Metropolitan growth, mass transit, planning regimes, and institutional structures

I blogged here highlighting the nature of metropolitan growth. 

Historically, metropolitan growth is characterised by separation of the geographies of residences and workplaces, facilitated by mass transit systems. Accordingly, all great cities are workplaces in the day. This is made possible by mass transit infrastructure that shrinks the commutes from the hinterland and expands the potential workforce for the metropolitan centre. This concentrates labour and talent and creates economies of scale, which creates the conditions for economic specialisation and innovation. 

London is a great example of this phenomenon and it has been documented. Stephan Heblich, Stephen J Redding, and Daniel M Sturm have an excellent paper that demonstrates this trajectory of metropolitan growth using spatially disaggregated data for London from 1801-1921, spurred by the invention of the steam engine. 

In counterfactuals, we find that removing the entire railway network reduces the population and the value of land and buildings in London by up to 51.5 and 53.3 percent respectively, and decreases net commuting into the historical center of London by more than 300,000 workers.

The figures below show the residential population trends for the City of London and Greater London, expressed as an index relative to its 1801 value. After being constant in the first half of the 19th century, the population of the City declined sharply by around 90% to 13,709 over 1851 to 1921, whereas that of the GL area rose from 2.69 million to 7.39 million. This trend is not confined to the City, but most central parts of GL. 

Following the arrival of the railway, we find a reduction in population growth relative to trend in parishes close to the center of Greater London, and an increase in population growth relative to trend in parishes further from the center of Greater London.

The graphic below shows the population trends in the City over day and night, with absolute numbers. 

While the residential (night) population of the central city declined sharply, workplace (day) employment increased, reflecting a shift toward a commuting-based urban structure. Lower commuting costs allowed locations to specialize: High-productivity areas became centers of employment, while areas with better living conditions became residential. Transport therefore did not simply support urban growth. It changed the internal structure of cities. By decoupling production from residence, it allowed cities to expand spatially while maintaining high levels of economic concentration. This increased the scale at which agglomeration forces could operate, since firms could cluster in highly productive locations without being constrained by the need for workers to live nearby.

What triggered this trend was the expansion of the railway network after 1836. It facilitated the dispersion of population to the suburbs and a rapid increase in population for the metropolitan area as a whole.

Mass transit shrunk commute times between the City and other London boroughs. The graphic captures the reduction in travel time in terms of an index that is a population weighted sum of the bilateral travel time reductions (between the City and the boroughs). 

Following the construction of the first railway in 1836, we observe a sharp reduction in the City of London’s population-weighted average travel times to other boroughs, which is followed shortly thereafter by the precipitous decline in its population discussed above. Although the decline in population-weighted average travel times slightly leads the decline in the City’s of London’s population, both follow a similar trajectory, with their rates of decline slowing towards the end of the 19th century and beginning of the 20th century.

This allowed City of London to specialise as a workplace, and attract commuters from the hinterland to work there.

Consistent with the City of London increasingly specializing as a workplace, we find that the steep decline in its night population from 1851 onwards coincides with a sharp rise in its day population. This evidence of an increasing specialization of locations as either workplaces or residences is also consistent with a sharp rise in public transport journeys per head of population from around 7 in 1834 to just under 400 in 1921.

This specialisation had its impact on the value of land and buildings, which manifests in the form of the City’s share of GL’s rateable value of developed land. 

In the early-19th century, this share declines from around 14 to 9 percent, which is consistent with a geographical expansion in the built-up area of Greater London. As this expansion occurs, and undeveloped land becomes developed, the share of already-developed land in overall land values tends to fall, because the denominator of this share increases. In contrast to this pattern, in the years after 1851 when the City of London experiences the largest declines in its residential population, its rateable value share increases from 9 to 11 percent. Finally, in the decades at the end of the 19th century, the pattern of a decline in this rateable value share again reasserts itself, consistent with the continuing geographical expansion in the built-up area of Greater London. Therefore, the change in the City of London’s pattern of specialization in the decades immediately after 1851 increases the relative value of this location.

The online appendix to the paper by Heblich et al documents strikingly similar trend of urban development in Paris, Berlin, Boston, Chicago, New York, and Philadelphia over the same period (1800-1921).

I want to qualify this story given the evidence we have from developing country cities. There are perhaps three big differences in the metropolitan development trajectories of the Western and developing world cities. The first is that in the latter the mass transit systems generally followed the development of urban settlements. This is unlike the former where, as we have seen, mass transit shaped the form of urban development. The second is that the residents of the latter cities now have access to the option of car commutes, a choice unavailable for the suburban residents in the Western cities in the 19th and early 20th centuries. 

