investment faces two questions on implementation on June 13
War history	June 2007	No	U.S. ruling that victims of Agent Orange would not be compensated. There is no explicit discussion on this issue
Transportation	September 2007	Yes	Collapse of the Can Tho Bridge. The Minister of Transportation was called before the VNA and VNA Chair Trong mentioned the Can Tho Bridge as an issue for debate on June 17, 2007
Electricity	June 2010	Yes	Wave of power outages. On June 12, 2010, Deputy PM Nguyen Sinh Hung faces more than 10 questions on the subject
Nondelegated issues
Libya	February 2011	No	Vietnam overseas workers stranded in Libya when civil war erupts. Libya never mentioned in any VNA speech. Labor export not discussed in June 2011. Ministers of Agriculture, Education, Finance, and State Bank queried
Corruption	October 2012	Yes	PM faces internal challenge over corruption. He does face questions in query session from isolated delegates on role in corruption. Reason for negative relationship is correlation between corruption and land in February 2012
Defense	December 2007, June 2011, July 2012	No	Three Chinese actions in South China Sea. After no instance are the Ministers of Foreign Affairs or Defense called to take questions
Appointments	January 2011, January 2013	No	2011 Party Congress and December 2012 appointment of Nguyen Ba Thanh to center. In neither case are the appointments made by the party questioned by the VNA
Justice	November 2011	No	Cu Huy Ha Vu arrested. Neither his name nor the names of any other dissidents are mentioned in the VNA. When the justice ministry is questioned in 2013, it is only on the issue of lower level judge quality

VNA = Vietnam National Assembly; VN = Vietnam; PM = prime minister.

In terms of the nondelegated issues, for four of the issues with either an unclear or negative relationship, there was no discussion in the VNA. For the South China Sea, after each spike in interest, the VNA was not afforded the opportunity to question the ministers of foreign affairs or defense. Similarly, for issues related to the judiciary and police, after the arrest of prominent dissident Cu Huy Ha Vu, the VNA was not encouraged to debate the wisdom of this decision. In fact, no dissident’s name was ever mentioned in any speech on the floor of the VNA. The one issue that was discussed was corruption, where the prime minister was challenged just after the Politburo unsuccessfully attempted to oust him in 2012. The reason for the negative correlation is that the corruption topic was also relevant to the land topic, which was heavily discussed in June 2012.

The examination of these specific issues also allays some of the endogeneity concerns. Each of the events that precipitated public interest was exogenous to the VNA legislative agenda. Natural disasters and the Can Tho bridge collapse were clearly exogenous and others, such as the stimulus package, taxes, and land compensation followed government announcements or events rather than legislative investigations. Internet interest in those issues surged immediately following the events or announcements rather than after announcements that the VNA would query the relevant ministers. Even if the announcement that the ministers would take the stand caused a modest spike in interest, these announcements are made after VNA sessions begin and were thus excluded from the measure of Internet interest used in the analysis.

In addressing the alternative explanation that the observed behavior of the legislature is a function of the VCP, allowing debate on nonsensitive issues as a safety release valve rather than a desire to check the bureaucracy, no ironclad rejoinder emerges from the data. Indeed, both this theory and the bureaucratic oversight theory are consistent with the observed patterns. However, the bureaucratic oversight story is the most compelling explanation when one considers the evolution of the VNA. As I argue in the next section, the VCP has steadily increased the level to which the VNA can debate issues publicly. Most notably, in 1994, they allowed VNA query sessions to be televised. It seems unlikely that certain issues became less sensitive in 1994. Indeed, given the economic crises that precipitated the economic opening, economic issues were perhaps more sensitive than ever at that point. The best explanation for the shift is that, in the late 1980s, the VCP devolved more authority to the state in certain policy domains.

Finally, on the weakness of the findings, the results show that the effect of public interest is only consistently significant in the fractional probit models. The coefficients are consistent for the OLS models, but only significant at 95% levels for the models that use 150 topics to categorize VNA speeches. Even in the probit models, the coefficients are small. The obvious conclusion is that responsiveness, while present, is much weaker than in democratic contexts. The primary reason for this is that the VNA, like most autocratic legislatures, meets far less frequently than democratic legislatures. With fewer data points, finding significance is less likely. More substantively, infrequent, shorter sessions limit responsiveness because more issues will accrue during the longer recesses and delegates have less time to respond. Another reason is that even within state-delegated issues, autocratic leaders may periodically step in and temporarily seize control of an issue if it becomes too hot. This is perhaps what occurred in the case of the swine flu scare, which was extremely salient, delegated to the state, yet not discussed in the VNA. The perceptible, but weak effect, speaks to the overall argument of this article that responsiveness to issue salience is evident, but limited.

