From FII Dependence to DII Dominance: Behavioral Dynamics and Minskyan Risk in India’s Stock Market

Abstract

This study examines how market leadership in Indian equities has structurally shifted away from foreign institutional investors (FIIs) toward domestic institutional investors (DIIs) and mutual funds (MFs), and it evaluates the systemic risks created by this rebalancing. Using monthly transaction data from April 2007 to January 2026, we analyze evolving investment patterns among FIIs, DIIs, and MFs by employing trend analysis, Pearson’s and Spearman’s correlation analyses, phase decomposition, stationarity tests, Granger causality analysis, ARIMA modelling, and GARCH volatility estimation. Since 2021, FIIs have recorded cumulative net outflows exceeding ₹8.68 lakh crore (US$95.36 billion), while DIIs mainly led by mutual funds financed largely through Systematic Investment Plans (SIPs) have made net purchases of over ₹19.37 lakh crore (US$212.67 billion), effectively absorbing FII selling and helping to maintain elevated index levels. The trend continues with SENSEX having remained above 80,000 points through 2025 despite persistent FII disengagement. The DII share of total market purchases rose from approximately 39% in 2017 to over 54% by January 2026, documenting a structural shift in market composition. The results show that DII flows have stayed positively and significantly correlated with SENSEX, with FII flows being significantly negatively correlated. Granger causality tests suggest market-responsive rather than market-driving behavior by domestic institutions. Drawing upon Minsky’s financial instability hypothesis and behavioral finance frameworks, we interpret that prolonged domestic absorption of FII exists where direct fundamental evidence is unavailable. The Minsky-type fragility interpretation is offered as a structured hypothesis for future empirical investigation. The findings carry important implications for retail investors, fund managers, and regulators.

1. Introduction

The Indian capital market has undergone a profound structural transformation over the past two decades (Chakrabarti, 2001; Mukherjee et al., 2002). From being a market sensitive to the fluctuations in foreign institutional investors (FIIs), now formally reclassified as foreign portfolio investors (FPIs), the Indian equity market has evolved into one predominantly influenced by domestic institutional investors (DIIs), primarily mutual funds (MFs) and insurance companies (Dhingra et al., 2016; Suresh & Kok Loang, 2024; Malik et al., 2025). While this shift reflects greater market depth and financial inclusion, it introduces latent risks that have received insufficient attention in the academic literature (Kok Loang, 2024; Mahapatra & Mishra, 2020; Bagchi et al., 2022; Singh, 2026). A growing body of popular commentary and practitioner analysis has observed that FIIs have been systematically divesting Indian equities since 2021, yet the SENSEX—India’s flagship equity benchmark—has continued to hover near historic highs. The year 2022 saw FII net outflows of approximately Rs. 2.78 lakh crore (US$30.54 billion), followed by Rs. 3.04 lakh crore (US$33.40 billion) in 2024 and Rs. 3.06 lakh crore (US$33.60 billion) in 2025. Paradoxically, the SENSEX crossed 85,000 points in November 2025. This apparent contradiction, we argue, is the product of sustained and accelerating DII and mutual fund inflows, increasingly funded by the retail investor ecosystem through the SIP mechanism (Choi & Yoon, 2020; Babajide & Adetiloye, 2012; C. Cao et al., 2011).

The scale of this substitution is striking. For the period 2021 to January 2026, FII net outflows amounted to approximately Rs. 10.38 lakh crore (US$113.96 billion), while DII net inflows stood at Rs. 19.37 lakh crore (US$212.67 billion) and mutual fund equity net inflows aggregated Rs. 13.55 lakh crore (US$148.77 billion). In essence, domestic institutions have not merely offset FII selling; they have exceeded it by a ratio of nearly 1.9:1, actively inflating valuations beyond what cross-border capital alone might justify.

This paper investigates whether the structural reliance on domestic liquidity, particularly retail-driven SIP inflows to sustain elevated market valuations, constitutes an emerging asset price bubble. We draw on Minsky’s (1978, 1986, 1992) financial instability hypothesis, which posits that prolonged periods of stability and rising asset prices generate the very conditions for financial fragility, and on the behavioral finance literature on herding and sentiment-driven investment (Pertiwi et al., 2019; Lu & Li, 2023; Hwang & Salmon, 2004; Gavrilakis & Floros, 2021). Our analysis utilizes monthly data from April 2007 to January 2026, including the Global Financial Crisis (GFC) of 2008, the COVID-19 shock of 2020, and the post-COVID-19 taper tantrum of 2021–2022 (Goodell, 2020).

The remainder of the paper is structured as follows: Section 2 reviews the relevant literature; Section 3 describes the data and methodology; Section 4 presents empirical findings; Section 5 discusses implications and risks; and Section 6 concludes with policy recommendations.

This study makes three distinct academic contributions. First, it provides one of the most comprehensive empirical analyses of the FII-to-DII structural transition in India using a near-two-decade monthly dataset, going beyond existing practitioner commentary to rigorously quantify the scale and timing of this shift. Second, it introduces novel metrics, the DII absorption ratio (DAR), mutual fund absorption ratio (MFAR), and DII dominance share (DDS), to systematically track domestic substitution of foreign capital. Third, it offers an integrated theoretical framework linking Minsky’s financial instability hypothesis with behavioral finance perspectives on narrative-driven retail participation, contextualizing India’s structural market shift within established theories of financial fragility. In doing so, the paper extends the literature on emerging market capital flows by demonstrating that domestic liquidity can sustain equity valuations through periods of sustained foreign exit and that this very sustainability may create the conditions for future instability.