The third is that the Western cities built their mass transit when their regulatory regimes (zoning, building codes, and land-use planning) were themselves being invented so that the two co-evolved. In contrast, developing world cities are encumbered with fossilised regulatory regimes whose change requires overcoming political-economic lock-in that Western cities did not face during their development phase.

The three factors together make the retrofitting problem harder - transit comes after the form, cars offer an alternative, and the regulatory apparatus is sticky. 

It is probable that such cities could remain entrapped in a model of metropolitan growth characterised by suburban sprawl revolving around car-commutes, unaffordable housing, and traffic congestion. Further, the last two have a self-reinforcing nature - increasing unaffordability expands the sprawl, increases vehicle density on the road (both in numbers and time spent on the road). This equilibrium erodes the productivity and competitiveness of the urban agglomeration and sets the stage for its gradual decline. This is apparent in the lived realities of metropolitan cities across the developing world. 

The best research evidence so far in this regard comes from the study by Liming Chen, Rana Hasan, Yi Jiang, and Andrii Parkhomenko (and here) on the impact of the opening of the Bangalore metro in 2011. They show that in a regime of low building heights, while the metro increases citywide output and welfare, those gains are several times larger with transit-oriented development (TOD). Using a quantitative spatial model calibrated to Bengaluru, they show the same metro can produce a 32 percentage point spread depending on the planning regime. The bigger takeaway is that mass transit with a bad planning regime produces a worser outcome than without mass transit. 

In their comprehensive survey of mass transit, World Bank economists Hiroaki Suzuki, Robert Cervero, and Kanako Iuchishow that transit systems whose returns were disappointing almost always lacked integrated land-use regimes, while those that performed best (Hong Kong, Singapore, Curitiba, Tokyo) treated transit and land use as a single planning object.

The takeaway is stark and simple - regulatory norms are what converts transit accessibility into measurable urban-form outcomes, and without it, the accessibility exists but the form does not change.

So, in these developing country cities where the urban form is already laid out, and where the convenience of car commutes can often trump, what can be done? 

The only option is to retrofit pre-existing neighbourhoods, and make mass transit commutes attractive both in terms of convenience and cost. In these contexts, it is essential for mass transit facilities to be complemented with urban planning norms that encourages the concentration of incremental development around the stations and also incentivises the redevelopment of the station surroundings. This is the case for TOD and is critical to discourage the use of cars and choose mass transit for work commutes. 

So, what are the successful examples of retrofitting?

In 1979, Curitiba introduced Bus Rapid Transit (BRT) with zoning relaxation only along the BRT corridors (“structural axes”), with density tapering sharply away from it. The result is that after 50 years, Curitiba has the highest transit mode share of any Latin American city, and the structural axes are visibly the densest and most economically dynamic parts of the city. 

Another example of retrofitting is that of Arlington County in the Washington metropolitan area which after the Orange Line was built in 1979 upzoned the five stations along the Rosslyn-Ballston corridor to high-density commercial and mixed-use, while preserving single-family character immediately beyond the station catchments. Today the corridor hosts over 50,000 jobs and over 40,000 residents within half a mile of the stations, generates roughly half of the County’s property tax revenue from 11% of its land, and ridership on the Orange Line is three times what was forecast.

Another example is that of Copenhagen which regulated and channelled post-war suburban growth along the S-train (commuter rail) corridors, keeping the spaces between the radiating lines (or fingers) as green wedges. The regulatory decision was made as the suburbs began forming and required active enforcement against sprawl. Today 70% of Copenhagen’s population lives within 1 km of an S-train or Metro station.

Other examples include Hong Kong’s “Rail + Property” model (where the MTR Corporation receives development rights at every new station and co-develops the station area with private partners), Seoul’s subway Line 2 which opened in 1980 (built through what was then farmland south of the Han river, and the Gangnam district was then zoned for high-density mixed-use development around every station), Shanghai’s Pudong area (the metro was built in parallel with the development of essentially a greenfield area east of the Huangpu, and the zoning placed Lujiazui’s financial district at a specific metro interchange), and London’s Docklands and the Jubilee Line extension from 1990s (which created dense developments at Canary Wharf, North Greenwich, and Stratford). 

So, what for India?

I had blogged here pointing to three requirements - relaxing development control regulations around stations, right kind of value capture instruments and their sequencing, and an incentive compatible institutional architecture. 

The third one is the most important, and can address the other two. It is about aligning the incentives between those constructing the transit system (and repaying the debt mobilised) and those responsible for the land use and other decisions that determine the commercial viability of the system. This is perhaps the hardest given India’s political economy and its less than satisfactory track record with such kind of coordination failure bridging reforms across sectors. 

