A school with a building, a salaried teacher and a free midday meal can still teach little if the teacher does not turn up. A ration shop stocked by the state can still turn away an eligible family if the dealer is never checked. Most failures that practitioners see in the field are failures of governance: the rules exist and the money is spent, but the people paid to deliver are not answerable to the people they serve.

The World Bank's Worldwide Governance Indicators (WGI) define governance as "the traditions and institutions by which authority in a country is exercised". The definition is broad on purpose. It covers how governments are chosen and replaced, whether they can make and carry out sound policy, and whether citizens and the state respect the institutions that govern their interactions.

The WGI summarise governance in six composite indicators, published annually for more than 200 economies from 1996 to 2025 in the 2026 release. Each dimension combines many underlying survey and expert questions. The descriptions below are short summaries of what each dimension tries to capture.

The 2026 WGI draw on perception data from 35 cross-country sources: household surveys, firm surveys and expert assessments. More than 400 underlying questions are mapped to the six dimensions, rescaled from 0 to 1, and combined. The release reports estimates in a standard statistical unit and as scores on a 0–100 scale.

Governance indicators are useful for comparing broad patterns across countries and over long periods. They are a poor guide to what is happening in a particular district, department or scheme, which is where most practitioners work. Treat them as a starting question, then go and look.

From the 1990s, donors began to attach governance conditions to aid: civil service reform, anti-corruption commissions, transparency laws. Critics argued that the list of desirable reforms grew longer than any poor country could manage, and that copying the institutions of rich countries rarely produced their results. Lant Pritchett, Michael Woolcock and Matt Andrews later called this isomorphic mimicry: adopting the forms of functional states while the function stays absent (Section 11).

South Asian states have expanded rights-based entitlements faster than their capacity to deliver them. India legislated a right to information in 2005 and guaranteed rural employment, now under the VB-G RAM G Act 2025 with 125 days a year. Nepal rebuilt its whole system of local government under the 2015 Constitution. The gap between what the law promises and what a citizen receives is where governance lives.

Accountability is a relationship between two parties. One holds power or resources on behalf of the other. The first must explain and justify what it did, and the second must be able to impose a consequence if the explanation is not good enough. Remove either part and what remains is weaker: explanation without sanction is public relations, sanction without explanation is arbitrary.

The political scientist Guillermo O'Donnell drew a widely used distinction in 1998. Vertical accountability runs from citizens to rulers, mainly through elections and a free press. Horizontal accountability runs between state agencies: courts, auditors, ombudsmen and legislatures checking the executive.

Catalina Smulovitz and Enrique Peruzzotti (Journal of Democracy, 2000) added societal accountability: citizens' associations, movements and media that expose wrongdoing and push horizontal agencies into action between elections. In South Asia the standard example is the Mazdoor Kisan Shakti Sangathan, which used public hearings (jan sunwais) against corruption in Rajasthan's public works in the 1990s.

The World Bank's World Development Report 2004, Making Services Work for Poor People, mapped service delivery as a triangle of citizens or clients, politicians and policymakers, and providers. It defined four relationships of accountability, and its vocabulary is still used in most donor governance work.

In a market, the customer pays the seller and can walk away, so accountability is direct. For health, education and water, the report notes, society has chosen to provide the service through government. Citizens must then work through the long route: voice to politicians, who use the compact to steer providers. The short route is client power over providers directly.

The report illustrates the failure with a survey of primary health care facilities in Bangladesh, which found the absentee rate among doctors to be 74 percent. Doctors posted to remote areas are rarely monitored, so the penalty for absence is low. The figure is old and should not be read as current; it is useful as a reminder of how large the gap between posts filled and services delivered can be.

The triangle is most useful as a diagnostic. Take one service failure and ask which relationship broke. The answer changes the remedy. A worked example, labelled Illustrative, follows the chain for a village where the anganwadi centre is open only two days a week.

Economists describe accountability failures as principal-agent problems. A principal wants a task done and hires an agent to do it. The agent knows more about its own effort than the principal does, and may have different goals. Unless the principal can observe effort or results and reward or sanction them, the agent can shirk, divert resources or serve someone else.

Public service delivery is a chain of principal-agent relationships. Citizens are principals for legislators, legislators for ministers, ministers for departments, departments for district officers, and so on down to the teacher or nurse. Information is lost at every link, and each agent can blame the next one down.

Textbook principal-agent fixes assume one principal with one clear goal. Public agencies face several principals with conflicting goals, and tasks that are hard to measure. These features explain why simple performance pay often disappoints in government.