2. Literature Review

2.1. FII and DII Dynamics in Emerging Markets

The relationship between foreign institutional flows and equity market performance in emerging economies has been extensively studied (Naik & Padhi, 2015; Feng et al., 2020; J. Cao et al., 2024; Panda et al., 2025). Bae, Chan, and Ng (Bae et al., 2004) demonstrate that foreign investors tend to be momentum traders, buying into rising markets and exiting during corrections, amplifying volatility in host markets. In the Indian context, Gordon and Gupta (2003) established that FII inflows are significant positive predictors of SENSEX returns, a finding corroborated by Mukherjee, Bose, and Coondoo (Mukherjee et al., 2002) who employed Granger causality analysis to show bidirectional causality between FII flows and equity indices. Chakrabarti (2001) noted that FII flows are primarily return-chasing rather than fundamentals-driven in India, making domestic markets vulnerable to abrupt reversals. More recent work by Dhingra, Gandhi, and Bulsara (Dhingra et al., 2016; Banerjee, 2011) confirms that while FII flows are positively correlated with market returns over shorter horizons, they introduce significant tail risk during global risk-off episodes. There is a growing literature on DII countercyclicality wherein domestic institutions absorb FII selling and especially DII purchases systematically increase when FII net selling intensifies, suggesting a stabilizing role. (Naik & Padhi, 2015; Alsahali et al., 2025; Froot et al., 2001).

More recent scholarship has begun to examine the role of domestic institutional investors in emerging market price discovery and stability (Forbes & Rigobon, 1999; Froot et al., 2001; Alomran & Alsahali, 2023). Studies document that DII flows have become increasingly trend-following, partially mirroring the momentum behavior previously attributed to FIIs (Tesar & Werner, 1995; Choe, 1999; Titman et al., 1995). Voronkova and Bohl (2005), studying Polish pension funds, demonstrate that domestic institutional investors can structurally support prices over extended periods but amplify corrections when redemption pressures arises in a dynamic environment directly relevant to India’s SIP ecosystem (Urquhart & Mcgroarty, 2016). In the context of retail investor participation, Kumar and Lee (2006) show that retail order flow driven by sentiment is more persistent than traditionally assumed, creating durable price pressure rather than transient noise. Gahlot (2019) examines the growing participation of first-generation Indian retail investors through digital platforms and finds significant herding tendencies correlated with media narratives, providing micro-level foundations for the aggregate patterns observed in this study. Collectively, this emerging body of work positions India’s DII-SIP ecosystem as a distinctive structural feature of its equity markets, warranting the focused empirical investigation this paper undertakes. On bubble diagnostics, the broader literature suggests that sustained institutional buying can depress realized volatility below its fundamental level, though the direction of causation between flows and prices remains empirically ambiguous (Chang et al., 1999; Da et al., 2015). This finding is corroborated by the Granger causality results, showing that market movements lead to institutional flows rather than the reverse. This paper’s empirical contribution is therefore best positioned at the intersection of two strands: the structural capital flow literature documenting the FII-to-DII transition and the market stability literature on the implications of domestically driven liquidity regimes.

2.2. Retail Investor Participation and Mutual Fund Flows

The democratization of Indian equity investing through the Systematic Investment Plan (SIP) mechanism represents one of the most significant structural shifts in the market’s history. AMFI data indicate that monthly SIP contributions crossed Rs. 25,000 crore (US $ 2.75 billion) in 2024–25, with over 90 million active SIP accounts1. SIP-driven inflows exhibit lower sensitivity to short-term market volatility compared to lump-sum equity investments, creating a relatively stable, recurring demand for equity assets (Kaminsky et al., 2000; Grinblatt & Keloharju, 2000).

However, this stability may carry a Minsky-type paradox. As Kindleberger and Aliber (2005) elaborated upon Minsky’s framework, the more stable a market appears, supported by steady inflows, the more investors and institutions take on risk, amplifying potential future instability. The SIP narrative, reinforced by distributor networks and favorable regulatory messaging, may be creating an expectation among retail investors that equity markets offer risk-adjusted returns superior to other asset classes, irrespective of valuation levels (Bae et al., 2002; Shleifer, 1986).

2.3. Bubble Identification and Market Overvaluation

The theoretical foundation for identifying asset price bubbles rests on the divergence between intrinsic value and market price. Shiller (2009) advanced the concept of irrational exuberance, noting that sustained price appreciation unsupported by earnings growth creates conditions for eventual violent mean reversion. The cyclically adjusted price-to-earnings (CAPE) ratio and the market capitalization-to-GDP ratio, which are among the few key Buffett Indicators, have been frequently cited to argue that Indian equities are structurally overvalued (Chakrabarti, 2001; Singh, 2026).

Brunnermeier (2008) notes that bubbles are characterized by three features: unsustainable price trajectories, concentration of leverage or liquidity risk in particular agent types, and narrative-driven investor behavior. Often the three elements are present to varying degrees, with indexes having more than doubled since 2020 lows without commensurate earnings growth and domestic retail investors increasingly leveraged via derivatives and margin financing (Gordon & Gupta, 2003; Ang et al., 2006).

2.4. Low-Volatility Indices and Systemic Risk Concealment

The low volatility anomaly is one of the key concepts often used to explain recent DII behavior in financial markets. In this anomaly, low-volatility stocks outperform high-volatility ones on a risk-adjusted basis (Shiller, 2009; Ang et al., 2006; Shleifer, 1986). However, Baker, Bradley, and Wurgler (Baker et al., 2011) caution that this anomaly can be attributed to benchmark constraints and investor sentiment and may reverse sharply during liquidity crises. The BSE Low Volatility Index, which tracks stocks with the lowest price volatility, offers an additional lens for understanding market risk perception. The convergence of Low Volatility Index performance with SENSEX performance (correlation = 0.993 in our sample) is consistent with conditions described in this literature as market-wide volatility compression, though the present study does not provide direct causal evidence that DII buying is the mechanism responsible. The empirical results in Section 4.6 suggest that institutional flows follow rather than lead market movements, which cautions against interpreting the low-volatility convergence as driven by domestic capital in any proven causal sense.

3. Data and Methodology

3.1. Data Sources and Description

This study utilizes monthly data from April 2007 to January 2026, comprising 226 observations across five datasets obtained from the Securities and Exchange Board of India (SEBI), BSE India, and the Association of Mutual Funds in India (AMFI).

The datasets include the following:

(i)

FII/FPI transaction data: Gross purchases, gross sales, and net purchases/sales in Indian equity markets (in ₹ crore).

(ii)

DII transaction data: Corresponding metrics for gross purchases, gross sales, and net purchases/sales.

(iii)

Mutual fund investments in equity and debt: MF-specific equity and debt transactions, available from November 2001 for extended historical context.

(iv)

BSE SENSEX monthly OHLC data: From January 2007 to February 2026.