Incidentally, every one of the aforementioned examples of successful retrofitting of mass transit shares two institutional features. The first is a unified authority with jurisdiction over both transit and land use in the station catchment. The second is the presence of value capture mechanisms that let the transit entity benefit directly from land-value uplift. In the Indian metropolitan areas, the simplest solution would be make the urban development authorities own the metro or regional rail entities. Needless to say, it may also be the politically hardest. 

In conclusion, the empirical evidence suggests that mass transit without the complementary planning regime and institutional structure that facilitates TOD actively produces worse outcomes than no transit at all, because it legitimises continued sprawl by appearing to offer a mobility solution while in fact raising car dependence.

A framework for deregulation in the Indian bureaucracy

I have blogged on the issue of deregulation as an organisational culture earlier. This post is an attempt to provide a framework for thinking about the bias within the Indian bureaucracy towards the prevention of inclusion errors and process abuse. 

Two features of India’s bureaucracy are of relevance. 

One, a welfare policy to support the needy and a promotional policy to incentivise something, comes filled with access barriers. Two, a regulatory policy to permit or regulate something comes tightly prescribed. 

In the former, there is a reflex within the bureaucracy to prioritise waste/leak/abuse avoidance over inclusion, whereas in the latter, there is a tendency to over-prescribe so as to limit abuse even if it inconveniences genuine applicants. This is the rational choice arising from the institutional incentives. 

In both cases, the precautionary effort detracts from the realisation of the primary objective. The pursuit of the ineligible (the false positive or the Type I error) comes at the exclusion or harassment of the eligible (the false negative or the Type II error). 

All bureaucracies are prone to this precautionary bias. The clearest exemplar of this bureaucratic bias is cited in the oft-cited paper on regulatory economics concerning the invisible graveyard in the US FDA’s drug approvals process. The works of Sam Peltzman, Henry Miller, and Alex Tabarrok have shown that a drug regulator treats approving a harmful drug (visible deaths, Congressional hearings) as catastrophic and not approving a beneficial drug (invisible deaths among those who never got it) as costless, so it errs massively toward over-caution. Studies have estimated the ratio of Type II (delay) cost to Type I (bad-approval) cost in the thousands to one.

However, there are compelling reasons to argue that this bias is excessive in the Indian bureaucracy. 

Root cause reforms aimed at deregulation should target this bias and not be confined tojust one-off changes to existing regulations. I blogged here and here on this issue. 

There is an important contrast on this with the judiciary. There, the norm of the Blackstone ratio deliberately and explicitly instills that it is better to let ten guilty escape than allow one to suffer. The bureaucracy needs its equivalent norm. 

As a conceptual framework to understand it, there are perhaps three drivers within the Indian bureaucracy of this bias.

1. The false positive is salient when detected, whereas the false negative is always a silent counterfactual. Even the oversight mechanisms focus on detecting and pursuing the false positives, while invariably staying silent on the false negatives. In public debates, too, media stories play up the false positives while largely overlooking the false negatives. 

2. The bureaucrat gets personally hauled up in the case of a false positive that comes to light. Even if honest mistakes are safeguarded now, and they are finally exonerated, the public scrutiny and process harassment is a worse punishment.

3. The bureaucratic approvals process, especially the finance department concurrence (and also legal and personnel department concurrences), focuses far more on the false negative and mitigating them than on the efficiency and effectiveness of the achievement of its objectives. 

These three drivers must be seen against the backdrop of a system with a very high level of trust deficit between the government and the beneficiaries (in the case of welfare and promotional policies) and the regulated (in the case of regulatory policies). In the Douglass-North sense, there is a low-trust institutional equilibrium inherited from the colonial control-state, which has hardened into a shared prior (“the applicant is a cheat”) and a norm (“process compliance is safety”). 

There is also the historical inheritance of the colonial guardian bureaucracy built to maintain order and extract revenue, standing above and apart from the governed, and which elevates procedural purity over substantive correctness. At a social and political level, there is a stigmatisation of private benefit and a valorisation of public gain, which often views the two in binary zero-sum game terms. There is a widespread chalta hai (it’s alright) attitude that condones such delays, hassles, and exclusions, and does not provide credible enough channels to provide feedback and redress them. Finally, I am not sure whether ease of doing business and ease of living are yet the kind of electoral goods that generally force real, actual changes. 

In contrast, the development trajectories of today’s advanced countries and China involved deliberate trade-offs that erred on the side of inclusion and lowering transaction costs. The regulatory excess in the US (captured in the so-called “everything bagel liberalism”) and the precautionary principles in the EU are more modern trends that emerged well after these countries reached their developed country status. 