The clearest evidence of agency problems in service delivery comes from unannounced visits. Nazmul Chaudhury, Jeffrey Hammer, Michael Kremer, Karthik Muralidharan and Halsey Rogers sent enumerators to primary schools and health clinics in Bangladesh, Ecuador, India, Indonesia, Peru and Uganda and recorded who was present.

A decade later, Karthik Muralidharan, Jishnu Das, Alaka Holla and Aakash Mohpal revisited schools in 1,297 villages across India. Inputs had improved a great deal: buildings, toilets, teacher numbers. Absence had fallen only modestly. They estimated the salary cost of unauthorised absence and compared two ways of getting more teaching time.

Every accountability instrument in this course is a version of one of three responses to the principal-agent problem. Knowing which one you are using helps you predict where it will fail.

Absence data invite a story of lazy workers. Field researchers find more varied causes: official duties such as election work and surveys, long travel to remote postings, unpaid salaries, unfilled posts covered by one person, and rules that reward paperwork over service. A practitioner who blames frontline staff for a system failure loses their cooperation and misses the fix.

A common working definition is the abuse of public office or entrusted power for private gain. Indian law is narrower and more specific: the Prevention of Corruption Act 1988 defines specific offences, among them a public servant taking an undue advantage, criminal misconduct and habitual offending. Development work meets corruption in several forms, and each calls for a different response.

Under the 1988 Act as first enacted, PRS Legislative Research notes, a bribe giver was charged with abetment. The Prevention of Corruption (Amendment) Act 2018 (No. 16 of 2018) treats bribe giving as an offence in its own right and, in section 9, makes commercial organisations liable for bribes given on their behalf. It also inserted section 17A, which bars police from inquiring into an offence relatable to a public servant's official decision or recommendation without prior approval.

Transparency International's Corruption Perceptions Index (CPI) scores countries from 0 (highly corrupt) to 100 (very clean) on perceived public sector corruption, combining expert and business surveys. It is the most quoted corruption number in the world. Here are the 2025 scores for South Asia, out of 182 countries ranked.

Perception indices ask experts what they think. Experience surveys ask citizens what happened to them. Transparency International's Global Corruption Barometer Asia 2020 surveyed nearly 20,000 people in 17 countries, mostly by telephone because of COVID-19, between March 2019 and September 2020.

Perception scores are slow to move, sensitive to scandals and media coverage, and silent about where in the system the problem sits. They can also be gamed: governments that want a better score may manage the appearance of reform. The Centre for Law and Democracy's RTI Rating shows a related problem from the other side: it rates the text of laws, and Afghanistan's 2014 law tops its table.

The strongest corruption research measures the gap between what was recorded and what happened. Each method below has been used in published studies cited in this course, and each can be adapted to programme monitoring at modest cost.

Anti-corruption work in India uses many instruments that are often discussed one at a time. Grouping them by what they do shows where a system is thin. A district with strong detection and no sanction produces audit paragraphs and no change; a district with harsh sanctions and no detection punishes only the unlucky.

Decentralisation moves decisions closer to citizens. The argument for it is informational and political: local governments know local needs, and local voters can see and judge results. The argument against is about capacity and capture: local bodies may lack skilled staff, and local elites may control them more easily than they could control a distant state.

India gave constitutional status to rural and urban local government through the Constitution (Seventy-third Amendment) Act 1992, which inserted Part IX on panchayats, and the Seventy-fourth Amendment Act 1992, which inserted Part IXA on municipalities. The 73rd Amendment came into force on 24 April 1993.

Article 243G says the Legislature of a State may, by law, endow panchayats with the powers and authority needed to function as institutions of self-government, including the preparation and implementation of plans for economic development and social justice, and schemes on the Eleventh Schedule matters. Article 243W says the same for municipalities and the Twelfth Schedule.

Article 243D reserves seats for Scheduled Castes and Scheduled Tribes in proportion to their population in the panchayat area, and not less than one-third of all directly elected seats for women, including within the SC and ST quotas. Offices of chairpersons are also reserved as state law provides, often by rotation.

Devolution is often assessed on three dimensions. A local government with functions but no money cannot act; with money but no staff it cannot spend well; with staff who answer to a state department it cannot direct them. Practitioners who work with panchayats should check all three before assigning them a role in a programme.

The Reserve Bank of India's study Finances of Panchayati Raj Institutions, released on 24 January 2024, used data for 2020–21 to 2022–23. It found that revenue receipts are dominated by grants-in-aid from the Centre and the states, and that panchayats raise very little from local taxes, fees and charges.