(v)

BSE Low Volatility Index monthly closing values: From December 2015 to January 2026 (123 observations).

3.2. Analytical Framework

We employ a multi-method approach encompassing: (a) descriptive trend analysis with phase decomposition; (b) Pearson’s and Spearman’s rank correlation to assess the relationship between institutional flows and market levels; (c) structural shift analysis demarcating key inflection points in FII–DII flow dynamics; (d) crisis episode analysis examining four distinct stress periods; (e) risk identification through a Minsky-framework overlay; (f) time-series diagnostics including Augmented Dickey–Fuller unit root tests, Granger causality analysis, ARIMA modelling, and ARCH/GARCH volatility estimation. Pearson’s and Spearman’s correlations are preferred over vector autoregression (VAR) or full structural causality models as the primary analytical tool because the study’s central objective is to characterize the direction and strength of association between institutional flows and market levels across the full sample and sub-periods rather than estimate a structural causal model. VAR frameworks, while more flexible, require stationarity or explicit cointegration assumptions and impose a symmetrical bivariate structure that may obscure the fundamentally asymmetric nature of DII and FII flow dynamics. That said, to also consider the causal relationship, we present Granger causality tests on first-differenced series in Section 3.3. Importantly, all reported correlations are interpreted as measures of co-movement and association, not as evidence of causality.

3.2.1. Trend Analysis and Phase Decomposition

A linear time-trend regression is estimated for the SENSEX closing index to characterize the long-run appreciation trajectory:

S_t = α + βt + ε_t

(1)

where S_t denotes the SENSEX closing value in month t, t is a time index (t = 1, 2, …, T), α is the intercept, β is the estimated monthly appreciation (slope), and ε_t is the error term. The coefficient β is estimated by Ordinary Least Squares (OLS). The coefficient of determination R² is reported to assess the explanatory power of the linear trend.

The full sample is subsequently partitioned into five phases based on structural breaks in net FII flow direction, DII dominance emergence, and macroeconomic context.

The full analytical period (April 2007–January 2026) is subdivided into five distinct phases based on dominant flow characteristics: Phase I (2007–2012): FII dominance with cyclical DII support; Phase II (2013–2016): FII recovery and early SIP mobilization; Phase III (2017–2020): FII outflow onset and DII countercyclicality emergence; Phase IV (2021–2023): accelerated FII exit and DII structural absorption; and Phase V (2024–January 2026): DII hegemony and bubble risk crystallization.

3.2.2. Pearson’s Correlation Analysis

To assess the linear co-movement between institutional net flows and SENSEX levels, the Pearson product–moment correlation coefficient is estimated for each institutional category k (FII, DII, MF Equity) as:

r_k,S = Σ_t=1^T (F_k,t − F_k)(S_t − S)⁄√[Σ_t=1^T(F_k,t − F_k)² · Σ_t=1^T(S_t − S)²]

(2)

where F_k,t is the net flow of institutional category k in month t, S_t is the SENSEX closing value in month t, and F_k and S are their respective sample means. Statistical significance is evaluated using the t-statistic:

t = r_k,S √(T − 2)⁄√(1 − r_k,S²)  ~ t_(T−2)

(3)

All correlations are tested at the 1% significance level (α = 0.01). Given the possibility of non-normality in monthly flow distributions—due to large episodic transactions and fat tails—Spearman’s rank correlations are computed as a robustness check.

3.2.3. Spearman’s Rank Correlation

The Spearman rank correlation coefficient provides a non-parametric alternative that is robust to outliers and distributional assumptions:

ρ_k,S = 1 − [6 Σ_t=1^T d_t²]⁄[T(T² − 1)]

(4)

where d_t = rank(F_k,t) − rank(S_t) is the difference in ranks of observation t for flows and SENSEX levels respectively. Under H₀, ρ = 0, and the test statistic is approximately t-distributed with (T − 2) degrees of freedom for large T.

Bivariate Pearson’s correlations are computed between FII net flows, DII net flows, MF equity net flows, and SENSEX closing values. Given potential non-normality in monthly flow data, Spearman’s rank correlations serve as robustness checks. All correlations are tested for statistical significance at the 1% level. A DII-to-FII absorption ratio is constructed as the ratio of absolute DII net inflows to absolute FII net outflows in each calendar year to quantify the degree of domestic substitution. The share of DII gross purchases in total institutional gross purchases is computed annually to assess structural dominance trends.

3.2.4. DII Absorption Ratio

To quantify the degree to which domestic institutions offset FII selling in a given year y, we define the DII absorption ratio (DAR) as:

DAR_y = |DII Net Inflow_y|⁄|FII Net Outflow_y|

(5)

A DAR > 1 indicates that DII inflows exceeded FII outflows in absolute terms, i.e., domestic institutions more than fully offset foreign selling and injected additional net capital into the market. A DAR < 1 indicates partial absorption. When FII flows are net positive each year, DAR is not computed for that year. For the period 2021–2025, DAR is further decomposed into an MF-specific mutual fund absorption ratio (MFAR):

MFAR_y = |MF Equity Net Inflow_y|⁄|FII Net Outflow_y|

(6)

3.2.5. Institutional Dominance Share

The structural shift in market composition is captured through the DII Dominance Share (DDS), defined as the proportion of total gross institutional purchases attributable to DIIs in year y:

DDS_y = DII Gross Purchase_y⁄(FII Gross Purchase_y + DII Gross Purchase_y) × 100

(7)

DDS captures the compositional weight of domestic capital in total market activity and is distinct from net flows, as it reflects the volume of market participation regardless of the direction of net investment. A rising DDS trend indicates increasing domestic structural control over price discovery.

3.2.6. Cumulative Flow Divergence Index

To visualize the growing structural imbalance between FII outflows and DII inflows, a Cumulative Flow Divergence Index (CFDI) is constructed for the Phase IV–V period (January 2021 onwards):

CFDI_t = Σ_τ=1^t DII Net_τ + Σ_τ=1^t FII Net_τ

(8)

where τ = 1 corresponds to January 2021. When FII net flows are negative, their contribution to CFDI is negative, amplifying the divergence signal when DII net flows are simultaneously positive. A rising CFDI signals accelerating domestic liquidity injection; a persistently high CFDI in the context of flat or declining earnings growth is treated as a bubble risk indicator consistent with Minsky’s instability hypothesis.

