Patterns of Coalition Politics in India

Coalition Government: The Concept

A coalition is a government formed by two or more parties that individually lack a majority and agree to share executive power on negotiated terms. William Riker described coalitions as power-sharing arrangements, and the phrase is exact: what partners bargain over is not only policy but the distribution of offices, and every such arrangement carries within it the possibility of renegotiation.

• Coalition is structurally a feature of the parliamentary system rather than the presidential one. The reason is where the executive gets its authority from.
  • In a presidential system the executive is separately elected for a fixed term and does not depend on a legislative majority to survive. A President facing a hostile legislature has divided government, not a coalition.
  • In a parliamentary system the executive is drawn from the legislature and survives only while it retains the confidence of the lower house. Where no party commands that confidence alone, the only route to a government is an agreement between parties — which is what a coalition is.
  • India’s is a parliamentary system by design. Article 75(3) makes the Council of Ministers collectively responsible to the Lok Sabha, and that single clause is what converts a fragmented House into a bargaining arena.
• The immediate trigger is the hung house — a legislature in which no single party holds half the seats plus one.
  • A hung house does not compel coalition; it compels negotiation. The outcomes available are a majority coalition, a minority government surviving on outside support, or fresh elections.
  • The frequency of hung houses is a function of party-system fragmentation, and fragmentation in India is a product of social diversity meeting a first-past-the-post system that only manufactures majorities where competition is nationally two-cornered. The party-system note carries that argument; here it is the premise.
• Coalition is not an Indian pathology. It is the normal condition of government across continental Europe and much of the parliamentary world. What varies between those systems is not the fact of coalition but the institutional apparatus built around it — and that, rather than coalition itself, is where India’s difficulty lies.

The question is never whether coalitions are desirable. In a society of India’s diversity they are largely unavoidable. The question is what institutions make them work.

The Formal Theory of Coalition Formation

Coalition theory began as an application of game theory to politics, and its founding assumption is that parties are rational actors maximising a payoff. That assumption makes the theory testable — and makes India an interesting case, because India departs from almost every prediction it generates.

Riker and the size principle

• Riker’s The Theory of Political Coalitions supplies the discipline’s founding proposition. If politics is a zero-sum game over a fixed pool of offices and spoils, then every additional partner dilutes each existing partner’s share.
  • It follows that rational actors will build the smallest coalition that is still capable of winning — the minimum winning coalition. This is Riker’s size principle.
  • The corollary is that oversized coalitions are irrational: a partner brought in unnecessarily is payoff given away for nothing.
  • Riker’s model is office-seeking, not policy-seeking. It assumes parties want power and says nothing about what they want it for.
• India violates the size principle systematically and openly. The NDA of 1999 contained more than twenty parties; the United Front of 1996 contained thirteen; the NDA formed in 2024 carries far more partners than the BJP’s 240 seats required to reach 272.
  • The redundancy is deliberate: spare partners function as insurance against the defection of a pivotal one, which is how Eswaran Sridharan reads the 2024 formation.
  • The point generalises. Where support is conditional and revocable, buying more than the minimum is not irrational; it is rational risk management. Riker’s model assumes commitments are firm. In India they are not.
• His characterisation of the 2024 government is worth keeping: a surplus majority coalition without a majority party — more partners than the arithmetic requires, and no single party holding a majority of its own.

The families of coalition theory

Coalition theory since Riker has run along two tracks: theories about office, which count seats, and theories about policy, which count ideological distance.

• Minimal winning coalition. Every member is necessary — remove any one and the coalition loses its majority. The formulation goes back to von Neumann and Morgenstern.
• Minimum size coalition. Riker’s sharper version: among minimal winning coalitions, actors choose the smallest total seat surplus over the majority mark, because surplus is wasted payoff.
• Bargaining proposition. Leiserson‘s refinement: parties economise not on seats but on the number of actors, because fewer parties are easier to negotiate with, easier to hold together and cheaper to police.
• Minimal range coalition. De Swaan introduced policy: parties are arrayed on an ideological scale, and the preferred coalition has the narrowest range between its most extreme members, because narrow range means fewer disputes.
• Minimal connected winning coalition. Axelrod‘s formulation: partners should be adjacent to one another on the policy dimension, forming an unbroken chain. A coalition that jumps over an intervening party to recruit a distant one is unstable, because the gap it spans must be paid for continuously.
• Gamson’s law adds the distributive rule: office payoffs are allocated roughly in proportion to the seats each partner contributes. It is among the most robustly confirmed findings in comparative politics and among the least applicable in India, where pivotality rather than size determines what a partner extracts.

Theory	Core prediction	Indian record
Minimal winning (von Neumann and Morgenstern)	No partner is surplus to requirement	Contradicted; surplus partners are the norm
Minimum size (Riker)	Smallest seat surplus above the majority	Contradicted; oversized coalitions consistently formed
Bargaining proposition (Leiserson)	Fewest possible partners	Contradicted; thirteen-party and twenty-party alliances
Minimal range (De Swaan)	Narrowest ideological span	Contradicted; rainbow coalitions spanning left to right
Minimal connected winning (Axelrod)	Ideologically adjacent partners only	Contradicted; the National Front supported by the BJP and the Left at once
Gamson’s law	Payoff proportional to seats contributed	Contradicted; payoff proportional to pivotality, so a 12-seat party can outbid a 29-seat one

• The pattern in that table is the finding, not an embarrassment for India. Every theory in the list assumes a single national policy space in which parties can be ranked.
  • India’s competition runs on several dimensions at once — economic, secular–majoritarian, centre–state, caste and linguistic — and parties adjacent on one are distant on another.
  • With that many dimensions ideological adjacency ceases to be computable, and coalition formation reduces to arithmetic and state-level convenience. That is the structural reason Indian coalitions are rainbow coalitions, and a better explanation than mere opportunism.

Office-seeking and policy-seeking coalitions

• An office-seeking coalition forms to capture and distribute power; it bargains over which ministries, how many, and for whom.
• A policy-seeking coalition forms to enact a programme; it bargains over what the government will and will not do.
• Most Indian coalitions have been predominantly office-seeking, with a policy layer bolted on afterwards. The National Agenda for Governance and the Common Minimum Programme are the two conspicuous attempts to reverse the order — to write the policy first and let the offices follow — and they are also the two coalitions that completed full terms. That correspondence is not accidental.

