Does Resource Nationalism Restrain Corporate Investment? Evidence from Indonesia’s Nickel Downstreaming Policy

Abstract

This paper investigates the microeconomic consequences of contemporary resource nationalism by evaluating Indonesia’s 2020 prohibition on raw nickel exports. While state-mandated downstreaming aims to capture domestic industrial rents, it imposes severe adjustment frictions on incumbent extractors. Using a Difference-in-Differences framework on a panel of publicly listed mining corporations from 2017 to 2026, we estimate the causal impact of forced vertical integration on firm valuation, equity market returns, and capital deployment. We address the small-cluster inferential challenges typical in extractive capital markets by employing dynamic event-study specifications, temporal placebos, and wild cluster bootstrap resampling. Contrary to polarized narratives predicting either sector-wide value destruction or immediate speculative windfalls, aggregate firm valuation (Tobin’s Q) remained statistically invariant to the regulatory shock. While pure-play nickel extractors initially exhibited a secondary market premium, this effect dissipated entirely under robust finite-sample corrections. The primary empirical legacy of the mandate centers on corporate investment. The evidence suggests a structural reduction in physical capital expenditure among nickel-exposed firms, as entities engaged in acute capital rationing to accommodate the massive financing requirements of mandated smelter construction, effectively restraining conventional upstream investment.

1. Introduction

The global pivot to clean energy has transformed the geopolitical economy of critical minerals and re-energized resource nationalism in emerging markets. Instead of just extracting and exporting raw commodities, governments are increasingly pursuing aggressive industrial policies to capture more of the value chain domestically. Perhaps the most consequential contemporary application of this strategy is Indonesia’s hilirisasi (downstreaming) policy. Indonesia is also the world’s biggest holder of nickel reserves, a key input for stainless steel and electric vehicle (EV) batteries. Indonesia imposed a sweeping ban on export of raw (unprocessed) nickel ore that took effect on 1 January 2020.

The macroeconomic success of this mandate is frequently cited, with its attendant surge in foreign direct investment, rapid expansion of domestic industrial parks, and many-fold increase in the export value of processed nickel, but the microeconomic burden imposed on incumbent mining corporations is poorly understood. From a corporate finance perspective, forced downstreaming introduces a profound tension between state developmental goals and shareholder value maximization.

This regulatory shock raises a critical empirical question: how do equity markets price abrupt, state-mandated vertical integration? Investors might view the required smelter construction as a value-destroying expropriation of free cash flow, burdened by severe execution risks, capital rationing, and sovereign hold-up hazards. Conversely, markets may reward the policy if captive downstream processing shields margins from upstream trade barriers and enables firms to capture long-term industrial rents in the EV battery ecosystem.

The importance of answering this question is because the current literature on Indonesia’s nickel ban has been mostly focused on geopolitical trade disputes, environmental sustainability, and aggregate macroeconomic changes. The firm-level financial transmission of this policy has largely escaped rigorous causal evaluation. As other resource-rich nations look to emulate the Indonesian model by leveraging their own critical mineral endowments, policymakers and investors urgently need empirical clarity on the corporate adjustment costs inherent in resource nationalism.

This study adds to the existing literature by investigating the causal effect of the nickel export ban in 2020 on the financial performance, capital investment, and market value of the publicly listed mining companies in the Indonesia Stock Exchange (IDX). We identify the policy shock by means of a DID approach and a balanced quarterly panel over the period 2017–2026. This study compares the net financial effects of the downstreaming imperative for firms directly exposed to the mandate (i.e., pure-play nickel extractors and diversified nickel-exposed holding companies) with a counterfactual control group of unaffected coal and non-nickel mining entities.

The methodological novelty of this paper lies in its rigorous handling of the inferential challenges inherent in extractive-sector capital markets. Recognizing that listed mining conglomerates constitute a small number of cross-sectional clusters, we move beyond conventional panel estimations by applying dynamic event-study specifications, placebo-date falsification tests, and wild cluster bootstrap resampling. This approach ensures that estimated shifts in valuation and investment are not statistical artifacts of cyclical commodity trends or small-sample downward bias.