In contrast, the development trajectories of today’s advanced countries and China involve deliberate trade-offs that straddle a fine line between enabling and abuse. The regulatory excess in the US (captured in the discourse of everything bagel liberalism) and the precautionary principles in the EU are more modern trends that emerged well after these countries reached their developed country status. 

As an analytical framework, the public loss function is to minimise social welfare loss by minimising inclusion and exclusion errors and compliance costs. The social optimum is when the marginal reduction in leakages and abuse is equal to the marginal increase in inclusion or convenience. Since abuse is generally rare and genuine cases are the overwhelming majority, the social optimum requires tolerating some leakage. 

However, this public loss function comes up against the official’s private loss function. The aforesaid reasons amplify an accountability multiplier that forces his/her private loss function to minimise inclusions and abuse, even at the cost of exclusions and compliance costs. 

The asymmetricity or gap between the two loss functions is the distrust wedge, the gap between the procedural strictness that the society would choose and the strictness that the incentive structure would produce. 

Any remedy to address the root cause of the problem should revolve around shrinking the wedge, either by lowering the accountability multiplier (stop over-punishing honest false positives), making exclusion and compliance costs visible and costly, or lowering the compliance costs. 

It is about internalising the implied shadow price of avoiding abuse or leakage. How many units of genuine exclusion or compliance cost is the system willing to incur to prevent one unit of abuse or leakage? This is the marginal trade-off that serious deregulation policy must target. There must be an institutional ownership (and bearing of the associated costs) of the cost of exclusion and harassment. 

So, in terms of possible solutions, here are a few suggestions.

Efforts to change deeply entrenched norms and behaviours are complex endeavours. They are collective social and political choices that get adopted by the bureaucracy as strong and salient organisational narratives. They are best and most effectively done when they are done directly to change beliefs, norms, and behaviours. 

The Indian state has generally struggled when it comes to instilling values to change collective behaviours. 

It does better with specific interventions. But there is a risk in reducing such efforts to technical measures. The Ease of Doing Business ranking is a salutary example of the form taking over the substance and subverting the objective. 

Nevertheless, the three-point conceptual framework articulated earlier can be the baseline for reforms. 

The single highest-value and genuine root-cause fix to the problem is to make exclusion visible and costly. It is useful to have an institution, ideally within the Department (even with all the flaws or conflicts of interest), or one like the NITI Aayog, entrusted with the task of capturing/quantifying and updating the exclusion and compliance costs (the false negatives). This would include the publication of exclusion/rejection/non-uptake rates and median clearance times for every benefit and licence, by geography and office, as standard MIS. To also provide perspective, these metrics can be posted against some desirable benchmarks. This may be done for a prioritised set of activities for certain departments. 

This may be an essential requirement to foster a culture that evaluates public services also in terms of their false negatives and compliance costs. 

The first step is to quantify and make the costs salient and institutionally aware. An apprenticeship, a scholarship, or a credit-linked subsidy scheme should be evaluated primarily on its uptake. A welfare benefit should also be evaluated on the cost of access and the extent of its exclusions. The compliance cost of sectoral regulatory regimes (environment, development control regulations, building byelaws, land, labour, cyber security, etc.) should be quantified, as also those for various kinds of registrations and licensing (shops, schools, restaurants, clinics, visas, land, etc.). 

This could be supplemented with an explicit articulation, for internal circulation, of an exclusion or compliance impact statement with every policy or regulation. This would provide policymakers and regulators the institutional freedom to account for convenience, access, and implementation flexibility, and avoid being exclusively hooked on loss prevention. 

Auditors and oversight entities must explicitly incorporate this dimension in their audits, assessments, and investigations that tend to focus only on the false positives. This would offer the combination of deliberative process protection, the vigilance-intake filter that separates the honest error from corruption, and the avoidance of the ex-ante (not hindsight) audit standard. 

Finally, the approval process should incorporate these principles across the gates. The financial, legal, and personnel concurrence/vetting should be circumscribed to ensure that the precautions and safeguards that are proposed don’t detract from the objectives. 

There is a high risk of these efforts spawning more bureaucracy. So it must be done thoughtfully and not in a checklist-filling manner, as was done in the case of EoDB rankings.

Further, given that most policies and regulations are administered digitally, a low-hanging fruit would be extensive adoption of data analytics. The digital trails of administrative and transactional data could be analysed to detect abuses and address them. The emergence of AI-enabled applications equips us with the tools to do this easily and effectively. The challenge in each activity should be to define the likely problems and analyse to identify signatures of abuse. 