Article 243I requires the Governor to constitute a State Finance Commission every fifth year to recommend how state taxes are shared with panchayats, which taxes they may levy, and what grants they receive. Article 243Y extends the same commission to municipalities. The Union Finance Commission also recommends grants to local bodies.

Kerala is the best-known Indian case of deep devolution. An evaluation published by the Planning Commission (now hosted by DMEO) records that the state announced in July 1996 a decision to devolve 35 to 40% of plan funds to local governments, and launched the People's Plan Campaign in August 1996 with mass mobilisation through organisations including the Kerala Sasthra Sahitya Parishad.

India's neighbours have taken different routes. Nepal's 2015 Constitution created a three-tier federal system with 753 local levels, each with its own executive. Pakistan's Constitution, in Article 140A, requires each province by law to establish a local government system and devolve political, administrative and financial responsibility to elected local representatives, with elections held by the Election Commission of Pakistan.

Moving decisions to the village can move them into the hands of the village's dominant caste, landlords or contractors. Gram sabhas may be held on paper, beneficiary lists may favour the sarpanch's network, and reserved office holders may be pressured. Decentralisation increases accountability only where those left out have voice, information and some protection.

India's RTI Act gives every citizen the right to request information from any public authority and requires authorities to publish a defined set of information without being asked. It overrides the Official Secrets Act 1923 and other laws to the extent of inconsistency (section 22). Public authorities include bodies owned, controlled or substantially financed by government, and NGOs substantially financed directly or indirectly by government funds (section 2(h)).

You do not need to memorise the Act, but these provisions come up in almost every RTI request, appeal and training session. Section numbers refer to the Right to Information Act 2005.

The RTI process has a fixed sequence with time limits at every step. Missing a deadline can end a case, so record dates carefully. Section 24 exempts the intelligence and security organisations listed in the Second Schedule, except for information on allegations of corruption and human rights violations.

Section 8(1) lists the grounds on which information may be refused, among them national security and sovereignty, information forbidden by a court, legislative privilege, commercial confidence, fiduciary relationships, information that would impede an investigation, and personal information under clause (j). Section 8(3) makes most records more than twenty years old disclosable.

The 2005 Act fixed the term of Chief Information Commissioners and Information Commissioners at five years (or age 65) and tied the central commissioners' salaries to those of the Election Commission. PRS Legislative Research's summary of the 2019 Bill records that it removed the fixed term and let the central government notify the term, salaries and conditions of service for both central and state commissioners.

Section 44(3) of the Digital Personal Data Protection Act 2023 (No. 22 of 2023, assented to on 11 August 2023) substituted a new clause (j) in section 8(1) of the RTI Act. MeitY's notification G.S.R. 843(E) of 13 November 2025 brought s44(3) into force on its publication in the Gazette that day; most of the DPDP Act itself commences only on 13 May 2027.

Information about officials' conduct, beneficiary lists, muster rolls and contractor payments all relate to identifiable people. Under the substituted clause a PIO can now refuse such requests by citing s8(1)(j) alone. Two arguments remain open to requesters, and both will be tested in the commissions and courts.

A right is only as good as the body that enforces it. Satark Nagrik Sangathan's annual report card assessed all 29 information commissions for July 2024 to June 2025. Its findings show that delay and vacancy, more than the text of the law, now decide whether an RTI request succeeds.

Courts have used both the RTI Act and Article 19(1)(a) to widen access to information about public power. Two Constitution Bench decisions are worth knowing because they concern institutions at the top of the system.

Afghanistan, Bangladesh, India, the Maldives, Nepal, Pakistan and Sri Lanka all have right to information laws. The Centre for Law and Democracy's RTI Rating scores the legal text out of 150 on seven categories, from scope to appeals and sanctions. It rates the law as written and does not assess how commissions work in practice, and its India entry is based on the 2005 Act without the 2019 and 2023 amendments.

A social audit is a public verification of a programme's records by the people it is meant to serve. Muster rolls, bills, vouchers and measurement books are collected, checked against what workers and residents report, and read out at a public meeting where officials must answer. It combines the answerability of an audit with the public pressure of a hearing.

Social audit in India grew from the Mazdoor Kisan Shakti Sangathan's public hearings (jan sunwais) in Rajasthan in the 1990s, where villagers read out official spending records and compared them with what was built. Section 17 of the Mahatma Gandhi National Rural Employment Guarantee Act 2005 required the Gram Sabha to conduct regular social audits of all projects under the scheme. Andhra Pradesh built the first large-scale system: a pilot in February 2006, a mass social audit across 600 villages of Anantapur district in September 2006 and, from May 2009, an autonomous Society for Social Audit, Accountability and Transparency (SSAAT).