3.2.7. Return and Volatility Analysis: SENSEX vs. BSE Low Volatility Index

For the overlapping sample period (December 2015–January 2026), monthly log-returns are computed for both indices:

R_i,t = ln(P_i,t) − ln(P_i,t−1)

(9)

where P_i,t is the closing value of index i ∈ {SENSEX, Low Volatility} in month t. Annualized volatility is estimated as:

σ_i^ann = σ_i^monthly × √12

(10)

The convergence of SENSEX and Low Volatility Index return trajectories alongside suppressed volatility in the latter is interpreted through the lens of the low-volatility anomaly literature (Ang et al., 2006; Baker et al., 2011) as a signal of artificially compressed market-wide risk perception driven by sustained DII buying activity. This study does not establish a causal mechanism between DII buying and volatility compression; the Granger results suggest that institutional flows are market-responsive, making it difficult to attribute volatility patterns to domestic capital as a proven driver.

3.2.8. Minsky Risk Overlay

While formal bubble detection tests (e.g., Phillips–Shi–Yu BSADF) are beyond the scope of this study, a qualitative Minsky Risk Score (MRS) is constructed for each phase using three binary criteria: It is emphasized that the MRS is intended as a supplementary discussion tool to organize the Minsky-type conditions narrative, not as a formal empirical test or proof of bubble dynamics. The score is author-constructed with discretionary thresholds, and its results carry no independent evidential weight beyond what the underlying component data (DAR, CAGR, DDS) already show in

Table 1. Statistical tests used and data considered.

No.	Method	Equation(s)	Purpose	Data Used
1	Linear Trend Regression (OLS)	(1)	Characterize SENSEX long-run trajectory	SENSEX monthly close
2	Pearson’s Correlation	(2), (3)	Linear co-movement of flows with SENSEX	FII/DII/MF net flows, SENSEX
3	Spearman’s Rank Correlation	(4)	Non-parametric robustness check	FII/DII/MF net flows, SENSEX
4	DII Absorption Ratio (DAR, MFAR)	(5), (6)	Quantify domestic offset of FII exit	Annual net flows by category
5	DII Dominance Share (DDS)	(7)	Structural market composition shift	Gross purchases: FII and DII
6	Cumulative Flow Divergence Index	(8)	Measure of growing liquidity imbalance	Monthly net flows post-2021
7	Log-Return and Volatility Analysis	(9), (10)	Index risk suppression assessment	SENSEX, BSE LowVol Index
8	Minsky Risk Score (MRS)	(11)	Phase-wise systemic fragility indicator	DAR, CAGR, DDS per phase

Note: All monetary values are in Indian rupees (Rs. crore). Statistical tests employ a two-tailed 1% significance threshold.

. It should not be cited or interpreted as empirical evidence of fragility. The threshold choices are grounded in the following rationale: DAR > 1 is selected because it marks the point at which domestic capital injection exceeds foreign withdrawal in absolute terms, indicating net domestic-driven price support; a SENSEX CAGR exceeding nominal GDP growth is used as a conventional proxy for valuation expansion beyond economic fundamentals; and DDS > 45% is chosen as the threshold at which domestic institutions account for a majority-approaching share of gross institutional turnover, constituting structural dominance. Readers should interpret scores of 2 or 3 as suggestive of elevated Minsky-type fragility conditions rather than confirmed financial instability.

MRS_phase = C₁ + C₂ + C₃

(11)

where C₁ = 1 if DAR > 1 (domestic flows exceeding FII outflows); C₂ = 1 if SENSEX CAGR exceeds estimated nominal GDP growth in the same period; and C₃ = 1 if DDS > 45% (structural domestic dominance). MRS ∈ {0, 1, 2, 3}, with MRS = 3 signaling the highest level of Minsky-type fragility. This criterion-based approach draws on Kindleberger and Aliber’s (2005) checklist methodology for bubble identification.

3.3. Statistical Tests

Considering the literature review of recent studies, supplementary time-series analysis is conducted. First, Augmented Dickey–Fuller (ADF) unit root tests are applied to all flow series and SENSEX levels to characterize their integration order and inform appropriate model specification. Second, Granger causality tests are performed on first-differenced series to examine whether past values of institutional flows contain incremental predictive information for future SENSEX changes and vice versa. These tests directly address the concern that correlation-based inferences may reflect common trending behavior rather than economically meaningful linkage. Third, an ARIMA(1,0,1) model is estimated on monthly SENSEX log-returns to characterize the conditional mean dynamics. Fourth, Engle’s ARCH-LM test is applied to model residuals to test for conditional heteroskedasticity, and a GARCH (1,1) model is estimated to capture volatility clustering in returns (Dantas & Oliveira, 2018).

Table 1 summarizes each method, its purpose, and the data used.

4. Empirical Findings

4.1. Long-Run Trend Analysis

As observed in

Table 2. Annual institutional net flows and SENSEX performance (Rs crore = USD 109.79 million).

Year	FII Net (₹ Cr)	DII Net (₹ Cr)	MF Equity Net (₹ Cr)	SENSEX Close (Approx.)
2008	−101,803	+72,967	+11,753	~9648
2012	+101,166	−55,800	−20,947	~19,427
2015	−20,374	+67,587	+71,000	~26,118
2017	−44,109	+90,738	+117,044	~34,057
2018	−73,212	+109,662	+113,333	~35,527
2020	+65,246	−35,663	−59,833	~47,751
2021	−91,626	+94,846	+44,780	~58,254
2022	−278,429	+275,726	+167,932	~60,840
2023	−16,325	+181,482	+168,555	~72,241
2024	−304,217	+527,438	+437,237	~78,140
2025	−306,419	+788,184	+493,875	~85,221

Source: SEBI, BSE India. Compiled by authors.

, over the full sample period, cumulative FII net flows aggregate to approximately Rs. 8.68 lakh crore (US$95.30 billion) of net outflows, while DII net inflows aggregate to approximately Rs. 22.23 lakh crore (US$244.07 billion), and MF equity net inflows aggregate to Rs. 16.69 lakh crore (US $ 183.24 billion). The divergence is remarkable: a market that was effectively FII-net-positive until 2016 has been DII-net-positive in every year from 2017 onwards, with the gap accelerating sharply from 2022.