Lijphart and the case for oversized coalitions

• Arend Lijphart turned Riker’s logic on its head for divided societies. Where a society is segmented by language, religion or ethnicity, majority rule permanently excludes a segment that can never become a majority.
• Consociational democracy is his answer, resting on four devices: the grand coalition of the leaders of all significant segments; a mutual veto on matters vital to each; proportionality in representation and in the allocation of funds and posts; and segmental autonomy.
• The grand coalition is precisely the oversized coalition Riker’s model condemns as irrational. For Lijphart it is the point: inclusion is the payoff, and buying a segment’s participation is cheaper than managing its alienation.
• This is the strongest theoretical defence of Indian coalition practice: an oversized rainbow coalition looks wasteful on the size principle and looks like elementary prudence on the consociational one. Kothari’s account of Congress as a grand coalition — a party internalising the bargaining a consociational system does between parties — is the same insight applied to a dominant-party era.
  • The qualification matters: India is consociational in practice without being consociational in design. There is no formal mutual veto, no proportionality rule and no constitutional guarantee of segmental inclusion. Inclusion happens because arithmetic forces it, and it can therefore stop happening when arithmetic stops forcing it.

Sartori, polarised pluralism and the centrifugal drive

• Giovanni Sartori‘s contribution to coalition analysis is the warning that fragmentation and polarisation are different things, and only their combination is dangerous.
  • Moderate pluralism — three to five relevant parties, competition centripetal and directed at the median voter — produces northern Europe’s stable alternating coalitions.
  • Polarised pluralism — more than five relevant parties, anti-system parties, bilateral oppositions that cannot combine, an occupied centre — produces a centrifugal drive in which parties are rewarded for moving away from the centre.
• India between 1989 and 1998 showed most symptoms of polarised pluralism — a crowded centre, oppositions on both flanks that could pull down a government but never jointly form one, and outbidding on identity rather than convergence on programme.
• The symptoms receded rather than the disease being cured. From 1998 competition re-bipolarised around two pre-poll alliances and the centrifugal drive weakened, which is why Indian coalition instability is concentrated in the years when the party system was fragmented and the fragments were unconnected.

Types of Coalition: A Working Classification

[Suggested image: a two-axis chart placing India’s central coalitions by pre-poll versus post-poll formation and by ideological range]

Distinction	Type A	Type B	Indian illustration
When formed	Pre-poll — seats shared, mandate sought jointly	Post-poll — assembled after results	NDA 1998 and 1999, UPA 2004 against Janata 1977, National Front 1989, United Front 1996
Legislative footing	Majority coalition — partners in government hold a majority	Minority government with outside support — supporters sustain it without joining	NDA 1999–2004 against National Front 1989–91 and United Front 1996–98
Size	Minimum winning — no surplus partner	Oversized — partners beyond the number required	Rare in India; the NDA and UPA formations and the 2024 NDA are all oversized
Ideological structure	Connected — partners adjacent on a policy dimension	Rainbow — partners drawn from across the spectrum	Left Democratic Front in Kerala against the National Front backed by the BJP and the Left together
Purpose	Positive — formed to implement an agreed programme	Negative — formed to keep a third party out	NDA 1999 with its National Agenda against Janata 1977 and the wider anti-Congress and anti-BJP fronts
Command	Core-party coalition — one partner far larger than the rest	Balanced coalition — comparable partners	NDA and UPA against the United Front, where no partner dominated

• These distinctions are predictive, not decorative. The strongest correlate of survival in the Indian record is the first row and the second strongest is the last: a coalition formed before the election and led by a party far larger than any partner usually completes its term.

Coalition Is Not New to India

The habit of treating 1989 as the beginning of Indian coalition politics is a national-level illusion. Coalition government is as old as representative government in India, and older than the Republic.

The provincial ministries of 1937

• The Government of India Act 1935 created provincial autonomy and ministries responsible to provincial legislatures, and the 1937 elections produced the first Indian governments that had to be assembled rather than simply won.
• Congress won clear majorities in several provinces but not everywhere. In Punjab the Unionist Party governed with the Akalis and the Muslim League; in Bengal and Assam ministries rested on negotiated blocs of communal, regional and independent members.
• The lesson carried into the Republic largely unlearned: the Indian coalition problem has always been one of assembling territorial and communal blocs, not of reconciling programmes.

Congress as a grand coalition

• Kothari’s reading of the Congress system is directly a coalition argument. Congress was not merely the largest party but a system — a party of consensus that internalised the country’s cleavages and settled them through factional bargaining inside itself, while other parties operated as parties of pressure on its margins.
• On that reading Congress was a grand coalition in the Lijphartian sense, and the Nehruvian programme was a standing common minimum programme negotiated between its factions.
• The implication reframes the whole topic. India did not move from single-party to coalition government in 1989; it moved from coalition inside one party to coalition between many. What changed was the visibility and location of bargaining — and bargaining between separately accountable parties is inherently noisier than bargaining inside one high command. The party-system note develops the Congress system itself.

1967 and the Samyukta Vidhayak Dal

• The fourth general election of 1967 is the real beginning of coalition government in India. Congress lost its majority in nine states; in eight of them non-Congress governments took office for the first time, and in the ninth Congress could govern only in coalition.
• These Samyukta Vidhayak Dal — united legislature party — governments appeared in Bihar, Uttar Pradesh, Punjab, Haryana, West Bengal, Odisha and Madhya Pradesh. They were rainbow and negative coalitions in their purest form, combining the Jana Sangh, the Socialists, the Communists, Swatantra and Congress dissidents on the single proposition of keeping Congress out.
• They were also spectacularly unstable, and they generated the defection epidemic — the Aya Ram Gaya Ram period — that would eventually produce the Tenth Schedule eighteen years later. The Indian coalition problem and the Indian defection problem were born in the same year, and they have never been separable since.