To conclude, the stock of policies and rules with a precautionary bias is an output whose generator is the distrust wedge, or the asymmetric loss functions. The revision of rules one at a time merely empties the reservoir that the generator keeps replenishing. Therefore, the durable reform is to do a root cause analysis and recalibrate the generator with some of the aforesaid measures. This can shift the precautionary bias that is the current individual rational choice within the Indian bureaucracy. 

Examining the structures and incentives of the Chinese administrative state

Zhuoran Li has a neat description of China’s frontline bureaucracy at the township level that resonates with many other accounts. It is worth extracting in detail and discussing its lessons for countries like India. 

The township, which includes both the rural and urban areas, is responsible for the implementation of most functions of the government such as education, healthcare, public welfare, and municipal services. This is a description of the administrative hierarchy at the township level.

At the top of the township government hierarchy are local leaders, who are township cadres with administrative rankings. A township typically has three section-level (科级) leaders: the township Chinese Communist Party (CCP) secretary, the township mayor, and the chief of the Township People’s Congress, who is often the former township party secretary transitioning to retirement. In addition, there are approximately 10 vice section level (副科级) leaders, including vice party secretaries, vice mayors, the discipline inspection secretary (纪检书记), and the heads of several key offices. Officials heading less critical offices or serving as vice chiefs of any office do not hold administrative rankings and, therefore, are not considered leaders. 

The next level below government leaders are civil servants with no rank who hold bianzhi (编制), which refers to established posts within government offices. Bianzhi status grants significant benefits, including superior welfare compared to regular citizens and lifelong employment (unless the employee commits a crime). Within the bianzhi workforce, “cadres” and “officials” hold civil servant bianzhi, while other bianzhi workers have public institution bianzhi – the same designation given to teachers and doctors in public hospitals. Only those with civil servant bianzhi can reasonably expect opportunities for promotion. In contrast, those with public institution bianzhi have no career advancement prospects.

Li points to several career progression challenges at the township level. 

The path to promotion begins only when an official is elevated to the vice section level, becoming a township leader. Making it to vice section level is not difficult. Officials attest that almost all cadres can expect this promotion as long as they don’t make any serious mistakes. However, further promotion from here becomes extremely difficult. Therefore, most township leaders do not view promotion as a strong incentive and often describe themselves as “not caring much about promotion.”

Promotion opportunities at the township level are rare, as leaders face intense competition and strict age cutoffs. As Chien-wen Kou and Wen-Hsuan Tsai noted, the age cutoff for promotion to vice section and section levels is 40, while the cutoff for vice division (副处) level is 45. In reality, interviews indicate that in many places, the cutoff age for the vice section level can be as early as 35. Promotion opportunities are further constrained by the competition among vice section leaders, 10 of whom may vie for just two section-level positions. Moreover, township party secretaries and mayors are often not promoted from the pool of vice section leaders within the township. Instead, county-level governments may appoint section-level leaders through temporary transfers (挂职), effectively fast-tracking their promotion… As a result, township governments are often referred to as a “black hole for young cadres,” trapping many in perpetual vice section level assignments. They frequently surpass the promotion cutoff age and remain indefinitely in vice section positions. 

Promotion prospects do not improve for section-level leaders, as they face even fiercer competition. A county typically has only around 10 vice division (副处) level positions. Township and street (街道) party secretaries and mayors compete for these roles alongside section chiefs from the county government. As several cadres have concluded, promotion opportunities are largely reserved for those with the right connections…

The lack of promotion incentives is a widespread phenomenon. According to research by Jianxing Yu and Xiang Gao, more than 70 percent of surveyed officials in Zhejiang stated that they know their terminal position early in their careers. Therefore, since early in their career, many local officials decide not to participate in the promotion tournament.

This is a good nice description of how officials manage to work the system to get assigments of their choice. 

Since promotion opportunities are rare, township leaders have developed different career calculations. Some are primarily motivated by the goal of “escaping the township.” Many township leaders keep their families in the county center to access better housing and educational opportunities for their children, while residing in township government dormitories during the workweek. Consequently, they aspire to secure an administrative reassignment (平调) to the county government as a section chief, which would allow them to work closer to their families. Obtaining such a reassignment depends largely on strong personal connections with county leadership and the Organization Bureau. Others, however, prefer to remain in the township because they value the benefits of “holding real power” (大权在握). Many leaders enjoy the authority that comes with being “number one” (一把手) – and the ability to leverage this nearly unchecked power to enrich themselves. For such figures, an administrative reassignment to the county government is seen as a career downgrade.

The similarity with India’s district and below level bureaucracy is remarkable. 

This about contract workers.