The SENSEX rose from approximately 13,000 in April 2007 to over 85,700 in November 2025, representing a compound annual growth rate (CAGR) of approximately 10.6% over 18 years. However, this growth masks significant compositional heterogeneity: the period 2007–2017 was characterized by episodic FII-led rallies and corrections, whereas 2017–2026 increasingly reflects DII-sustained appreciation. The linear trend regression of SENSEX on time yields a slope of approximately 299 index points per month (R² = 0.86), indicating a remarkably steady upward trajectory that is inconsistent with normal cyclical market behavior in an environment of rising global interest rates.

4.2. Correlation Analysis

Table 3. Pearson’s and Spearman’s correlations of institutional flows with SENSEX.

Variable Pair	Pearson’s r	p-Value	Spearman’s ρ	p-Value
FII Net vs. SENSEX	−0.365	<0.001	−0.208	0.002
DII Net vs. SENSEX	+0.686	<0.001	+0.486	<0.001
MF Equity Net vs. SENSEX	+0.705	<0.001	+0.613	<0.001

Source: Authors.

presents the Pearson and Spearman rank correlations between institutional net flows and SENSEX closing levels over the full sample period. The results are statistically significant at the 1% level for all pairs and reveal a striking bifurcation.

The negative and significant correlation between FII net flows and SENSEX (r = −0.365) indicates that higher SENSEX levels are associated with net FII selling, suggesting that FIIs systematically use elevated market levels to book profits and repatriate capital. Conversely, the strong positive correlation of DII (r = +0.686) and MF flows (r = +0.705) with SENSEX confirms that DII buying intensifies as markets rise, a pro-cyclical pattern that is consistent with SIP-driven automatic monthly accumulation regardless of valuation. The Spearman correlations, being rank-based and robust to outliers, confirm the directional relationships, albeit at lower magnitudes, underscoring the non-linear nature of these dynamics. These are descriptive correlations between non-stationary level series and should be read exclusively as measures of co-movement. They do not establish that DII or MF flows drive market levels; the Granger causality results in Section 4.6 demonstrate the opposite temporal ordering. Any subsequent reference to the DII-SENSEX relationship in this paper should be understood within this strictly descriptive and associational interpretation

It is important to emphasize that these correlations reflect statistical co-movement and should not be interpreted as evidence of causality. Both institutional flows and SENSEX levels may be driven by common underlying factors—such as macroeconomic conditions, global risk sentiment, or earnings expectations—and the reported associations are consistent with multiple causal structures. The directional Granger causality tests in Section 4.6 provide additional evidence on the temporal ordering of these relationships. A further methodological consideration is that SENSEX levels and certain flow series are non-stationary (as confirmed by ADF tests in Section 4.6), which means that correlations computed at levels may partially reflect common trending behavior; readers should accordingly interpret these figures as descriptive characterizations of co-movement rather than structural parameters.

4.3. Phase Analysis and Structural Shifts

Our five-phase decomposition reveals a structural evolution with profound risk implications.

Phase I (2007–2012) was characterized by high FII volatility—the GFC of 2008 triggered FII outflows of Rs. 1.02 lakh crore (US $11.20 billion), the largest single-year outflow at that time—partially absorbed by DII purchases of Rs. 72,967 crore (US$8.01 billion). MF equity flows were relatively modest. Crucially, the SENSEX corrected from ~21,000 to ~8000, demonstrating that DII flows at their then-level were insufficient to prevent sharp market corrections.

Phase II (2013–2016) saw FII recovery (cumulative net inflow of Rs. 2.17 lakh crore) alongside the emergence of SIP culture. The launch of the ‘Mutual Fund Sahi Hai’ campaign in 2017 and regulatory emphasis on financial inclusion through MFs catalyzed structural changes in savings behavior.

Phase III (2017–2020) marked the inflection. FII flows turned persistently negative from 2017 (Rs. 44,109 crore (US$4.84 billion) outflow) and 2018 (Rs. 73,212 crore (US$8.04 billion) outflow), while MF equity net inflows in 2017 exceeded Rs. 1.17 lakh crore (US $12.85 billion)—the first year MF flows decisively overwhelmed FII outflows. The COVID-19 year 2020 is an exception: FII flows turned positive briefly on account of global quantitative easing, while DIIs and MFs absorbed pandemic-related redemptions.

Phase IV (2021–2023) represents the critical transition to structural DII dominance. The cumulative FII net outflow over this period was Rs. 3.86 lakh crore (US$42.38 billion), while DII and MF equity collectively absorbed Rs. 5.52 lakh crore (US$60.60 billion), driving the SENSEX from ~47,000 to ~72,000. This phase coincides with aggressive US Federal Reserve rate hikes (2022–2023), which historically triggered large-scale EM outflows but were significantly cushioned by domestic flows in India.

Phase V (2024–January 2026) represents what we term DII hegemony. In 2024 alone, FII net outflows of Rs. 3.04 lakh crore were offset by DII net inflows of Rs. 5.27 lakh crore (US $57.86 billion) and MF equity net inflows of Rs. 4.37 lakh crore (US $47.98 billion). The DII absorption ratio—DII net inflows as a multiple of FII net outflows—reached 1.73 in 2024 and 2.57 in 2025. The DII share of total gross institutional purchases crossed 54% in 2025–26, compared to approximately 39% in 2017, as shown in

Table 4. DII share in total institutional gross purchases (%).

Year	DII Share (%)	Year	DII Share (%)
2017	39.7	2022	45.4
2018	42.7	2023	42.8
2019	40.4	2024	45.4
2020	38.7	2025	52.1
2021	41.1	Jan-26	54.8

Source: Authors.

.

4.4. Crisis Episode Analysis

Four stress periods are examined to assess whether domestic flows successfully supported markets and what the associated cost was:

(1) GFC 2008–09: FII outflows of Rs. 1.10 lakh crore (US$12.08 billion); DII inflows of Rs. 83,411 crore (US $9.16 billion); SENSEX fell ~55%. Domestic absorption was insufficient.