Janata 1977

• The Janata Party was a post-poll merger rather than a coalition — the Congress (O), Jana Sangh, Bharatiya Lok Dal and Socialists dissolved themselves into a single party after the 1977 verdict.
• It was held together by opposition to the Emergency and to Indira Gandhi, and by nothing else. Once that negative objective was achieved, the components had no shared programme to fall back on.
• It collapsed on the dual membership question — whether members drawn from the Jana Sangh could simultaneously belong to the RSS — which was less a doctrinal dispute than incompatibility surfacing under the pressure of office. Morarji Desai’s government lasted 856 days.

State coalition culture: Kerala and West Bengal

• Coalition politics has been continuously normal in the states since 1967, and state coalition culture has been markedly more mature than the Centre’s — the most consistently underweighted fact in the topic.
• Kerala is the exemplary case. Since the late 1970s two institutionalised pre-poll fronts — the Left Democratic Front and the United Democratic Front — have alternated in power, each with a recognised leading party, an agreed seat-sharing formula, a published programme and a standing coordination mechanism.
  • The 2026 verdict, returning the Congress-led UDF and displacing the LDF after two terms, is alternation working as designed — not a crisis of coalition.
• West Bengal offered the other model: the Left Front, dominated overwhelmingly by the CPI(M) with smaller partners in settled subordinate roles, governed continuously for thirty-four years from 1977 to 2011 — the longest-serving democratically elected coalition anywhere. Its stability came from asymmetry between the core party and its partners.
• The two cases isolate the variable. Kerala shows balanced coalitions can be stable when institutionalised; West Bengal shows asymmetric ones are stable almost automatically. What both had and the Centre lacked was a settled front with a known composition, leader and programme going into the election.

The Phases of Coalition at the Centre

Phase	Form	Why it held or failed
Janata 1977–80	Post-poll merger of dissimilar parties into a single party	Held only by opposition to the Emergency; no shared programme; collapsed on the dual-membership question
Charan Singh 1979–80	Minority government on Congress’s outside support, never faced the House	Support withdrawn before the confidence vote; 170 days without ever proving a majority
National Front 1989–91	Minority government with outside support from the BJP and the Left simultaneously	Withdrawal from both ends — the BJP over the rath yatra arrest, the Left over Mandal; two governments in two years
Chandra Shekhar 1990–91	Breakaway faction of 64 MPs governing on Congress’s outside support	A hostage government with no independent base; Congress withdrew within months
United Front 1996–98	Thirteen-party post-poll coalition with Congress support from outside	Two Prime Ministers, Deve Gowda and Gujral; Congress withdrew support twice; no partner large enough to anchor it
NDA 1998–99	Pre-poll alliance, majority government	Held thirteen months; fell when the AIADMK withdrew over a portfolio demand — confidence motion lost 269 to 270
NDA 1999–2004	Pre-poll alliance of over twenty parties with a National Agenda for Governance	Completed a full term — the first non-Congress government ever to do so; contentious items set aside in advance
UPA 2004–09	Pre-poll alliance plus a Common Minimum Programme, with Left support from outside	Completed the term; the Left withdrew in July 2008 over the nuclear deal and the government survived the trust vote 275 to 256
UPA 2009–14	Pre-poll alliance, larger Congress contingent	Completed the term but was paralysed — TMC and DMK exits, the 2G and coal allocation episodes, a collapse of decision-making
NDA 2014–19 and 2019–24	Pre-poll alliance with a single-party majority inside it	Allies present but not arithmetically needed; coalition in form, single-party government in substance
NDA 2024–	Pre-poll alliance in which the leading party is 32 short of a majority	Held so far; TDP and JD(U) pivotal; the BJP retained the major portfolios, so constraint has appeared more in policy than in office

The interludes and the logic of outside support

• The two interlude governments illustrate one failure mode: the hostage Prime Minister.
  • Charan Singh was made Prime Minister in July 1979 by Congress, the very party the Janata coalition had formed to defeat, and Congress withdrew support before he ever faced the Lok Sabha.
  • Chandra Shekhar in November 1990 led a breakaway of around 64 MPs sustained by Congress from outside — an arrangement in which the supporting party had every incentive to withdraw when it chose and none to prolong the government’s life.
• The point generalises to all outside support. A party supporting from outside collects the benefits of influence without the costs of responsibility: it extracts concessions, disclaims blame, and times its exit for electoral advantage. Outside support is the least stable of all coalition arrangements, and India has used it more than any comparable democracy.

The National Front, 1989–91

• The National Front, led by the Janata Dal under V.P. Singh, formed a minority government after the 1989 election with outside support from the BJP and the Left simultaneously.
• This is a textbook impossibility. On any minimal-connected-winning logic a government cannot be sustained by parties at opposite ends of the range, because any policy satisfying one supporter offends the other.
  • Mandal — the decision in August 1990 to implement the Second Backward Classes Commission’s reservation recommendation — pleased the Front’s social-justice constituency and split its support base.
  • The rath yatra and L.K. Advani’s arrest in Bihar in October 1990 caused the BJP to withdraw, and the government fell within days.
• The National Front is the clearest Indian demonstration of Sartori’s bilateral opposition: two supporting blocs that could each destroy the government but could never combine to replace it, so the only available outcome was collapse.

The United Front, 1996–98

• The 1996 election produced the most fragmented Lok Sabha to that point. The BJP as the largest single party was invited first, and Vajpayee resigned after thirteen days without facing a confidence vote.
• The United Front that followed was a thirteen-party post-poll coalition with Congress supporting from outside — a coalition whose largest component was smaller than its outside supporter.
  • Two Prime Ministers in twenty months — H.D. Deve Gowda from June 1996 and I.K. Gujral from April 1997 — both removed not by the House but by the withdrawal of Congress support, twice.
  • The Front did institutionalise something: a Steering Committee and a Common Minimum Programme were both used, and the Gujral Doctrine showed a coalition capable of a distinctive foreign-policy line. The machinery existed; the arithmetic did not.
• It is therefore the best evidence that a common programme is necessary but not sufficient. What the Front lacked was a core party large enough to make defection costly.