Many township governments often employ more contract workers than official bianzhi workers. They perform similar tasks and often work closely with officials. In fact, township contract workers are often more competent than their counterparts. Unlike officials, who are required to rotate posts and localities under the Civil Servant Law, contract workers often remain in the same position for their entire career. This continuity allows them to develop a deeper understanding of local conditions and acquire specialized skills… In theory, a township government can terminate a contract worker’s position at any time. However, the specialized skills that contract workers possess make them irreplaceable, granting them de facto job security… township governments are highly reluctant to fire contract workers unless they commit a crime or a serious mistake.

The primary distinction between contract workers and regular officials is that contract workers do not receive the welfare benefits that officials are entitled to. They only receive basic wage, which is lower than bianzhi workers’, and do not enjoy any bonuses or subsidies… contract workers enjoy more freedom compared to officials. They are not subject to investigations by the commissions for discipline inspection at various levels, since many of them are neither civil servants nor CCP members. Additionally, they are exempt from sitting through the grueling meetings or participating in political activities that are typically reserved for officials and Communist Party members.

Li has this summary of the local government performance.

China’s township governments are best described with one word: stagnation. One cadre captured the reality of township government succinctly: “Everyone is stuck.” Such stagnation will inevitably lead to organizational decay. The resiliency of the CCP comes from its grassroots, and a stable Chinese state depends on local governments delivering services. Decaying local governments will lead to long-term social consequences.

So what are the lessons for countries like India from the Chinese administrative structure? 

1. In India’s administrative hierarchy, the township appears akin to the block, the county to the sub-division, and the prefecture to the district. However, unlike in India, the Chinese administrative units appear to have been created keeping in mind their size and functional importance. Accordingly, the largest cities like Beijing and Shanghai have province-level status, some larger cities have prefecture-level status, and some others have county-level status. This is in contrast with India where all municipalities and districts are alike in their administrative powers. In the Chinese administrative structure, Mumbai and Bangalore would have the powers associated with state governments. 

2. The bianzhi are similar to the field staff in local governments in India, and the Vice-Section level and above forms its leadership. In terms of career progression, in most states, their progression is limited to their local government. In this respect, their prospects are similar to their Chinese counterparts. Further, local governments in India are generally are headed by officials appointed from the state governments. 

The portrait of the bianzhi painted by Li is not flattering. They appear no more or less motivated or institutionally primed (to perform) than their Indian counterparts. As Yuen Yuen Ang and others have pointed out, these levels of government are also rife with petty corruption and speed money. This requires more understanding. 

It is possible that most of the government actions in terms of economic growth happens by way of policy decisions on economic growth and investment happens at the level of prefecture and provinces. The frontline bureaucracy, represented by the township level, may be merely delivering public services, for which China’s higher baseline of state capability may be sufficient. 

3. The role of contract workers appear similar to that in India. It is not uncommon across government levels in India for contract employees, or consultants, to have become more indispensable than the regular employees. The Chinese frontline state appears as much dependent on the non-regular employees as India’s. 

All these raises questions concerning the frontline functionaries and the state they represent. Is the median office clerk or teacher or nurse or tax inspector or policeman or land official in China more capable and/or more motivated than their Indian counterpart? Or is it that there are more of them than in India, thereby reducing the administrative burden? Or is the system less politicised at the frontline, given the single, unified Commuist Party in China? Or is it about greater administrative discipline and bureaucratic efficiency in monitoring and execution?

4. The Chinese officials at the larger county and at the prefecture levels appear similar to the IAS officers, who though enter laterally. The Xuandiaosheng, a Party-run track for top university graduates, but starting in village or township posts and then fast-tracked, appears the nearest to a selected elite stream. 

In fact, one study by Genia Kostka and Yu Xiaofan of career backgrounds of 898 municipal (or prefecture-level) Party secretaries between 1990-2011, found that only 5.4% had mainly county-level careers, and under 1% came directly from a county post. About 56% were mainly municipal, 32% provincial and 3% central. The study points to a “glass ceiling” for county cadres.

Our analysis finds that obstacles hindering leading county cadres from rising to the upper ranks of the municipal apparatus include age and education restrictions; limited positions at the municipal level; the lack of networks linking to provincial leaders; and provincial leaders’ appointment preferences. Limited career advancement opportunities for county cadres have important implications for understanding the incentive foundations of China’s bureaucracy. With only slim prospects for leading county cadres to rise to top municipal positions, the political incentives outlined in the cadre evaluation system might not be as effective in steering cadre behavior as currently claimed in the literature.

This trend is also reflected in the career profiles of top government and Party leaders like Xi Jinping and Li Qiang. 

5. The appointments and promotions into the government and Party at each level are managed by one level above, by the Organisation Department of that level. Interestingly, unlike other Departments, this Department functions as a Party body and not a government ministry. The centre appoints ministers, governors and provincial Party secretaries, and each Party committee manages cadres one level below it. So the province, not Beijing, picks prefecture heads. This points to a significant level of delegation of responsibilities in personnel management and a highly capable, objective, and professional Organisation Departments across levels.