The remainder of this article is structured as follows. Section 2 reviews the literature review on resource nationalism, capital adjustment costs, and corporate valuation. Section 3 details the institutional setting, operationalization of variables, and the structural econometric framework. Section 4 presents the empirical results, including dynamic treatment trajectories and robustness diagnostics. Finally, Section 5 concludes with policy implications for resource governance and corporate strategy.

2. Literature Review and Theoretical Framework

2.1. The Political Economy of Resource Nationalism and Ownership Risk

The contemporary shift in the governance of critical minerals reflects an evolution in resource nationalism. Historically, sovereign interventions in extractive sectors were characterized by direct state expropriation and the nationalization of foreign assets, particularly during commodity booms (Bremmer & Johnston, 2009; Stevens, 2008). Modern resource nationalism, however, operates through more sophisticated regulatory architectures—such as windfall taxes, forced divestments, and, centrally to this study, export prohibitions designed to force downstream industrialization (Mahdavi, 2014; Vivoda, 2009).

From a property rights perspective, these sudden regulatory pivots functionally erode ownership security. Secure property rights are foundational to capital-intensive extraction; when the state unexpectedly alters the rules of resource monetization, it imposes a quasi-tax on future returns (Acemoglu et al., 2001; Bohn & Deacon, 2000). Cust and Harding (2020) and Engel and Fischer (2008) formalize this dynamic, demonstrating that institutions prone to sudden regulatory shifts inherently discourage upfront capital deployment. A raw mineral export ban accelerates this ownership risk. By severing upstream extraction from global spot markets—such as the London Metal Exchange—the state forces incumbent mining firms into captive domestic supply chains, fundamentally altering their operational equilibrium and stripping away established market premiums (Humphreys, 2015; Radetzki, 2008).

The current rush for electric vehicle battery precursors is changing the landscape of mineral extraction. It is obvious that extractive host nations use their endowments to push for supply chain localization, a move that blends industrial policy with the geopolitical weaponization of critical minerals. Recent structural analyses show that export prohibitions, in emerging markets, are a tool to push foreign capital into domestic metallurgical processing. This removes the option that multinational extractive firms once had for where to refine outputs (International Energy Agency, 2023; Juhász et al., 2024; Leruth et al., 2022).

2.2. Policy Uncertainty, Expropriation, and Corporate Valuation

The interaction between state intervention and firm valuation is heavily mediated by the pricing of uncertainty. Asset pricing literature establishes that severe fluctuations in government policy command a substantial risk premium, depressing equity valuations when state behavior becomes erratic (Brogaard & Detzel, 2015; Pástor & Veronesi, 2012, 2013). This political risk premium is acutely sensitive in extractive industries, where asset mobility is zero and cash flows rely entirely on state concessions (Haslam & Ary Tanimoune, 2016; Manzano & Monaldi, 2008).

Neoclassical models heavily penalize the execution risk of forced integration yet developmental state literature offers a compelling counter framework. Aggressive downstreaming mandates can theoretically generate deep agglomeration economies. By cutting off mineral outflows, the state forces the capitalization of domestic industrial parks, domestic industrial parks then facilitate technology spillovers through mandatory joint ventures with advanced foreign off-takers (Gereffi, 2019). For mining firms, navigating initial capital shock offers a distinct long-term structural advantage. Successfully monopolizing high-value nodes within the EV battery supply chain shields domestic profit margins from upstream global trade volatility. The result of monopolizing high-value nodes, within the EV battery supply chain, can reward compliant mining firms with local monopoly rents that eventually surpass the initial sunk costs of smelter construction (Aiginger & Rodrik, 2020).

2.3. Capital Adjustment Costs and the Neoclassical Investment Framework

The friction between sudden policy mandates and physical capital allocation is best understood through the neoclassical theory of investment. The theoretical bridge between a firm’s market valuation and its optimal capital stock is captured by marginal q—the market value of an additional unit of capital relative to its replacement cost (Hayashi, 1982). While standard models assume smooth capital accumulation proportional to q (Abel & Eberly, 1994; Erickson & Whited, 2000), real-world industrial investment is characterized by non-convex adjustment costs and severe irreversibility (Caballero & Engel, 1999; Cooper & Haltiwanger, 2006).