The NDA, 1998–2004

• The National Democratic Alliance is the turning point in Indian coalition history, and the reason is method rather than luck.
  • It was a pre-poll alliance with seat-sharing agreed before the election, so voters knew what they were electing and partners had already paid the cost of association.
  • It adopted a National Agenda for Governance which explicitly set aside the three most contentious items on the BJP’s own agenda — the Ram temple at Ayodhya, Article 370 and a uniform civil code. The core party disarmed itself in advance in exchange for partners.
  • It ran a coordination committee with a convenor, George Fernandes, as a standing mechanism for managing grievances rather than improvising through crisis.
• The 1998 government nevertheless fell after thirteen months, when the AIADMK withdrew over a refused portfolio demand and the confidence motion was lost by one vote, 269 to 270, on 17 April 1999. The alliance responded by broadening the model rather than abandoning it.
• The 1999–2004 government completed a full five-year term — the first non-Congress government ever to do so. Its record spans the Pokhran-II tests of 1998, the Golden Quadrilateral, the Sarva Shiksha Abhiyan and the Freedom of Information Act 2002.
• The counterfactual is worth stating. A BJP governing alone in 1999 would not have shelved Ayodhya, Article 370 or the uniform civil code. Coalition did not merely constrain the government; it changed the content of policy — the strongest available evidence that coalition constraint is real and not cosmetic.

The UPA, 2004–14

• The United Progressive Alliance copied the NDA’s method and extended it.
  • A pre-poll understanding in several states and a Common Minimum Programme published in May 2004, listing what the government would do and, by omission, what it would not.
  • Left support from outside with around 59 MPs, formalised through a UPA–Left Coordination Committee — the closest India has come to institutionalising outside support.
  • The National Advisory Council, chaired by Sonia Gandhi, is an unusual coalition institution with no constitutional standing: an extra-governmental body that drafted and pushed the CMP’s social legislation. It gave the coalition a policy engine outside the Cabinet and created the two-power-centre problem critics identify as the root of UPA-II’s paralysis.
• UPA-I’s legislative record is the strongest argument that coalitions can govern: the Right to Information Act 2005, MGNREGA 2005, the Forest Rights Act 2006, the Right to Education Act 2009, and the Indo-US civil nuclear agreement.
  • Several were coalition products in a direct sense: the Left pushed the employment guarantee toward a statutory right rather than a scheme, and the RTI’s final shape owed much to civil-society pressure channelled through the NAC. A single-party government with a comfortable majority would plausibly have legislated less, not more.
• The nuclear deal is the counter-case. The Left withdrew support in July 2008; the government survived the trust vote of 22 July 2008 by 275 to 256 with SP and BSP help. A core party can override a partner on a first-order national interest — but doing so costs the coalition its programmatic anchor.
• UPA-II is the standard exhibit for coalition paralysis, and the charge is largely fair.
  • TMC left over retail foreign investment and fuel pricing and DMK left in 2013, reducing the government to a minority sustained by parties formally in opposition; the 2G spectrum and coal allocation episodes broke over it, and the response was a freeze on decision-making rather than a correction.
  • The paralysis was real, but its cause was specific: not coalition as such, but a coalition whose written programme had been exhausted, whose coordination machinery had lapsed, and in which authority and accountability sat in different places.

2014 and 2019: coalition in form, majority in substance

• The BJP won 282 seats in 2014 and 303 in 2019 — the first single-party majorities since 1984. The NDA continued to exist and allies continued to hold ministries.
• But the allies were not arithmetically necessary, and everything about the period follows from that.
  • Portfolio allocation reverted to the Prime Minister’s discretion; ministers could be dropped and reshuffled at will; a partner’s withdrawal was an inconvenience rather than a threat. The Shiv Sena’s exit in 2019 and the Akali Dal’s over the farm laws in 2020 cost the government nothing.
  • Sridharan’s caution is important: even in 2014 the majority was built partly on pre-poll alliances in specific states. A single-party majority obtained through a pre-poll coalition is not the same thing as single-party dominance, even where it governs like it.
• This period is the control case for the whole argument, because it shows what disappears when coalition constraint does: the shelving of contentious items, the pre-legislative consultation, the deference to partner states. Article 370 was read down in 2019, a uniform civil code re-entered the agenda, and centralising instruments revived. The comparison with the National Agenda for Governance of 1998 is the cleanest available measure of what coalition constraint does.

2024 and after: the return of arithmetic

• In 2024 the BJP won 240 seats — 32 short of a majority — and the NDA’s 293 required TDP’s 16 and JD(U)’s 12 to stay above the line. The government was formed on 9 June 2024 with Chandrababu Naidu and Nitish Kumar as the pivotal figures.
• What has not changed is the distribution of office. The BJP retained Home, Finance, Defence and External Affairs — the four ministries that define a government — and allies received comparatively light portfolios. On a Gamson’s-law reading the allies were underpaid relative to their pivotality, which tells us they were bargaining for state-directed benefits rather than for Cabinet weight.
• What has changed is visible in policy and in process.
  • Fiscal transfers to allied states. The July 2024 budget committed ₹15,000 crore to Andhra Pradesh’s new capital with further tranches promised, plus support for the Polavaram project under the Andhra Pradesh Reorganisation Act; and roughly ₹26,000 crore of expressway and highway projects for Bihar together with a power plant, airports and medical colleges. The opposition’s charge that this was budget-making by coalition arithmetic is not obviously wrong.
  • Reversals under pressure. The lateral entry advertisement for 45 posts was withdrawn in August 2024; the Broadcasting Services Bill was shelved; the Waqf Amendment Bill went to a joint committee rather than straight through; pension reform was reversed with the Unified Pension Scheme; and the caste census decision of 30 April 2025 overturned a settled position.
  • A defeated constitutional amendment. The Constitution (131st Amendment) Bill 2026, which would have de-linked women’s reservation from the 2027 census and raised the Lok Sabha ceiling from 550 to 850, was negatived in the Lok Sabha on 17 April 2026 for want of the special majority, after which the companion Delimitation Bill 2026 was not proceeded with. A Union government unable to carry its own constitutional amendment is a rare event in Indian parliamentary history, and it is the sharpest single indicator that the arithmetic binds.
• The honest assessment is that coalition constraint has reappeared in substance, but unevenly.
  • It binds hardest on legislation requiring numbers, on fiscal allocation to allied states, and on measures that touch a partner’s core social base — reservation, federal transfers, and religious-minority legislation for a partner with a Muslim vote share.
  • It binds least on executive action, on portfolio allocation, and on foreign and security policy, where the BJP’s control of the four principal ministries leaves partners with little purchase.
  • The government’s command of the numbers remains comfortable — the September 2025 Vice-Presidential election returned the NDA candidate 452 to 300 — so this is a policy constraint, not a survival constraint.
  • Both allies have strong incentives to stay. The NDA’s 202 of 243 seats in the November 2025 Bihar election and the TDP’s dominance in Andhra Pradesh entangle each partner’s state position with the alliance — Sridharan’s lock-in effect: a partner cannot walk out without strengthening its own state rival. Stability here is a product of state-level interest, not of coalition dharma.