Who Discusses Salient Issues?

An important extension to the findings in the previous section is who within the legislature addresses the salient issues. Although the party leadership may allow debate on certain topics, this does not mean that delegates will bite at the chance. Delegates may fear that criticizing government officials, who are in many cases higher ranked than the delegates, will punish them. Furthermore, it is possible in a system such as Vietnam’s for the current government leadership to move to party leadership positions in the future. For example, it was widely assumed that the prime minister of Vietnam from 2007 to 2016 was under consideration for the position of party general secretary in the 2016 party Congress (Schuler & Ostwald, 2016). Therefore, only some delegates with the capacity and independence from the central government may be willing to discuss salient issues (Malesky & Schuler, 2010). This section, therefore, considers who speaks on the salient issues.

Because the data used in the preceding analysis are aggregated to the VNA level, the dataset is not ideal to test this proposition. Furthermore, because we are only interested in speaking on salient issues as a dependent variable, using the proportion of a speech about a given topic does not allow us to isolate whether or not that topic was salient.

To tackle this question, I focus on the eight highly influential delegated topics driving the findings from Table 3. I then look within the sessions in which those topics were salient and search for speeches containing key terms related to those topics. Therefore, for the stimulus topic (see Table 3) I search for the word “stimulus” (kích cầu) in query session debates during the June or November 2009 VNA sessions when the issue was salient. I follow the same process for the other issues in Table 3.¹³ Using these counts, for each delegate I tally the number of speeches on a salient topic to generate a variable called “hot topic speeches.” As Online Appendix 5 shows, about 12% of delegates in the 12th and 13th VNA sessions raised a salient issue in an oversight speech by this measure.

I then match these counts to biographical data collected and coded from the VNA website for all delegates in the 12th (2007-2011) and 13th (2011-2016) VNAs. Following Malesky and Schuler (2010), the key variables of interest are whether or not the candidate was a full-time delegate, whether or not he or she was from a competitive district, and whether or not he or she was nominated by provincial-level institutions or national-level institutions (see Online Appendix 5 for the descriptive statistics on the delegates).¹⁴ I also include other control variables that could correlate with these variables that may also affect the propensity to speak such as age, gender, region, incumbency, Communist Party Central Committee membership, and education.

Table 4 presents the results of a negative binomial regression of delegate background on a count of hot topic speeches and a probit regression where the dependent variable is a binary indicator of whether or not a delegate ever made such a speech. The results show that the strongest factors determining whether or not a delegate will raise a salient issue are consistent with previous work on legislative speeches. Full-time member, nonparty members, and locally nominated delegates are more likely to discuss salient issues as compared with part-time, party member, or centrally nominated delegates. Furthermore, consistent with Malesky and Schuler (2010), it is the locally nominated full-time delegates who are most willing to raise topical issues. The fact that the Communist Party has steadily increased the number of full-time delegates since 1992 from 5% to 35% in 2011 and introduced the institution of locally nominated full-time delegates in 2002 is a sign that the party wants responsiveness on salient issues, so long as it maintains agenda control.

Table 4.

Who Makes Topical Speeches?

