Source: Ideas and Institutions Issue #32
Analysis
Freeing Up Defense Land for Urban Growth
Last month, the Indian government took a significant step toward potentially opening up more land for urban development within Indian cities. In April, the Ministry of Defence (MoD) initiated a plan to transfer the civilian sections of military cantonments in Indian cities to the respective municipal bodies. The MoD holds around 17 lakh acres of land in the country and is one of the country’s largest landholders. Of these 17 lakh acres, approximately 1.6 lakh acres of land are within sixty-two cantonments that are often located within or adjacent to cities. In some cities, this can account for a significant portion of the city’s total area. In Ahmedabad, for example, it is estimated that the cantonment area occupies 7 percent of the city’s area.
While cantonments were initially established at the peripheries of cities, city expansion has led to many of them now being located in key areas within these cities. This has been the case in cities such as Pune, Lucknow, and Delhi. As such, these cantonments often occupy prime real estate. In addition, the restrictions on development within cantonments have often hampered the city-wide deployment of infrastructure and services. Cantonment areas have extremely stringent floor space index requirements and building bylaws that impose significant constraints on growth. Additionally, cantonments prioritize lower population densities, making them anomalous within populous Indian cities.
For example, as per the 2001 census, Hyderabad’s overall population density was approximately 21,048 persons per square kilometer, but the density in the Secunderabad Cantonment Board in Hyderabad was just 5,131 persons per square kilometer. Similarly, as per the 2011 census, Delhi-NCR had an overall population density of 29,259 persons per square kilometer, whereas the Delhi Cantonment area’s population density was just 4,235 persons per square kilometer. This is a highly suboptimal use of scarce urban land in cities marked by very high population densities.
In addition, this land is also often underutilized. A 2010–11 report of the Comptroller and Auditor General (C&AG) found that “58,529 acres of acquired land were lying vacant. Out of this, 49,831 acres of land acquired between 1905 and 1990 were lying vacant since its acquisition.” With the phasing out of cantonment areas, this underutilization of prime land is likely to abate gradually.
The location of cantonments within cities has sometimes created tensions between the civilian population and the armed forces owing to the restrictions imposed in cantonment areas. Until 2018, for example, cantonment boards in many cities had closed down roads within cantonment areas that connected different parts of the city to one another. For example, the following has been observed in Hyderabad:
While the rest of the twin cities have seen construction of flyovers, Metro connectivity and other infrastructure developments, the residents of the areas under the Secunderabad Cantonment Board were caught in a time warp… since the local military authority have kept 21 key roads closed in the cantonment.
Cantonment boards have also been dismal at monetizing land resources on their own. In the 2010–11 report on defense estates, the C&AG found that “as of March 2010, 2500 acres of land valuing Rs. 11,033 crore was on lease for an annual rent of Rs. 2.13 crore.”
In addition to using land suboptimally, cantonment areas also fragment urban governance. Cantonment areas are governed by cantonment boards, not municipalities. This makes city planning arduous, as any planning involving cantonment areas requires coordination with cantonment boards that are not accountable to the city’s population. The civilian population that lives within cantonment areas also suffers from this fragmentation of governance. For example, mortgages are difficult since all land within the cantonment either belongs to the armed forces or is leased out. Recent amendments to the relevant rules, allowing people to convert from leasehold to freehold properties, are expected to make this easier.
Until recently, if a state or municipal government required any land from a cantonment board, it had to provide land of equal value to the cantonment. This made the provision of infrastructure difficult. In 2020, the union government modified these requirements to allow cantonments to receive “equal value infrastructure” development in return for land. This is one of the many moves, including new rules for cantonment land administration, that the government has initiated to make better use of high-value land resources available within cantonments located amid cities.
This decision to absorb cantonments into municipalities has been arrived at after several efforts to encourage or force better land use within cantonment areas. For example, last year, the government eased construction norms for building close to cantonments, which stipulated that there could be no buildings within 500 meters of a cantonment area.
While there are no clear indications or statements explaining why it is now that the government is moving to resolve these issues, there could be three possible reasons:
- The poor utilization of cantonment land, its misuse, and the inability of the cantonment administrations to use prime urban land properly. The government had proposed a new regulatory framework for cantonments under a draft Cantonment Act, 2020, to overhaul the regulatory framework for managing these areas, but seems to have gone in a different direction altogether now.
- The escalating demand for well-situated cantonment land from cities, municipalities, private builders, and real estate developers would have been significant. Most Indian cities have been expanding outward to cope with the demand for urban space. Cantonment lands present a great opportunity to provide land for urban development well within the core of the city. Demand for this land, combined with the poor utilization of cantonment land by the respective cantonment boards, may have led to these changes.