The Distinguishing Features of India’s Coalition Culture

Bidyut Chakrabarty‘s comparison of Indian and Western coalition practice is the most useful single frame in the topic, because it locates the problem in culture and method rather than in coalition as such.

Feature	Western practice	Indian practice
Basis of formation	Coalition by design — an agreed architecture of alliance	Coalition by political calculation — still searching for a design
Size	Minimum winning coalitions	Oversized coalitions
Ideology	Ideological convergence weighed before formation	Rainbow coalitions across incompatible positions
Purpose	Predominantly positive — to govern	Negative coalitions a regular feature — to stop someone governing
Internal method	Institutionalised consultation with partners	The core party has not practised consultation

• Coalition by design against coalition by calculation. A Western coalition begins from an agreed template: who leads, how disputes are settled, what the government will do, what happens if a partner objects. An Indian coalition typically begins from a count of seats after the votes are in, and the template, if it appears at all, is written afterwards under time pressure.
  • The consequence is that every crisis is a constitutional crisis in miniature. Germany’s partners argue inside a machinery built for arguing; India’s partners argue in public.
• Oversized rather than minimum winning. A violation of the size principle, and on Lijphart’s reading a defensible one. Chakrabarty’s criticism bites not against oversize but against oversize without an institutional structure to manage the numbers it brings in.
• Rainbow rather than connected. The National Front leaning on the BJP and the Left at once is the extreme case; the Janata Party combining the Jana Sangh and the Socialists the classic one; the SVD governments of 1967 the origin.
  • The defence: in a country of India’s diversity, ideological purity in coalition formation would leave large populations permanently unrepresented in government.
  • The cost: a rainbow coalition has no shared criterion for resolving disagreement, so there is no doctrine to appeal to and only power settles it — which is why Indian coalition disputes escalate so quickly to threats of withdrawal.
• Negative coalitions. A coalition formed to stop someone rather than to govern carries a fatal design flaw: its unifying purpose expires on the day it succeeds.
  • Janata 1977 is the paradigm — it defeated Indira Gandhi and then had nothing left to agree on.
  • The SVD governments after 1967 were negative coalitions in state after state, and almost none survived a full term.
  • The I.N.D.I.A. bloc formed in 2023 is arguably the current instance. It succeeded in its negative objective to the extent of denying the BJP a majority in 2024, and has since shown the predicted symptoms: no agreed leader, no common programme, and component parties competing against one another in West Bengal and Kerala in 2026 — the ordinary condition of a bloc that is an arithmetic rather than a project.
• The absence of consultation by the core party is Chakrabarty’s least quoted point and possibly his sharpest. Coalition management in India has meant placating partners after decisions, not including them before decisions.
  • Where consultation was institutionalised, coalitions survived — the NDA’s coordination committee, the UPA–Left committee, the Kerala fronts’ standing mechanisms. Where it was not, they did not. The correlation is close enough to be treated as causal.

The Generalisation the Record Actually Supports

The blanket proposition that coalition government is unstable does not survive contact with the Indian record. What the record shows is a sharp division inside the category of coalition, and the dividing line is when the alliance was formed and whether it wrote anything down.

• Pre-poll alliances with a written common programme have completed their terms. The NDA 1999–2004, UPA 2004–09 and UPA 2009–14 each ran a full five years. So has every Kerala front since the late 1970s, and the West Bengal Left Front for thirty-four years.
• Post-poll arrangements have not. Janata 1977–80, Charan Singh, the National Front, Chandra Shekhar and the United Front all failed, and none reached three years.
• The mechanism has four components.
  • Sunk cost. A pre-poll partner has shared seats and campaigned jointly, so it has already given up something that walking out would waste. A post-poll partner has given up nothing.
  • Mandate. Voters knew what they were voting for, so the coalition can claim a mandate as a coalition rather than as an improvisation, and cannot be accused of assembling a government the electorate did not choose.
  • The written programme. A common minimum programme states what the government will do and, more importantly, what it will not do, quarantining the issues that would otherwise break the alliance.
  • Asymmetry. Where the leading party is much larger than any partner, defection is survivable and therefore less often attempted. Where partners are comparable, every one of them is pivotal and behaves accordingly.

Indian coalitions are not unstable. Post-poll coalitions without a written programme are unstable, and India spent a decade making almost nothing but those.

The Charge Sheet Against Coalition Politics

The criticisms are real and worth stating as mechanisms rather than as a list, because most are consequences not of coalition as such but of the particular way India has conducted coalitions.

Policy paralysis

• Coalition raises the number of veto players, and each one raises the cost of any decision imposing concentrated losses — subsidy reform, labour law, land acquisition, public-sector ownership.
• UPA-II is the standard exhibit: retail foreign investment cost the government the TMC, and diesel pricing, land acquisition and the goods and services tax were delayed, diluted or abandoned.
• The counter-argument is worth taking seriously. The 1991 liberalisation, the most consequential economic reform in four decades, was executed by a minority government under Narasimha Rao; UPA-I legislated more social transformation than any single-party government since the 1970s. Paralysis is a possible outcome of coalition, not a necessary one, and its real predictors are the absence of a programme and the exhaustion of coordination machinery.

The decline of Parliament and the office of the Speaker

• Coalition shifts decision-making out of Parliament and into inter-party negotiation. The floor becomes where a bargain struck elsewhere is announced, and where partners perform their differences for their own constituencies.
• The Speaker’s office is the specific institutional casualty. In a hung house the Speaker holds decisions that determine whether a government lives — recognition of splits, the timing of disqualification proceedings, the admission of no-confidence motions.
  • The office was therefore converted into an object of coalition bargaining, and the convention that a Speaker resigns from the party is honoured almost nowhere.
  • Dignity and neutrality decline together: a Speaker whose political future depends on the survival of the arrangement he adjudicates is not a neutral adjudicator. The Speaker note carries the doctrine; the coalition point is that fragmented arithmetic made the office worth capturing.