Section 37 of the Viksit Bharat Guarantee for Rozgar and Ajeevika Mission (Gramin) Act 2025 (VB-G RAM G Act, No. 36 of 2025) repeals MGNREGA from the appointed date, which was 1 July 2026. The new Act guarantees 125 days of wage employment and keeps social audit, but places it in different provisions.

Practitioners trained under MGNREGA will find familiar elements in the new Act and some important shifts. The Act places the audit duty on the Gram Sabha itself and pairs it with technology-based monitoring. Whether audits stay independent of the implementing agency will depend on the rules and on how states use their existing Social Audit Units.

Randomised trials of community monitoring have produced mixed results, and the differences between them are instructive. The three below are among the most cited; each is summarised from its published abstract.

Social audit is one member of a larger family. The right tool depends on whether you want to measure satisfaction, check records, or negotiate improvements with providers. Each tool below has been used across South Asia by government departments, NGOs and research groups.

Reviewing the experiments and the Indian experience together suggests a set of conditions. None is sufficient alone. Use them as a checklist before you invest in a social accountability component.

Social audits and RTI requests expose people who benefit from leakage, and those people may retaliate. RTI users and social audit resource persons in India have faced threats and violence. A programme that encourages citizens to challenge officials has a duty to reduce the risks it creates.

A citizen charter states what service a department promises, to what standard and in what time. On its own it has no legal force. Several Indian states went further and passed right to service laws, which make delivery within a notified time a legal duty and allow penalties on officials who miss it. Madhya Pradesh was the first state to enact such a law, on 18 August 2010, and Bihar followed in 2011.

The Centralised Public Grievance Redress and Monitoring System (CPGRAMS) links all central ministries and departments, states and union territories on one portal, pgportal.gov.in. Citizens can file online, through a mobile app, or at Common Service Centres. A PIB backgrounder of 9 August 2026 sets out the government's figures.

Grievance redress fixes individual cases: a pension not credited, a certificate delayed. Accountability changes the behaviour that produces such cases. A grievance system becomes an accountability tool when complaints are analysed for patterns, published, and used to sanction or retrain the units that generate them.

The Lokpal is an anti-corruption ombudsman for the Union. The Act's preamble links it to India's ratification of the UN Convention Against Corruption. The President gave assent on 1 January 2014. The Lokpal can inquire into complaints of corruption against a wide range of public functionaries.

Passing the Act was the end of a long public campaign, but appointing the institution took more than five years. The first Lokpal, retired Supreme Court judge Pinaki Chandra Ghose, took office on 23 March 2019. After his term ended in May 2022 a member held additional charge until Justice A.M. Khanwilkar was appointed Chairperson from 10 March 2024.

Outsiders see symptoms; insiders see how corruption works. The Whistle Blowers Protection Act 2014 was meant to protect public servants and others who disclose corruption. It was published as Act No. 17 of 2014 on 12 May 2014, but the government has never notified a date for it to come into force. A Lok Sabha answer of 12 December 2024 says the Act first needs amendments against disclosures affecting sovereignty and security; the 2015 amendment Bill, passed by the Lok Sabha, lapsed when the Sixteenth Lok Sabha was dissolved in 2019.

Citizens and programmes often send every problem to the same office. Matching the problem to the right channel saves months. The table summarises the main routes discussed in this course.

The CAG is India's supreme audit institution and a central horizontal accountability agency. The Constitution protects its independence by tying its removal to the process for a Supreme Court judge and by routing its reports to the legislature, away from the executive that it audits.

An audit report changes nothing until someone acts on it. In India the main link is the legislature's Public Accounts Committee, which examines CAG reports and asks departments to explain. Media coverage of major reports adds pressure. Local body accounts are audited under state law, as Article 243J provides.

Two well-known studies show that audits can reduce corruption, and that publishing them can change elections. Both come from outside South Asia, but their designs have shaped Indian research and practice.

Direct Benefit Transfer (DBT) pays subsidies and benefits straight into beneficiaries' bank accounts, usually with Aadhaar-based identification, instead of routing cash or goods through intermediaries. A PIB release of 21 April 2025 reported an assessment by the BlueKraft Digital Foundation of DBT's effects from 2009 to 2024.