(2) COVID-19 (H1 2020): FII outflows of Rs. 69,662 crore (US$7.65 billion); DII inflows of Rs. 88,212 crore (US $9.68 billion); SENSEX fell 38% initially before recovering rapidly, aided by fiscal and monetary stimulus.

(3) Post-COVID-19 Taper 2021–22: FII outflows of Rs. 3.79 lakh crore (US $41.59 billion); DII inflows of Rs. 3.42 lakh crore (US $37.55 billion); SENSEX held broadly above 50,000—the first episode in which domestic flows genuinely offset FII selling with near-parity.

(4) FII Exit 2024–25: FII outflows of Rs. 6.11 lakh crore (US $67.09 billion); DII inflows of Rs. 13.16 lakh crore (US $144.49 billion); SENSEX remained above 80,000 throughout. This episode is unprecedented in its scale of domestic absorption.

The pattern reveals a critical asymmetry: while the scale of domestic institutional investor (DII) activity has expanded substantially, so too has the market’s reliance on this liquidity to maintain elevated valuations. Should systematic investment plan (SIP) inflows decelerate—due to economic downturns, rising unemployment, or adverse return experiences prompting widespread redemptions—the supportive buffer would erode more rapidly than commonly anticipated BSE Low Volatility Index as a signal.

4.5. BSE Low Volatility Index as a Signal

The BSE Low Volatility Index, available from December 2015, exhibits a near-perfect correlation with the SENSEX (r = 0.993). Average monthly returns are nearly identical: 1.06% for the SENSEX versus 1.05% for the Low Volatility Index. However, the Low Volatility Index displays lower standard deviation (3.80% per month versus 4.61% for SENSEX), consistent with its construction methodology.

The convergence of both indices in trajectory, with lower dispersion in the Low Volatility Index, is paradoxically concerning. Ang et al. (2006) note that when low-volatility strategies converge with broad market performance, it signals widespread volatility suppression rather than genuine low-risk conditions. In our sample, DII buying has effectively created an artificial floor, reducing observed market volatility below its fundamental level, which is a form of Minsky’s ‘stability breeds instability’ in action. This pattern is consistent with the conditions Baker et al. (2011) and Ang et al. (2006) associate with volatility compression in markets experiencing sustained institutional buying. However, the paper does not provide direct causal evidence that DII flows are the mechanism responsible for this compression. As established in Section 4.6, institutional flows are market-responsive rather than market-driving, which makes a DII-led volatility suppression narrative difficult to sustain empirically. The GARCH(1,1) persistence result (α + β = 0.903) indicates that when volatility does occur it is sustained—this is indirect and circumstantial evidence consistent with a compressed-volatility interpretation but should not be presented as proof of it.

4.6. Time-Series Diagnostics: Unit Root, Granger Causality, ARIMA and GARCH Analysis

This section presents supplementary time-series diagnostic results that support the findings in terms of addressing the key concepts of stationarity, causal ordering, and return dynamics. These tests complement the correlation and trend analysis in preceding sections and are intended to provide greater methodological rigor. All tests are performed on the 226-month sample (April 2007–January 2026).

4.6.1. Unit Root Tests

Augmented Dickey–Fuller (ADF) tests are applied to all four series at levels and first differences. Results are reported in

Table 5. Augmented Dickey–Fuller unit root test results.

Series	ADF Statistic	p-Value	Integration Order	Decision
FII Net Flow (Level)	−3.540	0.007 **	I(0)	Stationary
DII Net Flow (Level)	−0.100	0.949	I(1)	Non-Stationary
MF Equity Net (Level)	−0.128	0.947	I(1)	Non-Stationary
SENSEX Level	+0.898	0.993	I(1)	Non-Stationary
ΔFII Net Flow (1st Diff.)	−11.405	<0.001 ***	—	Stationary
ΔDII Net Flow (1st Diff.)	−5.405	<0.001 ***	—	Stationary
ΔMF Net Flow (1st Diff.)	−11.186	<0.001 ***	—	Stationary
ΔSENSEX (1st Diff.)	−14.756	<0.001 ***	—	Stationary

Note: ADF tests use AIC-selected lag length. ** significant at 5%; *** significant at 1%. Δ denotes first difference.

. The SENSEX level series is strongly non-stationary (ADF = +0.898, p = 0.993), as are DII net flow (p = 0.949) and MF equity net flow (p = 0.947). FII net flow is stationary at levels (p = 0.007), likely reflecting its high month-to-month variability. All four series are stationary at first differences (p < 0.001), confirming they are integrated of an order of one, I(1), except FII net, which is I(0). This finding has an important implication for the correlation analysis reported in Section 4.2: correlations between non-stationary level series (SENSEX, DII, MF) may in part reflect common trending behavior rather than a purely structural economic linkage. Thus, correlations are treated as descriptive measures of co-movement, with the Granger causality results below in Table 5 providing the time-series-appropriate causal evidence.

4.6.2. Granger Causality Tests

Granger causality tests are estimated on first-differenced series to ensure stationarity. Two directions are tested for each flow variable: (1) whether past changes in institutional flows predict future SENSEX changes (flows → market), and (2) whether past SENSEX changes predict future flow changes (market → flows). Results are reported in

Table 6. Granger causality test results (first-differenced series).

Null Hypothesis	Lag 1 F	p-Value	Lag 2 F	p-Value	Conclusion
Panel A: Flows → ΔSENSEX (H0: Flow does NOT cause SENSEX)
ΔFII does not cause ΔSENSEX	2.420	0.121	0.777	0.461	Fail to Reject H0
ΔDII does not cause ΔSENSEX	0.246	0.621	0.627	0.535	Fail to Reject H0
ΔMF does not cause ΔSENSEX	0.114	0.736	0.200	0.819	Fail to Reject H0
Panel B: ΔSENSEX → Flows (H0: SENSEX does NOT cause Flow)
ΔSENSEX does not cause ΔFII	15.062	<0.001 ***	11.458	<0.001 ***	Reject H0
ΔSENSEX does not cause ΔDII	8.953	0.003 ***	14.914	<0.001 ***	Reject H0
ΔSENSEX does not cause ΔMF	13.226	<0.001 ***	9.980	<0.001 ***	Reject H0

Note: Tests performed on first-differenced series. Only lag 1 and lag 2 F-statistics shown for brevity; all lags 1–4 yield consistent results. *** p < 0.001. H₀ = null hypothesis of no Granger causality.

for lags 1 through 4.