Variables	Negative binomial	Probit	Negative binomial	Probit
Centrally nominated	−0.513**	−0.263*	0.153	0.241
Centrally nominated	−0.226	−0.138	−0.342	−0.185
Full-time delegate	0.902***	0.564***	1.205***	0.863***
Full-time delegate	−0.221	−0.127	−0.252	−0.151
Centrally Nominated × Full-Time			−1.058**	−0.903***
Centrally Nominated × Full-Time			−0.441	−0.248
Election percentage	−0.0182*	−0.00505	−0.0201*	−0.00639
Election percentage	−0.0104	−0.00651	−0.0104	−0.00642
VCP party member	−0.852**	−0.445**	−0.761**	−0.412**
VCP party member	−0.342	−0.2	−0.334	−0.199
Age at election	0.02	0.00648	0.0128	0.000896
Age at election	−0.0153	−0.00856	−0.0158	−0.00873
Male	0.252	0.113	0.222	0.0936
Male	−0.26	−0.148	−0.262	−0.151
VCP central committee member	−0.182	−0.0993	−0.234	−0.163
VCP central committee member	−0.366	−0.185	−0.369	−0.188
Incumbent	0.566***	0.278**	0.540**	0.280**
Incumbent	−0.208	−0.125	−0.21	−0.124
Province south of 17th parallel	0.00576	0.0798	−0.00306	0.0957
Province south of 17th parallel	−0.204	−0.119	−0.207	−0.12
Committee fixed effects	Yes	Yes	Yes	Yes
Constant	−0.716	−0.803	−0.321	−0.476
Constant	−1.111	−0.647	−1.113	−0.635
Observations	962	962	962	962
r²_p	.0878	.101	.0936	.118
ll	−429.9	−325.3	−427.2	−319.2

Robust standard errors in parentheses. Models 1 and 3 are negative binomial regressions where the dependent variable is a count of topical speeches; Models 2 and 4 are probit regressions where the dependent is a binary indicator of whether or not the delegate ever issued a topical speech. VCP = Vietnam Communist Party. ll = Log-Likelihood. r2_p = Pseudo R-Squared.

p < .1. **p < .05. ***p < .01.

Discussion on the Generalizability of the Findings From Vietnam

The findings in this article examine the implications of my theory of public legislative responsiveness in the single case of Vietnam. However, in introducing the theory, I suggest that some countries such as China may not allow any public debate. The analysis naturally raises the question of why Vietnam’s legislators are allowed to debate salient issues within portfolios delegated to the government on the floor of the legislature on live national television, whereas others cannot. The theory in this article is that the difference should be explained by variation in the authority the party leadership in both countries delegate to their respective bureaucracies. To illustrate that the theory could be used to explain cross-national differences, I compare Vietnam with China, which is Vietnam’s closest match in terms of institutional design, but where the legislature is not allowed to debate issues publicly. Could differences in delegation explain the differences in legislative behavior? In Vietnam, as I have shown, selective delegation occurred. My theory would suggest that in China such delegation even in the nonsensitive issue areas should be lower. Is this the case? Furthermore, does such variation explain differences in the development of the two countries’ legislative institutions?

Although a full historical analysis is not possible in this article, evidence from the evolution of the party–state relationships in the two countries suggests that the argument presented here is highly plausible. In the 1980s, the ruling parties in both Vietnam and China considered a greater separation of the party and the state as a way of improving bureaucratic efficiency. In China, as Shirk (1993) outlines, from 1980 to 1989, Deng Xiaoping and Zhao Ziyang worked to separate the party from direct management of economic reforms (pp. 62-68). The goals of the reforms were to circumvent conservatives within the Politburo and to encourage risk taking by government bureaucrats. Although it appeared that policy making “shifted from the party to government after 1980” by reducing the influence of party committees over government ministries, the party groups were revived and the Politburo “appeared to take a more active and direct role in economic policy-making” after the Tiananmen Square crisis (Shirk, 1993, pp. 67-68). Therefore, the fledgling attempts at party–state separation in the late 1980s were short-circuited, thus suppressing an important incentive to enhance the NPC’s oversight role.

In Vietnam, by contrast, which did not directly experience a crisis comparable to Tiananmen Square, conservatives within the party enjoyed no such dominance over more reformist elements. Therefore, there was no comparable reversal of the reforms of 1989 where the number of party committees paralleling was reduced by half (Thayer, 1991). Although many additional reforms such as further reducing the size of the state-run economy stalled in the next decade (Womack, 1997), the reforms that were enacted were not reversed. Strong evidence of the permanence of the limited reforms in Vietnam is shown in the fact that the key reformer of the era—Vo Van Kiet—was never purged unlike his Chinese counterpart Zhao Ziyang. Furthermore, even today, the position of prime minister is more powerful vis-à-vis in the party general secretary in Vietnam than it is in China (Schuler & Ostwald, 2016). What is even more compelling from the standpoint of this analysis is that the reforms in Vietnam to extend autonomy to the government were coupled by reforms, promoted by political conservatives, to strengthen the role of the VNA. It was conservative party secretary Do Muoi who advocated a stronger role for the VNA in overseeing economic policy, which Pike (1992) suggests was “political power the Assembly never had before” (p. 77).