- The increasing fiscal constraints on the MoD, especially in light of its acquisition program and the implementation of the One Rank One Pension scheme, may have increased the incentives to take decisive measures with regard to cantonment land. It is worth noting here that the Fifteenth Finance Commission had been asked to examine the issue of capital outlay for defense modernization and had recommended, among other things, the monetization of land as one of the sources for meeting this expenditure. While handing over cantonment land to municipal authorities does not lead to monetization, it does signify a willingness to take decisive steps driven by fiscal considerations that may not have been taken otherwise.
Irrespective of the reason, it will be interesting to study the manner in which this process unfolds. Larger and more populated cantonments will be more difficult to close, and their closure will face greater resistance. In addition, it will also be interesting to note how land and real estate markets respond to these changes and how they decide to pursue goals of profit maximization by making more efficient use of newly available land.
—By Anirudh Burman
Review
Understanding the Debate on Democratic Backsliding Through Two Papers
In recent years, a narrative regarding the decline of democracy globally has become accepted among many scholars and journalists. The latest report by the V-Dem Institute states, “The level of democracy enjoyed by the average global citizen in 2022 is down to 1986 levels. More than 35 years of global advances in democracy have been wiped out in the last decade. The decline is most dramatic in the Asia-Pacific region, which is back to levels last recorded in 1978.” The V-Dem scores suggest that the present state of India’s democracy is at its worst since the first elections were conducted back in 1951–52.
In January of this year, Andrew Little and Anne Meng (“Little and Meng”) published a working paper that challenged the prevalent narrative on the backsliding of democracy globally by offering an alternative narrative based on more “objective” indicators. They identify four types of relatively more objective indicators: electoral loss for incumbents, the ability of the opposition to organize and compete, constraints on the executive, and freedom of the media. They then construct indices to depict changes over time.
Looking at electoral turnover, they find that the “rate of leader and ruling party turnover has remained fairly constant since the late 1990s.” They also find that the vote shares of incumbent leaders in executive elections and incumbent parties in legislatures have decreased in recent years. To assess the competitiveness in elections, they construct two indices. The first index reflects levels of de facto and de jure multipartyism through certain variables. The other index captures various other electoral process violations. They find that both the indices have been mostly stable since the 1990s. They also consider certain indices on executive and legislative competitiveness that assign higher numbers for elections where multiple parties compete and win and find that there is a clear upward trend for both the executive and legislative indices.
To measure constraints on the executive, they consider certain constitutional rules that limit executive power—rules regarding term limits, succession, resignation, or permanent incapacity, and dismissal. They find that in recent decades, rules for term limits have been flat, while rules for dismissal and succession have continued a slight upward trend. They note that while altering or removing term limits is the most popular strategy for evasion, the data also includes the "placeholder" strategy used by some leaders who temporarily stepped down and allowed a trusted person to serve as the leader, such as the arrangement between Putin and Medvedev between 2008 and 2012. To measure freedom of expression, they use data on the number of journalists jailed and killed. They found that, starting in the early 2000s, there has generally been an increase in journalists jailed but also a decrease in journalists killed for doing their job.
They construct a composite index using all these indicators. The index has a strong correlation with subjective measures, except that it does not show a dip in the last decade. They then explain what may be driving differences between their findings and those in other studies by presenting two theoretical models that highlight the problems that may arise with using subjective measures to detect changes in levels of democracy within countries and in the aggregate. The first model—the coder model—shows that if all coders receive “a common shock to the information they used to make decisions (such as increased media reporting on backsliding), decreases in democracy scores will reflect measurement bias, rather than real changes to democracy.” The other model—the leader model—is a strategic model that considers changes that may have caused leaders to switch to more subtle democratic violations that are not easily detected in the kinds of indicators Little and Meng have used.
Little and Meng argue that the leader model does not square with the fact that incumbent leaders and their parties are not doing better in the elections. After all, they suggest, the violations are mainly meant to retain power. They present some evidence that media reporting, scholarly attention, and popular discourse on backsliding have increased substantially since 2016, suggesting that the coder model might better explain the divergence between subjective and objective measures of democracy. This suggestion mainly rests on the observations that the increased media and scholarly attention to democratic erosion is much more dramatic than even subjective measures of democratic backsliding and that this attention has declined significantly since the end of the Trump presidency, which could correspond to a real shift or to a declining interest with Trump out of office.