The results reveal a clear and economically meaningful asymmetry. In the flows-to-market direction, none of the three institutional flow variables Granger-cause SENSEX changes at any lag (all p > 0.12). This suggests that once non-stationarity is accounted for, past changes in institutional buying or selling do not contain statistically significant incremental predictive information for future market movements. In the reverse direction, however, past SENSEX changes strongly and consistently Granger-cause changes in all three flow categories (all p < 0.001 across lags 1–4). This indicates that institutional investors—both foreign and domestic—respond to market price movements rather than leading them. The finding is consistent with momentum-chasing and return-responsive behavior documented in the broader institutional investor literature and provides important context for the level correlations reported in Section 4.2: the positive association between DII flows and SENSEX levels reflects market-responsive institutional behavior, not a unidirectional causal flow from domestic capital to price.

4.6.3. ARIMA and GARCH Volatility Analysis

An ARIMA(1,0,1) model is estimated on monthly SENSEX log-returns (expressed as percentage changes) to characterize conditional mean dynamics. The AR(1) coefficient is −0.969 and the MA(1) coefficient is +0.997, both highly significant (p < 0.001). The near-cancellation of the AR and MA terms is consistent with near-random-walk returns, where short-term serial dependence is largely removed after accounting for one lag. The Jarque–Bera statistic (197.75, p < 0.001) confirms non-normality in residuals, with excess kurtosis of 7.42, consistent with fat-tailed return distributions typical of equity markets.

Engle’s ARCH-LM test (five lags) applied to ARIMA residuals yields a test statistic of 17.19 (p = 0.004), strongly rejecting the null hypothesis of no conditional heteroskedasticity. This confirms that SENSEX monthly returns exhibit volatility clustering: periods of high volatility tend to be followed by periods of high volatility, and vice versa. A GARCH(1,1) model is therefore estimated, yielding an ARCH coefficient of α = 0.263 and a GARCH coefficient of β = 0.640. The sum α + β = 0.903 indicates high volatility persistence. These results are consistent with the suppressed-volatility narrative developed in Section 4.5: while DII buying may dampen realized volatility in the short run, the underlying conditional variance process remains highly persistent, implying that volatility, once it reasserts itself, is likely to be sustained. Results are summarized in

Table 7. ARIMA(1,0,1), ARCH-LM and GARCH(1,1) results—SENSEX monthly returns.

Parameter/Test	Coefficient/Statistic	p-Value
Panel A: ARIMA(1,0,1) Mean Model
Constant (μ)	0.793	0.059
AR(1)	−0.969	<0.001 ***
MA(1)	+0.997	<0.001 ***
Jarque–Bera (residuals)	197.75	<0.001 ***
Panel B: ARCH-LM Test (5 lags)
LM Statistic	17.190	0.004 ***
Panel C: GARCH(1,1) Volatility Model
ARCH (α)	0.263	0.471
GARCH (β)	0.640	0.131
α + β (Persistence sum)	0.903	—

Note: ARIMA estimated on monthly log-returns (%) of SENSEX, n = 225. ARCH-LM test uses 5 lags on ARIMA residuals. GARCH(1,1) estimated with normal distribution and robust standard errors. *** p < 0.001.

.

5. Discussion: Bubble Risk and Policy Implications

5.1. The Minsky Framework Applied to Indian Markets

Minsky’s (1986) financial instability hypothesis identifies three stages of borrower types: hedge (cash flows cover principal and interest), speculative (cash flows cover only interest), and Ponzi (cash flows cover neither, requiring asset appreciation for solvency). Applied to the Indian mutual fund equity ecosystem, the SIP-driven retail investor base shows signs that may indicate a shift toward more speculative behavior. Returns expectations appear calibrated to recent market performance. We observe that redemptions remain low because markets have not been corrected sharply and fund managers are compelled to deploy fresh inflows into an increasingly concentrated and overvalued set of large-cap equities. We emphasize that these observations are consistent with a Minsky-type reading but do not, by themselves, constitute proof of speculative or Ponzi dynamics. They should be read as cautionary indicators warranting further monitoring rather than as established findings of financial instability. A critical limitation must be stated prominently here: the paper’s dataset does not include earnings data, price-to-earnings ratios, cyclically adjusted valuation metrics, or any direct measure of market fundamentals. The Minsky-type interpretation therefore rests on flow dynamics and index co-movement alone. The paper does not demonstrate with the available data that Indian equity valuations are disconnected from fundamentals or that market prices are unsupported by earnings growth. This is an important constraint on all interpretations offered in this section. The fragility hypothesis is advanced as a plausible framework for understanding the structural shift, not as an empirically established conclusion.

The DII absorption ratio of 2.57:1 in 2025 implies that for every rupee of FII selling, domestic institutions are deploying ₹2.57. Whether this pace is sustainable depends on factors that this paper cannot assess empirically: earnings growth trajectories, macroeconomic conditions, and the sensitivity of SIP inflows to a sustained market downturn. The absorption ratio is presented as a descriptive indicator of the scale of the domestic–foreign substitution, not as evidence that a correction is imminent or that valuations are fundamentally unsupported.

5.2. Herding and Narrative Risk

Shiller (2009) introduced the concept of narrative economics, the idea that viral economic stories shape collective behavior and asset prices. The SIP narrative in India states to invest every month, stay long, and ignore volatility, which is a powerful and operationally sensible strategy in a fundamentally growing economy. However, it becomes systemically dangerous when it creates expectations of guaranteed equity outperformance and insulates fund inflows from valuation signals.

Survey evidence from AMFI and the National Council of Applied Economic Research (NCAER) suggest that a growing proportion of SIP investors are first-generation equity investors. They have been observed to experience only rising markets post-2020. Their behavior is under a sustained downturn with a risk of mass redemptions triggering forced selling by MFs, further depressing markets in a negative feedback loop, which also cannot be ignored and should not be dismissed.