Since most officials are also Party members, they move routinely between the Party and the government. As you go up, careers alternate between the two tracks. 

6. The mainstream accounts of China’s bureaucracy paints a picture of managed competition among local governments as the driver of its success. This calls for a highly capable set of leadership level officials, institutional capabilities, and strong alignment of institutional and personnel incentives. These accounts describe a highly competitive and empowered system that also aligns the career progression of officials to jurisdictional performance in expanding the economy and creating jobs. This requires a highly meritocratic set of incentives and their disciplined compliance and enforcement from the top. 

The deeper institutional innovation in China may be that of local governments (prefectures and provinces) being able to capture some of the returns to local economic development while their officials were simultaneously evaluated on that development and their career progression was dependent on the evaluation. This innovation has been underpinned by managed competition among local governments. 

7. Even with this administrative structure, the fuel that fired competition was financial resources - revenues from land sales (and the raising of debt through Local Government Financing Vehicles, LGFVs), share of tax revenues from local economic activity, the profits from ownership of stakes in State owned enterprises. This resource stream is also important since the local governments receive very little provincial or central support and are responsible for over 85% of the government spending. Further, China does not have property tax, and after 1994, the central government dispensed with the retention of marginal revenues beyond the contracted base and recentralised a substantial share of tax revenues. 

There are very few countries which can even remotely closely mimic this institutional structure and the scale of financial resource availability at the local government level. 

Imagine an Indian parallel, where larger cities and districts, headed by IAS officers, are empowered to sell lands, set up off-balance sheet entities to raise debt and finance infrastructure, establish companies and subsidise them, court private companies to build factories and provide them bespoke incentives, mobilise funds and invest in innovative sectors, etc. This would be far beyond their capabilities. Even if the capabilities were miraculously mobilised, creating the institutional and personnel incentives to do these activities with reasonable fidelity would be difficult, if not impossible. It must be borne in mind that even state governments in India struggle to do these activities with any degree of effectiveness. In fact, even the central government has struggled to effectively manage the disbursal of its flagship innovation fund. 

There are so many questions that come to mind. 

How does the Chinese adminstrative system maintain its discipline given, say, the discretion available to mayors and governors in picking winners and allocating them large amounts, or controlling state-backed venture funds (as outlined here)? How does this discipline co-exist with several accounts of widespread corruption and aggrandisement among senior officials at these levels? Or is it that the centre condones these excesses to some extent as long as the headline outcomes on economic output and jobs are delivered? Or is it that the extent of administrative powers and financial resources available endow local governments with the levers to generate economic outcomes even with the indiscipline? Or is it that the magnitude and scorching pace of economic growth has offset all the inefficiencies? 

How much of the success of the Chinese bureaucracy in delivering outcomes is due to the presence of a parallel Party structure, especially given that the Party Secretary outranks the mayor or governor? How much of the discipline within the government bureaucracy is due to the Party oversight? How much is it due to the personnel management of the government itself being done by the Party-controlled Organisation Department? How much of the discipline arises from the incentives on rising up the hierachy? How much of the motivation and drive to realise national goals among leaders is due to their Party membership and indoctrination? 

It is difficult to not come away with a strong belief that the Chinese administrative system is sui-generis and may offer limited lessons for others to emulate in both reforming their own less effective administrative systems and also in creating institutional structures to enable and sustain high growth rates. 

I had blogged here on China’s management of their public sector units, another sui-generis arrangement which might be difficult to replicate in countries like India.

Weekend reading links

1. The labour market effects of AI is not clear.

Even at the entry level, graduate unemployment has been rising since 2015, pointing to factors beyond AI.

2. China will grow old before it gets rich?

China’s total population peaked at 1.4 billion in 2021 and is now falling by three to four million people per year—roughly equivalent to losing one Los Angeles each year. By the 2040s, if China continues to follow East Asia’s descent to ultralow fertility rates, it will shed around one percent of its population annually—roughly ten million people, more than the population of New York City. At the same time, the proportion of China’s population aged 65 and older will surge, doubling from 15 percent today to around 30 percent in the 2040s, and then rising to well over 40 percent by the late twenty-first century. The size of that cohort in the United States, by contrast, is projected to rise far more gradually, from around 19 percent of the population today to just over 23 percent in 2060.

The essay by Carl Minzner  talks of how the centralisation, a paranoid state, and suppression of dissent is creating the conditions for discontent and social instability. Sample this. 