On May 26, nine scholars affiliated with the V-Dem Institute (“the V-Dem authors”) published a working paper that aims to offer a defense of the V-Dem approach, while specifically addressing the Little and Meng paper.
The V-Dem authors first seek to delineate a “common ground” for conceptualizing democracy and democratic backsliding. They suggest that a relatively more “encompassing” measure is most useful for monitoring the global state of democracy because, in their view, even though democracy is about election quality, it also encompasses political freedoms and the ability of political opposition to effectively compete in both elections and public discourse. They suggest that measures with sufficiently broad conceptualizations of democracy are more useful in an age when, as they argue, autocrats have learned how to abuse formal democratic institutions. They go on to suggest that V-Dem’s democracy measures are based on such a broad conceptualization.
On conceptualizing democratic backsliding, the V-Dem authors concede that there are myriad ways to do the conceptualization, but they highlight three conceptual points that they believe are important for measurement. First, does the conceptualization of democracy and its backsliding focus on competitive elections (less backsliding) or consider checks on executive power, protections of civil liberties, a critical media, and an active civil society (more backsliding)? Second, does the conceptualization of backsliding treat it as a short-term phenomenon (less backsliding) or a longer-term process, occurring over, say, five or ten years (more backsliding)? Third, does the conceptualization of backsliding pertain to democratic or even very democratic countries (less backsliding), or can backsliding theoretically occur in any country, regardless of its level of democracy (more backsliding)?
The V-Dem authors advocate for a broader conception of democracy and a conception of backsliding that treats it as a longer-term process that could theoretically occur in any country. They suggest that backsliding predominantly occurs through “processes driven by elected incumbents who gradually concentrate power in their own hands.” This is a subtle form of backsliding not conceptually included in a minimalist definition or the operationalization of democracy. They argue that this may be partly why “objective” measures of democracy, typically centered on electoral outcomes, show less backsliding.
The second argument that the V-Dem authors put forth is that we cannot uncritically rely on “objective” measures as unbiased indicators against which expert-coded indicators may be assessed. They argue that coding seemingly objective indicators typically requires multiple judgments by human coders. They cite the example of “fact-based” measures from V-Dem, which complement their expert-coded variables. Even these variables, they suggest, require several subjective decisions (for example, how to code independents, which round of an election to consider, and how to deal with diverging sources). This, however, is a weak argument because Little and Meng have picked indicators that require relatively less human judgement. In fact, it is for this reason that V-Dem does not rely on experts for coding these indicators, instead getting research assistants to code them. Similarly, their other illustrations also seem to miss this point. For instance, the V-Dem authors cite the lack of perfect inter-rater reliability in “objective” indicators in the National Elections Across Democracy and Autocracy (NELDA) project used by Little and Meng. NELDA codes objective indicators related to elections and reports an average of 83 percent of cases with agreement between two raters. They also argue that time-varying biases can also afflict objective measures, but they do not provide any examples of time-varying bias affecting the objective measures chosen by Little and Meng. Their examples relate to either biases that do not seem time-varying or measures that are not included by Little and Meng.
While some of the criticisms of the measures selected by Little and Meng are valid, they do not establish that, when it comes to reliability of measurement, there is no significant difference between the kinds of measures Little and Meng have picked and those that V-Dem gets experts to code. The V-Dem authors only leave matters at the suggestion that “subjective” expert evaluations, such as those of the V-Dem Project, “can be more reproducible than their ‘objective’ counterparts.” This is far from obvious, especially because it is not clear whether V-Dem’s expert coding indeed operationalizes the concept of democracy in a reliable way over time. That is what is being disputed in this debate.
The V-Dem authors then offer a defense of the V-Dem methodology, asserting that the following factors limit its potential for systematic expert bias: asking experts to code specific subcomponents of democracy and then aggregating them up, rigorous survey design, a systematic expert recruitment process, and the use of statistical methods to mitigate the impact of biased and unreliable experts. These arguments are questionable in the context of time-varying biases of the type being suggested by Little and Meng. It is not clear why such biases cannot affect the coding of subcomponents of V-Dem indices, especially when many of the subcomponents require subjective judgements. Similarly, while quality of survey design may help mitigate the chances of poor judgement, even implementing a well-designed survey may require relying on media reports, which may be feeding time-varying biases. Further, just because the experts are “scholars with advanced degrees… with deep knowledge of both the countries and sectors,” it does not mean that they cannot suffer from biases. Arguably, people most familiar with a context are more vulnerable to such biases and to groupthink. Relatedly, the V-Dem method for aggregating expert data is not a check against time-varying biases that affect a majority of experts. V-Dem uses a model to aggregate expert-coded data, which, inter alia, assumes that experts who diverge from other experts in terms of directionality are less reliable. It adjusts the contribution of such experts downward. This could cut both ways because it works to enforce groupthink when a majority of experts share a bias.