Judicial and presidential activism

• Institutions expand into vacuums. A weakened legislature and an executive occupied with survival leave space that other institutions occupy.
• Judicial activism in the 1990s — public interest litigation, environmental jurisprudence, the Vineet Narain directions on the CBI — grew fastest in exactly the years when governments were weakest. This is less a criticism of the Court than a description of what happens when the political branch stops deciding.
• Presidential activism followed the same logic and is treated below.

Instability, frequent elections and the cost of defection

• India held general elections in 1989, 1991, 1996, 1998 and 1999 and saw eight governments in ten years where two would have been expected. Each dissolution is a fiscal cost, an administrative interruption, and a period in which no government can take a decision that outlasts it.
• Horse-trading and defection are the arithmetic’s direct products. Where a government’s life turns on a handful of members, those members acquire a market value.
  • Money and muscle power enter here: the premium on individual legislators raises the value of winning a seat by any means and holding it afterwards.
  • Ethnic and caste mobilisation becomes a bargaining resource rather than only an electoral one: a party that can credibly claim to deliver a bloc converts that claim into portfolios, so the incentive to sharpen the bloc is continuous.

Corruption and the mechanism behind it

• The corruption charge is usually asserted rather than explained. The mechanism is specific.
  • A coalition partner holding a revenue-generating ministry — telecom, coal, mines, ports, civil aviation — is not removable by the Prime Minister without risking the government.
  • Central oversight of that ministry is therefore structurally weak, because the sanction that normally backs oversight cannot be used.
  • The minister’s accountability runs to a party leader outside the government whose interest is in resources rather than in the government’s reputation.
• The 2G spectrum episode involved the telecom portfolio held by the DMK, an ally the Prime Minister could not dismiss; the coal block allocation episode spanned several governments and produced a comparably large CAG finding. Both became defining scandals of UPA-II.
  • The qualification is necessary. The Supreme Court’s cancellation of 122 telecom licences in 2012, and of coal allocations in 2014, turned on allocation by discretion rather than auction — a problem of method single-party governments have also had. Coalition did not create discretionary allocation; it removed the Prime Minister’s ability to police it.

The Prime Minister in a Coalition

Coalition converts the Prime Minister from leader to manager. The general question of prime-ministerial power belongs to the note on that office; what follows is the coalition-specific part of the argument.

• The power to allocate portfolios is lost first. In a single-party government the Prime Minister decides who holds which ministry. In a coalition, which party gets which ministry is negotiated and which individual fills it is decided by that party’s leader — often a leader who is not in the Council of Ministers at all, and sometimes not in Parliament.
• The power to reshuffle and dismiss follows it. A Prime Minister cannot drop an ally’s minister without the ally treating it as an attack on the party, so the disciplinary sanction that makes the Cabinet a hierarchy ceases to exist for much of the Cabinet.
• Agenda control weakens. Legislation is pre-cleared with partners, measures offending a partner’s base are withheld, and political capital goes on holding the arrangement together rather than directing it.
• The observation that a coalition contains more than one Prime Minister captures the psychological reality: for each partner’s ministers and cadres, their own party leader is the authority that matters.
• The countervailing point. Prime-ministerial weakness under coalition is real but conditional on asymmetry. Vajpayee after 1999 and Manmohan Singh in UPA-I both led coalitions and both governed; Deve Gowda and Gujral led coalitions and were removed by a party outside their government. The variable is the gap between the core party and its partners, not the fact of coalition.

Collective Responsibility under Coalition

“The Council of Ministers shall be collectively responsible to the House of the People.” — Constitution of India, Article 75(3)

The proposition that this principle has eroded under coalition government is largely sustainable — with an important qualification about what it was ever like.

• What collective responsibility presupposes. Three things: that ministers owe their office to the Prime Minister; that Cabinet decisions, once taken, are defended in public by every minister; and that a minister who cannot defend a decision resigns rather than dissents. The convention is British and unwritten, and the Constitution states only the first half of it.
• Coalition breaks the first presupposition, and everything else follows. A minister who owes appointment, retention and future prospects to a party leader outside the Council of Ministers has divided loyalties, and when they conflict the party leader wins, because the party leader controls the minister’s career and the Prime Minister does not.
• The visible symptom is public dissent without resignation. Coalition ministers have opposed Cabinet decisions in public, voted against government positions in their own states, boycotted Cabinet meetings, and threatened withdrawal over pending legislation — and remained in office. The resign-or-defend convention is now honoured so rarely that its breach no longer causes comment.
• The judicial position is unchanged, which is itself the gap. In S.P. Anand v. H.D. Deve Gowda the Supreme Court affirmed the constitutional standing of collective responsibility, and the Court has consistently treated a Cabinet decision as the decision of the whole Council. The doctrine is intact in law and hollowed out in practice.
• Was it ever fully real? No — and that matters to how the question is answered.
  • Collective responsibility was already partly fictional under single-party dominance. Cabinet ministers in the Indira Gandhi years were appointees implementing decisions taken elsewhere; the Prime Minister’s Office and the party high command had already displaced the Cabinet as the site of decision.
  • Coalition therefore did not destroy the principle so much as change the direction of the damage. Under dominance it eroded by concentration, the Cabinet becoming too weak to deliberate collectively; under coalition it eroded by dispersal, the Cabinet becoming too fragmented to act collectively.
• What survives is the legal shell: the Council still falls as a body on an adverse vote and the government still speaks with one voice on paper. Collective responsibility now functions as a rule of legal accountability to the House rather than as a norm of ministerial conduct — a narrower thing than the Cabinet system it was borrowed from.