Constructing smelting infrastructure is the quintessential irreversible investment, subject to profound lumpiness (Dixit & Pindyck, 1994). A state mandate forcing mining firms to suddenly pivot from raw extraction (a relatively low capital-intensity process) to complex metallurgy imposes massive adjustment frictions. Because internal capital markets are constrained, forcing multi-billion-dollar downstream investments triggers acute capital rationing (Fazzari et al., 1988). When confronted with these non-convex costs and the policy uncertainty documented by Baker et al. (2016), Gulen and Ion (2016), and Julio and Yook (2012), extractive firms are likely to exhibit periods of investment inaction. The rational corporate response to a sudden, massively expensive state mandate is often to freeze conventional upstream capital expenditure until the regulatory and technological uncertainty resolves.

2.4. Synthesis and Hypothesis Development

Bridging these distinct literatures generates testable implications for the Indonesian mineral sector. If the costs of non-convex capital adjustments and sovereign hold-up risk dominate, the implementation of the export ban should theoretically depress the Tobin’s Q of exposed firms relative to unconstrained extractive peers (Peters & Taylor, 2017). The theoretical friction between sudden state mandates, ownership risk, and neoclassical capital adjustment generates two testable propositions for the Indonesian mining sector: Hypothesis 1: The implementation of the raw nickel export ban significantly alters the aggregate market valuation (Tobin’s Q) of exposed extractive firms, reflecting the market’s pricing of execution risk versus captive supply chain advantages. Hypothesis 2: Facing acute capital rationing and non-convex adjustment costs to meet mandated metallurgical infrastructure requirements, nickel-exposed firms will significantly reduce conventional upstream physical capital expenditure.

3. Model Specification and Data

3.1. Study Setting and Data Collection

3.1.1. Institutional Context: The Indonesian Mineral Hilirization Mandate

The empirical setting of this study is grounded in Indonesia’s mineral downstreaming policy—locally termed hilirisasi—which represents one of the most aggressive exercises of contemporary resource nationalism in emerging markets. Although the mandate for domestic processing was established under Law No. 4/2009 on Mineral and Coal Mining (UU Minerba), its operational timeline was repeatedly modified through phased export allowances, quota systems, and intermediate processing exemptions. The decisive policy shock materialized through the Ministry of Energy and Mineral Resources (ESDM) Regulation No. 11/2019, which halted the export of raw, unprocessed nickel ore (Harmonized System code 2604) effective 1 January 2020—advancing the previously scheduled timeline by two years.

From a corporate finance perspective, this intervention induces competing valuation dynamics. On one hand, forced integration imposes severe capital outlays, lengthens payback periods, amplifies technological execution risk, and exposes equity holders to sovereign hold-up hazards. On the other hand, vertical integration potentially allows mining entities to capture downstream industrial rents, insulate operating margins from upstream trade barriers, and benefit from long-term industrial subsidies and corporate income tax holidays (tax holidays administered by the Ministry of Finance). Whether the equity market prices hilirisasi as a value-destroying expropriation of free cash flow or as a strategic, value-accretive structural transformation remains an unresolved empirical question.

3.1.2. Sample Selection and Classification Framework

The sample was taken from all of the mining sector corporations listed on the Indonesia Stock Exchange (IDX) in the observation window, every quarter from the first quarter of 2017 to the first quarter of 2026. The window spans thirty-seven consecutive quarters (T = 37) so we can construct an empirical panel that divides the 1 January 2020 policy shock into a twelve-quarter pre-treatment period (2017:Q1–2019:Q4) and a twenty-five-quarter post-treatment period (2020:Q1–2026:Q1). Financial statement data, balance sheet aggregates, and capital expenditures were compiled from quarterly corporate filings submitted to the IDX. Secondary trading metrics, adjusted stock prices, and market capitalizations were sourced from the IDX daily trading database. Macroeconomic controls, including London Metal Exchange cash settlement prices for nickel and Bank Indonesia reference exchange rates (Jakarta Interbank Spot Dollar Rate, JISDOR), were aggregated to quarterly averages.

To avoid confounding our causal estimates with attrition biases, the estimation sample is constructed as a balanced cross-section of 18 publicly listed mining firms observed across all 37 quarters, yielding an aggregate balanced structure of N × T = 666 firm-quarter observations. The treatment status of a firm is determined by its operational exposure to the raw mineral export prohibition. Because corporate structures on the IDX frequently involve multi-tiered business models—where extractive concessions are housed within partially owned operating subsidiaries—relying on a single binary classification introduces non-trivial measurement error.