The V-Dem authors then argue that while the coder theory implies a fairly uniform erosion of democracy across countries and across components and indicators of democracy, the V-Dem data show that only a minority of countries have registered significant erosion in the electoral democracy index (EDI). This is a strawman argument, especially if we consider that coder bias may be more local rather than global, so that time-varying biases afflict experts coding for certain countries only. If this is the case, it does not matter whether no decline in democracy is seen in most countries. Such a theory is more plausible because V-Dem selects experts who work on specific countries, and such experts typically come from a community of scholars and may also be accessing similar sources of information on those countries.
Another argument they put forth is that since the V-Dem country experts can update or change their previous scores when coding annual updates for V-Dem, if expert bias is more prevalent in recent years, we should observe that many experts update their previous coding and do so in a “more pessimistic” direction. They find that very few experts change scores for any variable, and the ones who do change do not systematically alter their scores in a more pessimistic direction. The premise of the argument is not obvious. If experts’ bias is time-varying in such a manner that they are, for some countries, giving lower scores in recent years than is justified by the “reality” of democracy, why would they want to revise their scores for earlier years downward?
The V-Dem authors then critique the Little and Meng paper on three specific counts. First, they criticize the choice of indicators because they imply a minimalist or unsuitable conception of democracy and also because some of the indicators measure the phenomena in a manner that does not allow for nuanced judgement. For instance, some of the indicators, such as those on executive constraints, are dichotomous, therefore not allowing for proper differentiation. They also criticize some of the ordered categorical variables that Little and Meng use to measure democracy for the manner in which the categories within each variable scale onto the concept of democracy. All through this section, they use V-Dem’s Electoral Democracy Index as the standard measure of democracy to critique the measures put forth by Little and Meng. Assuming the former to have greater validity and using it to critique the latter is tricky, especially when, for one, the latter is meant to be a critique of the former. Many of these arguments can be turned around to critique the EDI itself.
Second, the V-Dem authors critique the way Little and Meng aggregate the indicators into a composite index toward the end of their paper. While one of the authors has clarified that the index in Little and Meng paper is not meant to be a usable objective "index," the index, along with the missing data for many country-years, could lead to absurd conclusions. For instance, as the V-Dem authors show, for a few years, China seemed to be scoring perfectly as a democracy because data for it was available only for a few (dichotomous) indicators on which it happened to score well. This assessment lacks face validity. The critique of the country scores on the composite index seems fair.
Third, the V-Dem authors present a critique of how Little and Meng deal with missing data. Little and Meng treat missing data “as-if” they are random, which is not a reasonable assumption to make, especially when, as the V-Dem authors argue, the data are missing for countries that seem to be otherwise scoring low on democracy indices. Anne Meng has noted on her Twitter account that the missing data are mostly from cases of no election (for the electoral competition variables) or no constitution (for the executive constraint variables). Since these missing data mostly pertain to country-years from the pre-1990s, it should not matter much for democratic backsliding measures in recent years. However, it would be good for Little and Meng to address this issue. Perhaps missing data for any country-year should be dealt with carefully and assumed to be random.
In their paper, the V-Dem authors frequently use face validity tests to critique the Little and Meng paper. However, such tests can also be applied to V-Dem’s score. For instance, in 2022, the V-Dem scores for India on two of the five high-level indices—the deliberative democracy index and the egalitarian democracy index—are lower than they were in 1975; for one index—the electoral democracy index—it is at the same level as it was in 1975; and for the other two indices—the liberal democracy index and the participatory democracy index—it is only slightly better than it was in 1975. The year 1975 was when the Emergency was declared in India, many opposition leaders were jailed, and civil liberties were substantially curbed.
While the V-Dem paper offers many arguments that Little and Meng should address in the next iteration of their paper, the debate is likely to continue. A key issue highlighted by Little and Meng is that in the period in which V-Dem and others claim democratic backsliding has been occurring, the incumbent leaders and parties have been losing more elections and their vote shares have also been declining. As Little and Meng argue, many “undemocratic actions,” including subtle ones, would be expected to affect electoral outcomes. For anyone to argue convincingly that democracy is in a decline globally, such findings would have to be either refuted or qualified by explaining why the myriad forms of alleged democratic decline are not showing up in increased electoral power for the incumbents.
—By Suyash Rai