Two large randomised evaluations show that payment reforms can cut leakage in Indian welfare programmes. Both were run with state governments at scale, which makes them unusually relevant to policy.

Muralidharan, Niehaus and Sukhtankar also evaluated reforms that added stricter biometric identity checks to India's largest social protection programme (NBER Working Paper 26744, 2020). Corruption fell, but at a substantial cost to legitimate beneficiaries.

Monitoring does not always need inspectors. Muralidharan, Niehaus, Sukhtankar and Weaver (AEJ Applied 2021) tested phone-based monitoring of a programme that transferred nearly a billion dollars to 5.7 million Indian farmers. In randomly chosen areas, officials were told that implementation would be measured through calls to beneficiaries.

New laws increasingly write technology into accountability. Section 24 of the VB-G RAM G Act 2025 lists biometric authentication of workers and transactions, geospatial planning with satellite imagery, mobile and dashboard monitoring of demand, works and payments, and weekly digital and physical disclosure of muster rolls, payments, sanctions, inspections and grievances.

Before designing any governance component, run this diagnostic for the service you care about. It takes a few days of document review and interviews, and it prevents the common mistake of applying a favourite tool to the wrong problem.

Once the diagnostic identifies the weak link, the choice of tool is easier. This table is a starting point drawn from the evidence in Sections 3 to 9; adapt it to your context.

Illustrative example. Families in a block report that their ration shop opens irregularly and gives less grain than entitled. A district NGO runs the diagnostic and finds: the dealer is related to the sarpanch, stock registers are not displayed, and complaints to the block supply officer go unanswered.

Many requests fail because they ask questions instead of asking for records, or because they ask for so much that the PIO can refuse as disproportionate. A good request is short, specific and tied to documents the authority already holds.

Accountability works best when it is designed in from the start, before problems appear. Whether you run a government scheme or an NGO project, the same four commitments make your own delivery answerable to the people you serve.

Governance outcomes can be measured at programme level without relying on national indices. The indicators below are concrete, observable and can be collected by a monitoring team. Pick a few that match your theory of change.

Accountability work collects sensitive information: names on beneficiary lists, testimony about officials, caste and income details. The Digital Personal Data Protection Act 2023 commences in stages (G.S.R. 843(E), 13 November 2025): its core duties in ss 3–17 apply from 13 May 2027. Section 17(2)(b) will then exempt processing necessary for research, archiving or statistical purposes, but only if the data are not used to take a decision specific to an individual and follow prescribed standards. Monitoring that acts on named cases falls outside it.

Most practitioners need officials' cooperation to get records, run hearings and fix problems. Accountability work can sour that relationship if it feels like an ambush. It can also be co-opted if the organisation depends on the department it monitors. Both risks can be managed.

In a 2009 Harvard Kennedy School working paper, Lant Pritchett argued that India's state is "flailing": its capable head, in the central ministries and elite services, is no longer reliably connected to the arms and legs of implementation. Using examples from health, education and routine tasks such as issuing driving licences, he showed agents of the state routinely failing to carry out assigned tasks, causing a large gap between de jure and de facto reality.

Pritchett, Woolcock and Andrews (Center for Global Development Working Paper 234, 2010) argued that many countries are in state capability traps: implementation capability is severely limited and improves, if at all, only very slowly. They named two techniques of failure. Andrews, Pritchett and Woolcock (World Development, 2013) proposed Problem-Driven Iterative Adaptation (PDIA) as a way out.

Devesh Kapur (Journal of Economic Perspectives, 2020) observed that the Indian state's performance runs from woefully inadequate to surprisingly impressive. It does better on macroeconomic than microeconomic outcomes, on episodic tasks with a built-in end (elections, the Kumbh Mela, vaccination drives) than on daily delivery, and where social hierarchy matters less.

Karthik Muralidharan's book Accelerating India's Development: A State-Led Roadmap for Effective Governance (India Viking, February 2024) analyses India's governance problems, especially in delivering essential public services, and sets out evidence-based reforms with an emphasis on state governments. He is co-founder of the Centre for Effective Governance of Indian States (CEGIS).

The studies cited in this course point to a cautious summary. These are patterns across specific settings. None is a universal law, and each depends on implementation quality. Use them to form hypotheses for your own context, then test.

Each point connects to a design choice, a legal duty or a measurement decision in programme work. Return to them when you review a programme's governance component.

Governance connects to most other courses in the ImpactMojo 101 Series. These are the most direct next steps, chosen to deepen the constitutional law, the policy process, the politics and the money behind the topics in this deck.