Although further historical work is necessary, this cursory analysis suggests that the different trajectories of the VNA and the NPC with regard to public position taking are connected to differences in delegation between the party and the state. In short, China’s communist leadership maintains more direct control of policy making through the Politburo Standing Committee, whereas Vietnam allows greater independent authority to the office of the prime minister (Malesky, Abrami, & Zheng, 2011). Because of this, Chinese political leadership feels less incentive to empower the NPC to oversee policies it formulates, whereas the Vietnamese party leadership has a greater incentive to empower the VNA to manage delegation to the government.

Conclusion

This article addresses the question as to whether delegates in authoritarian legislatures will be encouraged to engage in position taking on salient issues. I theorize that autocrats, most of whom are relatively unconstrained compared with democratic ruling parties in their ability to structure the agenda, will only allow such debate on issue areas they do not control directly. The findings from the automated content analysis of speeches in Vietnam confirm the theoretical prediction that legislative debate only follows public interest on issues that the autocrat delegates. Additional tests show that it is the most professional representatives that are most willing to take advantage of the opportunity. Finally, more speculatively, a comparison of Vietnam and China suggests that the delegation theory can help explain cross-national patterns in the public activity of legislatures under authoritarian rule.

In addition to providing a theory and evidence for a previously unexplored question, the findings have implications for work on authoritarian legislatures. Perhaps the most direct implication of this article is for research on representation in authoritarian regimes. Scholars increasingly suggest that authoritarian legislatures represent (Truex, 2016), or are at least responsive, to citizen concerns (Manion, 2014). In searching for representation, these studies examine connections between representative behavior and citizen preferences. This article contends that an additional part of representation, and probably the most critical one related to the safety valve function of legislatures, is responding to citizen concerns in a timely manner visible to the public. The theory and findings from this article suggests that the relationship between the autocrat, the executive, and the legislative leadership affects the degree to which this vital role is fulfilled. When autocrats control policy making, implementation, and legislative agendas, such responsiveness appears sharply circumscribed. If debate on salient issues is limited in this way, authoritarian legislatures may constitute a source of frustration as much as a safety valve for public discontent.

The results also have implications for the informational and bureaucratic oversight theories of authoritarian institutions more generally. Although information theories stress the role of institutions in transmitting information from citizens to the regime, they do not examine the reciprocal relationship. That is, they tend to ignore the role legislatures play in mobilizing public opinion to generate regime support or provide public pressure on bureaucrats. As such, in its pure form, the informational arguments should predict a purely private supply of information so that the autocrat can reap the informational benefits without suffering the risks. Conversely, bureaucratic oversight theories would predict exactly the opposite, where the benefits of public debate logically would imply that authoritarian legislatures should conduct debates on salient issues in public (Chen & Xu, 2017). The theory and findings offered in this article provide an explanation that can account for the variation that we see in real world with regard to the public provision of information through the legislature.

Finally, this article may point toward an explanation for the puzzling divergence in the legislative development between two regimes that are otherwise similar—Vietnam and China. While Vietnam’s legislature has steadily gained influence since its economic liberalization in 1986, China’s legislature has not increased in influence to the same extent. This article offers a potential explanation. Greater delegation from the party to the state in Vietnam incentivized party leadership to empower the legislature raise salient issues to act as a brake on the government. In China, where the Politburo Standing Committee has more direct control over policy, there is less incentive to allow such position taking. As such, NPC delegates are consigned to ducking rather than taking positions publicly. VNA delegates, on the contrary, are encouraged to take positions on issues the party delegates while ducking them on issues the party controls.

In terms of future research, this study points to the utility of research on the dynamics of media attention and public opinion. This article uses foreign and domestic media to generate potential topics of interest but does not look at the dynamics of which issues become salient and which do not. Future work could untangle the relationship between how information is disseminated in an authoritarian context, which issues become salient, and how that salience translates into pressure for political responsiveness. The account in this article suggests that salience driven by foreign media attention to issues that the party controls will not be responded to, regardless of whether or not the issue captures the attention of the domestic public. Future work could address this dynamic more directly.