5.3. Concentration and Systemic Risk

The structural concentration of domestic institutional flows in large-cap index constituents raises questions about market composition. As DII and MF inflows are channeled toward index-weighted stocks, participation in mid- and small-cap segments may become more uneven. The absence of a sustained price correction despite persistent FII selling is a descriptive observation that is consistent with domestic flows providing a degree of market support; however, this paper does not demonstrate that market signals have been suppressed or that price discovery has been distorted in any empirically verifiable sense. Without fundamental benchmarks such as earnings data or valuation ratios, it is not possible to determine from the available evidence whether the absence of correction reflects genuine resilience, a temporary liquidity effect, or a divergence from fundamentals. These are important distinctions that the current dataset cannot resolve, and the language in this section should be read as raising hypotheses rather than establishing findings.

5.4. Policy Recommendations

We offer three broad policy recommendations. First, SEBI and AMFI should enhance investor education to explicitly communicate valuation risk within SIP frameworks, moving beyond return-focused messaging to include downside scenario analysis. This suggestion can be implemented by mandatory inclusion of valuation-context disclosures in SIP account statements, like risk-grade labelling already required for debt funds. AMFI could issue periodic “market temperature” bulletins using standardized valuation metrics (e.g., Nifty P/E relative to 10-year median) to aid retail investor decision-making. Second, macroprudential oversight of MF equity concentration, particularly the monitoring of fund-level gross redemption capacity under stress scenarios, should be strengthened, potentially through mandatory liquidity stress tests analogous to those applied to banking institutions. Third, the development of more sophisticated domestic investor hedging instruments (long-dated put options, volatility products) would allow institutional investors to manage downside risk without resorting to wholesale liquidation, reducing the probability of a disorderly correction. AMFI could issue periodic “market temperature” bulletins using standardized valuation metrics (e.g., Nifty P/E relative to 10-year median) to aid retail investor decision-making. The second recommendation could be implemented through SEBI’s existing stress-testing framework for mutual funds, extended to include redemption surge scenarios calibrated at 20%, 30%, and 40% of AUM within a 30-day window. Results could be reported confidentially to SEBI on a quarterly basis, with mandatory public disclosure of aggregate sector-level liquidity coverage ratios. The third recommendation would require SEBI to facilitate market-making in long-dated equity options beyond current liquidity constraints, potentially through designated market-maker incentives or through IRDA facilitating insurance industry participation in volatility instruments as a natural hedge for guaranteed-return product liabilities.

6. Conclusions

This paper presents evidence of a structural transformation in the Indian equity market wherein FII selling has been systematically and increasingly absorbed by domestic institutions, principally mutual funds fueled by SIP inflows. Our analysis of monthly data from April 2007 to January 2026 documents that FIIs have net-sold approximately Rs. 8.68 lakh crore (US $ 95.36 billion) since 2021, while DIIs and MFs combined have invested Rs. 32.92 lakh crore (US $ 361.28 billion) over the same period—an absorption ratio without historical precedent in Indian markets.

The correlations between DII/MF flows and SENSEX levels are strong (r > 0.68), statistically significant, and directionally opposite to FII flows (r = −0.365), documenting a clear structural shift in the composition of institutional market participation. These are level-series correlations and should be read descriptively: they do not establish that DII or MF flows drive SENSEX levels. The Granger causality analysis demonstrates that the temporal ordering runs from market to flow, not from flows to market. The DII share of gross institutional purchases reached 54.8% in early 2026, compared to 39.7% in 2017, which is a structural shift in market composition that is documented robustly by this study.

Viewed through the Minsky framework, the stability of Indian equity markets during a period of historically large FII outflows may be interpreted as suggestive of structural dependence rather than underlying resilience, though this interpretation requires the important caveat that correlation and co-movement evidence alone cannot establish this claim definitively. The terms ‘structural dependence’ and ‘fragility’ are used here as interpretive labels for the documented flow pattern, not as empirically verified characterizations of the market’s underlying condition. Whether the market is genuinely fragile cannot be determined without earnings benchmarks and formal bubble detection methods that lie beyond this study’s scope. Whether this constitutes a bubble in the classical sense remains an open empirical question, dependent on future developments in corporate earnings, retail investor behavior, and global capital flows. The Granger causality results in Section 4.6 temper stronger causal claims: institutional flows respond to market movements rather than drive them, which suggests that the observed DII-SENSEX co-movement reflects adaptive investor behavior rather than unidirectional price inflation by domestic capital. The paper’s contribution lies in documenting the structural shift, quantifying its scale through novel metrics, and identifying the conditions under which fragility could emerge—conditions that the evidence suggests are present to a meaningful degree and merit close attention from regulators, market participants, and researchers.

To be precise about what this paper does and does not show: it robustly documents a structural compositional shift in Indian equity market participation from FII-led to DII-led, quantified through novel metrics (DAR, MFAR, DDS) and confirmed by nearly two decades of monthly data. It shows that DII and MF flows are strongly correlated with SENSEX levels in a descriptive sense and that the causal ordering runs from market movements to institutional flows, not the reverse. It shows that SENSEX returns exhibit volatility clustering and high persistence consistent with standard equity market behavior. The Minsky-type interpretation, the fragility hypothesis, and the structural dependence framing are offered as cautionary analytical perspectives grounded in the flow dynamics observed, not as empirically established conclusions. Future research incorporating earnings data, valuation ratios, and formal explosive-bubble detection methods (BSADF) will be necessary to test these hypotheses directly.

Future research should incorporate higher-frequency data (weekly or daily), apply regime-switching models to formally identify structural break points, and integrate corporate earnings growth data to assess the fundamental gap between earnings and valuations. International comparisons with other emerging markets experiencing similar DII transitions—notably China and Brazil—would further contextualize the systemic implications. Most critically, three extensions are necessary to move the core hypotheses of this paper from interpretive to empirically grounded: first, cointegration analysis between SENSEX levels and corporate earnings or CAPE ratio series to formally test the fundamentals-gap hypothesis; second, application of the Phillips–Shi–Yu (Phillips et al., 2015) BSADF recursive test to identify episodes of explosive price behavior in the SENSEX sub-periods; and third, threshold or Markov-switching VAR models to examine whether the market-to-flows causal ordering identified in the Granger tests is stable across bull and bear regimes or shifts during stress episodes. Until these extensions are completed, the fragility and bubble-risk interpretations advanced in this paper should be treated as structured hypotheses, not findings.