In 2023, Beijing created the Central Social Work Department and charged it with monitoring those suffering from any of the “four lacks” (no spouse or children, no job or stable income, no normal social interactions, or no financial assets, such as a house or car) or any of the “five losses” (failed investments, career or life setbacks, emotional imbalances, mental illness, or failed personal relationships). Beijing views such individuals as potential security threats capable of mass attacks, such as the November 2024 rampage in which a 62-year-old man, dissatisfied with the property settlement after his divorce, drove his car into a crowd and killed dozens.

3. Only 6.6% of Indians earn a middle class monthly salary of Rs 25,000. Even this is likely on the higher side, and I am not even talking about the low baseline. 

4. China and EU agree tacitly on an arrangement whereby EU will impose a quota ceiling on Chinese hybrid car imports without any retaliatory action by China. 

The agreement would more than halve projected exports over the next four years, reducing the number of hybrids shipped by China by “several millions”. On the current trajectory, Chinese exports to the EU would hit 800,000 this year. In 2027 alone, the deal would reduce imports by between 500,000 and 700,000 vehicles, cutting the EU’s expected trade deficit by between €10bn and €14bn, according to an EU official. The 2027 annual quota was expected to be initially about 350,000 vehicles, increasing by 6 per cent a year for the following three years, the official said. The EU’s trade deficit in goods with China hit €360bn in 2025 and has increased since. In the second quarter of 2026, it was €103bn, up from €93bn in the same period last year. Vehicles and machinery were the biggest contributors... The share of the EU market for plug-in hybrid electric vehicles controlled by Chinese imports rose to 25 per cent in the first half of this year and for other hybrid electric vehicles to 14 per cent, both up from 2 per cent in 2024.

5. China's debt service burden surges. 

Debt servicing payments alone are expected to account for 19.2 per cent of the central government’s general public budget this year, up from 12 per cent in 2014, according to a new study from the Conference Board. The problem is especially acute at the local government level, where debt interest payments have risen 16 per cent annually over the same period. Together, central and local government debt interest payments are projected to amount to Rmb1.4tn ($209bn) in 2026, up from about Rmb348bn in 2014... Public debt-to-GDP is set to hit 122 per cent by the end of 2026, up 48 percentage points from its pre-pandemic level, according to Morgan Stanley analysts... Economic growth slipped to 4.3 per cent in the second quarter, below the official full-year target range of 4.5 to 5 per cent, on the back of weakening retail sales and sharp falls in fixed-asset investment... While the central government runs a fiscal deficit of about 4 per cent, the augmented deficit, a wider measure of state spending, is running at about 10 per cent of GDP, according to Goldman Sachs... “Spending on debt interest payments rose 341 per cent” between 2013 and 2025, according to the Washington-based Center for Strategic and International Studies, “faster than any other major budget category”. The think-tank noted that about 14 per cent of US government spending went towards interest payments this year while Japan, one of the world’s most indebted governments, spent almost 26 per cent.

6. The remarkable stability of occupations.

Long-established ones are far more durable — more than 70 per cent of workers today are employed in occupations that existed in the late 19th century. Rather than eliminating these occupations, technology has transformed them.

7. Forward guidance (among other things) wins Nobel Prize precisely in the year the new Fed Chairman is taking steps to reverse it.

8. China clocks a 36.3% growth in dollar terms in its exports in September over the same month last year, the biggest increase since January-February 2021. Trade surplus for the first nine months was $938.1 bn, compared to $875 bn last year. 

This comes even as economic growth slows, reporting 4.3% in the second quarter, one of the lowest quarterly growth rates in decades and below the full year target band of 4.5-5%. 

9. While subsidies were a major contributor to Chinese companies becoming dominant in manufacturing, they have been far sighted enough to sharply increase their R&D spending and move up the value chain and become global manufacturing leaders across the value chain. 

10. Richard Waters puts the AI investment in perspective.

To take one recent estimate, from the Brookings Institution: Spending on the AI services produced by all those data centres would need to account for an astounding 9.2 per cent of US GDP by 2032 for their operators to make a decent economic return... The five biggest hyperscalers — Alphabet, Amazon, Meta, Microsoft and Oracle — have raised their capex budgets throughout the year and currently expect to spend $800bn, nearly twice as much as last year. That is set to eclipse their combined forecast cash flow of around $700bn... according to the Brookings report, it will take an extraordinary jump in AI demand to justify the construction boom. By 2032, customers would need to spend $3.7tn a year (the 9 per cent of US GDP). Tech research firm Gartner came up with a similar calculation this week, estimating that $8.5tn of AI revenue will be needed between now and 2030 to produce decent investment returns on all the new data centres. The actual AI revenue Gartner forecasts over that period: $3.6tn.