Supplemental Material

Online_Appendix – Supplemental material for Position Taking or Position Ducking? A Theory of Public Debate in Single-Party Legislatures

Supplemental material, Online_Appendix for Position Taking or Position Ducking? A Theory of Public Debate in Single-Party Legislatures by Paul Schuler in Comparative Political Studies

Footnotes

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

Supplemental Material

Supplemental material for this article is available online at the CPS website

Notes

Author Biography

Paul Schuler is an assistant professor at the University of Arizona. His research examines political institutions and political behavior in single-party regimes.

References

Blei

Jordan

(2003). Latent Dirichlet allocation. The Journal of Machine Learning Research, 3, 993-1022.

Blei

D. M.

(2012). Probabilistic topic models. Communications of the ACM, 55 (4), 77-84.

Chen

(2017). Why do authoritarian regimes allow citizens to voice opinions publicly? Journal of Politics, 79, 792-803.

China PM predicts “battle” for growth as target cut. (2016, March 5). BBC News.com. Retrieved from http://www.bbc.com/news/business-35723198

Cox

(2006). The organization of democratic legislatures. In Weingast

Wittman

(Eds.), Oxford handbook of political economy (pp. 141-161). New York, NY: Oxford University Press.

Cox

McCubbins

(1993). Legislative leviathan: Party government in the house. Cambridge, UK: Cambridge University Press.

Duc

(2012). Bên Thắng Cuộc II: Quyền Bính [The Winning Side]. New York, NY: Osin Books.

Egorov

Guriev

Sonin

(2009). Why resource-poor dictators allow freer media: A theory and evidence from panel data. American Political Science Review, 103, 645-668.

Gill

(1994). The collapse of a single-party system. Cambridge, UK: Cambridge University Press.

10.

Gladstone

Cowell

(2012, March 14). Iran’s President unfazed in parliamentary grilling. New York Times. Retrieved from http://www.nytimes.com/2012/03/15/world/middleeast/iran-ahmadinejad-questioned-before-parliament-majlis.html?_r=1

11.

Grimmer

Stewart

(2013). Text as data: Promise and pitfalls of automated content analysis for political texts. Political Analysis, 21, 1-31.

12.

Hopkins

(1970). The role of the M.P. in Tanzania. American Political Science Review, 64, 754-771.

13.

Huang

Yeh

Y.-Y.

(2019). Information from abroad: Foreign media, selective exposure and political support in China. British Journal of Political Science, 49(2), 611-636.

14.

King

Pan

Roberts

(2014). Reverse engineering censorship in China: Randomized experimentation and participant observation. Science, 345, 1-10.

15.

Lorentzen

(2014). China’s strategic censorship. American Journal of Political Science, 58, 402-414.

16.

Malesky

Abrami

Zheng

(2011). Institutions and inequality in single-party regimes: A comparative analysis of Vietnam and China. Comparative Politics, 43, 409-427.

17.

Malesky

Schuler

(2009). Paint-by-numbers democracy: The stakes, structure, and results of the 2007 Vietnamese National Assembly election. Journal of Vietnamese Studies, 4, 1-48.

18.

Malesky

Schuler

(2010). Nodding or needling: Analyzing delegate responsiveness in an authoritarian parliament. American Political Science Review, 104, 482-502.

19.

Maltzman

Sigelman

(1996). The politics of talk: Unconstrained floor time in the U.S. House of Representatives. Journal of Politics, 58, 819-830.

20.

Manion

(2014). Authoritarian parochialism: Local congressional representation in China. The China Quarterly, 218, 311-338.

21.

Manion

(2016). Information for autocrats: Representation in Chinese Local Congresses. Cambridge, UK: Cambridge University Press.

22.

Manion

(2017). “Good types” in authoritarian elections: The selectoral connection in Chinese Local Congresses. Comparative Political Studies, 50, 362-394.

23.

Mayhew

(1974). Congress: The electoral connection. New Haven, CT: Yale University Press.

24.