Coalitions and the President

• Presidential discretion is dormant when a party has a majority and revives the moment it does not. The Constitution says nothing about whom to invite in a hung house; Article 75(1) provides only that the Prime Minister is appointed by the President.
• Three questions arise and none has a textual answer: whom to invite, in what order, and on what evidence a claim of majority should be accepted.
• The Sarkaria Commission proposed an order of preference which the Punchhi Commission endorsed: the pre-poll alliance with the largest number of seats; then the largest single party claiming support from others; then a post-poll coalition with all partners in government; and last a post-poll alliance with some partners supporting from outside.
  • The order is significant in itself. It puts the pre-poll alliance above the largest single party — an institutional endorsement of exactly the finding the Indian record supports.
• The episodes are the jurisprudence.
  • 1996: President Shankar Dayal Sharma invited the leader of the largest single party; Vajpayee resigned after thirteen days without facing the House, discrediting that test as a sufficient rule.
  • 1998: President K.R. Narayanan invited Vajpayee only after requiring letters of support from allies, with a condition to win a confidence vote within a set period. This is the most important convention-building act of the coalition era: it substituted documentary evidence for the President’s own assessment, which disciplines claimants and protects the President from the charge of partisanship.
  • 1999: after the government fell by one vote, Narayanan asked the alternative claimant for proof of support and, receiving none, dissolved the House rather than install a government that would fall. The precedent is that a claim alone does not oblige the President to act on it.
• The counter-risk is real. Discretion exercised well by Narayanan is discretion available to be exercised badly, and the Governor’s parallel discretion in the states has repeatedly been used to favour a preferred claimant. The President note carries this question; the coalition point is that hung arithmetic is what returns the head of state to the political process at all.

Coalitions and Federalism

This is the most important positive consequence of coalition politics, and it is the one most often understated.

• Regional parties inside national government pulled policy toward state interests. From 1989 the Union executive contained parties whose entire electorate was in one state and whose survival depended on delivering to it. Federalism stopped being a question of constitutional interpretation and became one of daily bargaining.
• The clearest measurable effect is the collapse in the use of Article 356. President’s rule was used freely through the Congress-dominant decades and fell away sharply after 1989, for a straightforwardly political reason: the parties who would be its victims held the balance of power in Delhi, and no coalition could afford to dismiss a state government run by a partner or a potential partner.
  • S.R. Bommai v. Union of India (1994) supplied the legal restraint — the proclamation is justiciable, the floor of the House is the only proper test of majority, an Assembly cannot be dissolved before parliamentary approval, and a mala fide proclamation can be struck down and the government restored.
  • Coalition politics supplied the political restraint, and the two together are why the provision fell into disuse. Bommai alone would not have sufficed: courts strike proclamations down after the fact, while coalition arithmetic prevents them being made. The Centre–state note carries federalism generally.
• Other federal effects. The Inter-State Council was constituted in 1990; the Rajya Sabha gained significance because a coalition rarely controls it; and states acquired a voice in foreign policy — West Bengal on the Teesta agreement, Tamil Nadu on Sri Lanka.
• The counter-argument is strong. The single-party majorities of 2014 and 2019 saw a partial return of centralising instruments — the reorganisation of Jammu and Kashmir in 2019, the replacement of the Planning Commission, centrally sponsored schemes with Union branding, and friction over Governors and the GST Council’s weighting.
• And 2024 has begun to reverse it. The special-package demands of Bihar and Andhra Pradesh and their appearance in the July 2024 budget are federal bargaining of exactly the pre-2014 kind. India’s federal balance tracks the arithmetic of the Lok Sabha more closely than the text of the Constitution — an uncomfortable finding, because arithmetic changes and text does not.

Anti-Defection and Coalition Stability

• The Tenth Schedule, inserted by the Fifty-Second Amendment in 1985, disqualifies a member who voluntarily gives up membership of the party on whose ticket they were elected or who votes against a party whip — a direct response to the defection epidemic the coalition governments of 1967 had begun.
• The Ninety-First Amendment of 2003 tightened it: the one-third split exception was deleted, leaving merger of two-thirds as the only protected route, and the Council of Ministers was capped at 15% of the House to remove the incentive to defect for office.
• The law stabilised individual defection and did nothing about wholesale splitting, and coalition instability migrated accordingly.
  • Retail defection became impossible and wholesale defection became the standard technique. A faction mustering two-thirds is protected, so the object is now to capture the party itself — its name, symbol and whip — rather than to leave it.
  • The Shiv Sena split of 2022 and the Nationalist Congress Party split of 2023 are the defining cases. In both, a faction took the party’s official identity through a combination of Election Commission proceedings on the symbol and Speaker’s proceedings on disqualification, and in both the outcome was that the defectors, not the defected-from, held the party.
• The Speaker’s role is the mechanism’s weakest joint. The Tenth Schedule makes the Speaker the adjudicator, and the Speaker in a coalition legislature is a partisan officer whose own position depends on the government surviving.
  • Kihoto Hollohan v. Zachillhu (1992) held that the Speaker acts as a tribunal whose decision is subject to judicial review, and struck down the finality clause for want of state ratification — but confined review largely to the post-decisional stage.
  • That is precisely the gap that delay exploits. A Speaker who does not decide cannot be reviewed, and a disqualification petition left undecided for the remainder of a term achieves the defector’s object completely.
  • Keisham Meghachandra Singh (2020) addressed it directly: a Speaker should ordinarily decide within three months, and Parliament should reconsider vesting the power in the Speaker at all in favour of an independent tribunal headed by a retired senior judge. Neither recommendation has been legislated.
• The net effect on coalitions is double-edged. The Tenth Schedule stabilises a coalition against erosion by ones and twos — a real gain, since that is what destroyed the SVD governments — but leaves it exposed to capture of whole parties, while suppressing the individual legislator’s conscience and converting the member into an instrument of the whip.

The Positive Case

• Coalitions make Indian democracy more representative. A single-party government in a country of this size represents a plurality; a coalition of ten parties has between them won votes in nearly every region and from nearly every large social bloc — Lijphart’s argument applied to India.
• They force consultation and produce more broadly acceptable policy: a measure that has survived negotiation with partners from four regions is more likely to be implementable, because the objections have already been heard.
• They gave the states a real voice at the Centre, which no amount of constitutional drafting had achieved in the preceding four decades.
• They act as a hedge against concentration. A coalition disperses power among leaders who can each impose a cost, which is a check no constitutional provision supplies and one that operates continuously rather than after the fact.
• The output record of the coalition era is not the record of a paralysed state. Between 1998 and 2014 India conducted the nuclear tests and negotiated the civil nuclear agreement, built the Golden Quadrilateral, and legislated the Right to Information Act, MGNREGA, the Forest Rights Act and the Right to Education Act. Every year of those sixteen was a coalition year.