Mellon

(2014). Internet search data and issue salience: The properties of Google Trends as a measure of issue salience. Journal of Elections, Public Opinion, and Parties, 24, 45-72.

25.

Miller

(2008). The CCP Central Committee’s leading small groups. China Leadership Monitor, 26, 1-21.

26.

Mitchell

Hulse

(2002, October 11). Threats and responses: The vote; congresses authorizes bush to use force against Iraq, creating a broad mandate. New York Times. Retrieved from http://www.nytimes.com/2002/10/11/us/threats-responses-vote-congress-authorizes-bush-use-force-against-iraq-creating.html?pagewanted=all

27.

Những vấn đề cơ mật không chất vấn công khai [Confidential Issues not Queried in Public]. (2015, June 4). Baophapluat.vn (Legal News). Retrieved from http://baophapluat.vn/thoi-su/nhung-van-de-co-mat-khong-chat-van-cong-khai-220506.html

28.

Ostwald

Schuler

(2015). Myanmar’s landmark election: Unresolved questions. ISEAS Perspectives, 1-7.

29.

Parliament gives Blair the go-ahead for war. (2003, March 18). The Guardian. Retrieved from http://www.theguardian.com/politics/2003/mar/18/iraq.iraq6

30.

Pike

(1992). Vietnam in 1991: The turning point. Asian Survey, 32, 74-81.

31.

Proksch

S.-O.

Slapin

(2015). The politics of parliamentary debate. New York, NY: Cambridge University Press.

32.

Quinn

Monroe

Colaresi

Crespin

Radev

(2010). How to analyze political attention with minimal assumptions and costs. American Journal of Political Science, 54, 209-228.

33.

Ripberger

(2011). Capturing curiosity: Using Internet search trends to measure public attentiveness. Policy Studies Journal, 39, 239-259.

34.

Schuler

Malesky

(2015). Authoritarian legislatures. In Oxford handbook of legislatures (pp. 676-695). Oxford, UK: Oxford University Press.

35.

Schuler

Ostwald

(2016, January 16). Delayed transition: The end of consensus leadership in Vietnam? ISEAS Perspective, 2016(2), 1-9.

36.

Scott

(1990). Domination and the arts of resistance. New Haven, CT: Yale University Press.

37.

Shirk

(1993). The political logic of economic reform in China. Berkeley: University of California Press.

38.

Soroka

(2002). Agenda-setting dynamics in Canada. Vancouver, Canada: University of British Columbia Press.

39.

The structure of power in Iran. (2014). PBS.org. Retrieved from http://www.pbs.org/wgbh/pages/frontline/shows/tehran/inside/govt.html

40.

Thayer

(1991). Political developments in Vietnam: From the sixth to seventh National Party Congress. Canberra: Australian National University.

41.

Truex

(2014). Who believes the people’s daily? Bias and credibility in authoritarian media (Working paper).

42.

Truex

(2016). Making autocracy work: Representation and information in China. New York, NY: Cambridge University Press.

43.

United Nations Development Programme. (2017). PAPI 2016: The Viet Nam Provincial Governance and Public Administration Performance Index. Hanoi, Vietnam: Author.

44.

Vietnam lawmaker calls for investigation of Gov’t. (2010, November 2). Associated Press. Retrieved from advance-lexis-com.

45.

Vliegenthart

Walgrave

(2011). Content matters: The dynamics of parliamentary questioning in Belgium and Denmark. Comparative Political Studies, 44(8), 1031-1059.

46.

Walgrave

Soroka

Nuytemans

(2008). The mass media’s political agenda setting power: A longitudinal analysis of media, parliament, and government in Belgium (1993 to 2000). Comparative Political Studies, 41, 814-836.

47.

Wedeen

(2015). Ambiguities of domination: Politics, rhetoric, and symbols in contemporary Syria (pp. 1031-1059). Chicago, IL: University of Chicago Press.

48.

White

Nelson

(Eds.). (1982). Communist legislatures in comparative perspective. Albany: State University of New York Press.

49.

Wilkerson

Casas

(2017). Large-scale data computerized text analysis in political science: Opportunities and challenges. Annual Review of Political Science, 17, 529-544.

50.

Womack

(1997). Vietnam in 1996: Reform immobilism. Asian Survey, 37, 79-87.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.72 MB