Reform: Coalition Dharma and Institutional Design

Coalition dharma is the term of art for the unwritten code a coalition needs — consultation before decision, restraint in public criticism, and mutual acceptance of the leading party’s primacy on national questions and the partners’ on their own. Manmohan Singh’s observation that India lacks it is the standing diagnosis. A diagnosis phrased as a missing virtue prescribes nothing, which is why the useful reform debate is institutional.

Proposal	What it fixes	What it does not fix
Constructive vote of no confidence	Instability — a government cannot be removed until a successor is agreed	Policy paralysis; may entrench a government that has lost the capacity to govern
Chancellor model	The weakness of the head of government vis-à-vis partners	Sits awkwardly in a Westminster system; concentrates power the coalition exists to disperse
Direct election of the Prime Minister by the lower house	Clarity of mandate at formation; removes the head of state’s discretion	Says nothing about what happens when the coalition breaks mid-term
Institutionalised coordination committees	Consultation failure, the most common proximate cause of breakdown	Cannot substitute for arithmetic where no partner is dominant
Written common minimum programmes as a norm	Issue-driven collapse; quarantines contentious items in advance	Cannot bind a partner that decides the cost of exit is worth paying
Anti-defection reform	Delay and partisan adjudication by the Speaker	Wholesale party capture, unless the merger provision is also revisited
Electoral and party-finance reform	The money and muscle premium coalition bargaining amplifies	The underlying fragmentation that produces hung houses

The constructive vote of no confidence

The device most often proposed for India is Article 67 of the German Basic Law.

“The Bundestag can express its lack of confidence in the Federal Chancellor only by electing a successor with the majority of its members and by requesting the Federal President to dismiss the Federal Chancellor.” — Basic Law for the Federal Republic of Germany, Article 67

• It has succeeded once in Germany, when Helmut Kohl replaced Helmut Schmidt in October 1982; Spain has used its version once, in June 2018; Hungary, Poland, Slovenia, Belgium and Israel have variants.
• Its logic is that destruction should require construction. It rules out a combination that agrees on nothing except removing a government — precisely the negative coalition Chakrabarty identifies as the Indian pattern.
• What it would not do is more instructive. It would not have saved the National Front, which fell because its supporters left and no alternative majority existed; in that situation it produces not stability but a government nobody can remove and nobody can support.
• It therefore addresses removal, not paralysis — and India’s coalition problem since 1998 has been much more paralysis than removal.

The Chancellor and Japanese models

• The Chancellor model. In Germany the Chancellor sets the general guidelines of policy for each department under the Richtlinienkompetenz, ministers run their departments within those guidelines, and the Cabinet intervenes chiefly where two ministries conflict. It is a sharper instrument than the Westminster Cabinet India actually adopted.
  • The objection is that it is a transplant into an incompatible structure. India’s Cabinet rests on collective responsibility to the House, and a head of government with directive authority over departments weakens the collegiality Article 75(3) presupposes. It would also concentrate power in the office coalition politics has served to constrain.
• The Japanese model — a Prime Minister formally elected by the lower house — settles government formation inside Parliament rather than in the head of state’s discretion, which would have eliminated the 1996 and 1999 controversies. It is the least intrusive of the imports and the least discussed.

The undramatic reforms

• The reforms most likely to work are the least dramatic: a standing coordination committee with a convenor and a fixed schedule; a published common programme treated as a norm rather than an improvisation; and greater transparency by the core party, which reduces small partners’ leverage by removing the information asymmetry they trade on.
• These distinguished the coalitions that survived from those that did not, and none requires a constitutional amendment. The rest — anti-defection reform and electoral and party-finance reform — attack the money and muscle premium that coalition bargaining amplifies rather than coalition itself.

The Comparative Picture

Country	Institutional apparatus	Result
Germany	Written coalition agreements running to hundreds of pages; constructive vote of no confidence; strong Chancellor; a 5% threshold limiting fragmentation	Long-lived governments; one successful removal since 1949
Italy	No constructive vote; extreme fragmentation; weak party discipline; frequent post-poll realignment	Governments averaging little more than a year across the post-war period
Netherlands	Highly proportional system; long, formalised cabinet-formation negotiations producing a detailed agreement	Slow formation, then stable programmatic government
Belgium	Deep linguistic segmentation; formal consociational guarantees and parity rules	Record-length formation periods; governments hold once formed
Israel	Very low threshold; many small pivotal parties; payoffs disproportionate to size	Chronic instability; repeated elections

• The comparison isolates the variable. Every state in the table is governed by coalitions; stability tracks not fragmentation but the machinery built around it — thresholds, written agreements, formation procedures, removal rules.
• India resembles Italy in culture and Germany in aspiration. Like Italy it has fragmentation without agreed procedure, post-poll realignment and partners who negotiate in public. Unlike Italy it has no proportional system to blame: its fragmentation is social rather than mechanical, so threshold reform is not available as a remedy.

Coalition is not India’s problem. The countries that manage coalition well are the countries that have written down how.

Assessment

Coalition government in India has been judged against a Westminster ideal it was never going to meet. The record does not show that coalitions cannot govern; it shows that improvised coalitions cannot govern and negotiated ones can. The three governments that completed full terms between 1998 and 2014 each had a pre-poll alliance, a written programme, a coordination mechanism and a core party much larger than its partners — and every failure lacked at least two of the four.

• The reappearance of coalition constraint after 2024 is not a regression. It restores a check that operated for a quarter of a century and lapsed for a decade, and its early effects — federal bargaining, withdrawn measures, a constitutional amendment the government could not carry — are the ordinary functioning of a parliamentary system in which a majority must be assembled rather than assumed.
• What has still not been built is the design. India learned the pre-poll alliance and the common programme by trial and error and institutionalised neither, and it has not reformed the two mechanisms on which coalition stability most directly depends: anti-defection adjudication and the Speaker’s neutrality. Chakrabarty’s charge that Indian coalitions are still searching for a design remains accurate four decades on.