To accurately capture the complex corporate ownership structures present on the IDX, policy exposure is operationalized across multiple tiers—distinguishing between pure-play nickel extractors, integrated mining conglomerates, and general mineral producers—measured against a control group of unaffected coal and non-nickel entities. This tiered approach explicitly separates direct policy exposure from indirect regulatory spillover. The broad mineral group (comprising copper, gold, and bauxite) is not framed as being directly treated by the nickel export ban. Instead, this third tier serves as a test for sector-wide regulatory contagion, allowing us to measure whether the aggressive state intervention in the nickel sector induced precautionary capital hoarding among other extractive firms fearing future downstreaming mandates. As a result, we operationalize three different treatment definitions with a hierarchical structure:

• Direct Nickel Producers (Treat pure): A targeted category made up only of pure-play concession holders whose primary source of income is nickel ore mining (N₁ = 2 businesses; 74 observations). The counterfactual control group comprises all remaining listed mining enterprises (N₀ = 16 firms; 592 observations).
• Nickel-Exposed Group (Treat exposed): A broader classification that encompasses pure-play operators alongside mining holding conglomerates that hold controlling equity stakes or significant economic interests in active nickel-mining subsidiaries and processing facilities (N₁ = 3 firms; 111 observations). The control group consists of the 15 unexposed extractive entities (N₀ = 15 firms; 555 observations).
• Mineral Sector Group (Treat mineral): An aggregate classification encompassing all metallic mineral mining companies (N₁ = 8 firms; 296 observations)—including nickel, bauxite, copper, and gold-silver producers subject to broader processing mandates—benchmarked against thermal and metallurgical coal extraction companies (N₀ = 10 firms; 370 observations). Coal miners provide an ideal counterfactual baseline because they operate within the same institutional, labor, and fiscal environment of the Indonesian extractive sector and are exposed to equivalent sovereign risk premia and currency fluctuations, yet remained completely exempt from the raw nickel export ban.

Table 1 details the distribution of observations across the pre- and post-intervention periods. Across all three tiers, the pre-treatment period accounts for exactly 32.4% of the panel (216 firm-quarters), while the post-treatment period spans 67.6% (450 firm-quarters). This structure provides an extensive post-shock observation horizon (over six full calendar years), which is critical for evaluating long-cycle capital investments in smelter infrastructure and discerning transitory adjustment costs from permanent shifts in market valuation.

Table 1. Distribution of observations across the pre- and post-intervention periods.

Subsector Grouping	Policy		Total Observations
	Pre-Policy (2017Q1–2019Q4)	Post-Policy (2020Q1–2026Q2)
Panel A: Treat Mineral vs. Coal
Coal control	120	250	370
Mineral treatment	96	200	296
Total Firms	216	450	666
Panel B: Treat Nickel-Exposed vs. Others
Non-nickel-exposed control	180	375	555
Mickel-exposed treatment	36	75	111
Total Firms	216	450	666
Panel C: Treat Direct Nickel vs. Others
Non-direct-nickel control	192	400	592
Direct nickel treatment	24	50	74
Total Firms	216	450	666

Source: Author’s calculation.

Even though the group of firms that are directly affected is small in number, this limit shows the structure of the Indonesia Stock Exchange, not a problem with sampling. These companies are the set of publicly traded nickel producers that fall under the new regulatory framework. Typical large-sample inference often fails when there are a few treated clusters, which makes results look more significant than they really are and causes many true null hypotheses to be wrongly rejected. To stop this bias, our empirical approach purposely does not use the large-sample statistical methods.

Instead, we use a wild cluster bootstrap method. By resampling the data while keeping the cluster structure intact, this technique fixes the bias that comes from samples and gives trustworthy solid p-values. As a result, this strict econometric correction automatically makes our confidence intervals wider. Using a 10% significance level, capital expenditure of the usual 5% shows a very conservative estimation strategy. This method keeps the research honest, making sure that the reported parameters really show the precision of the market data we have.

3.1.3. Descriptive Statistics and Distributional Properties

The operationalization of variables captures firm valuation, corporate physical investment behavior, asset scale, and equity market performance. Table 2 reports descriptive statistics across the pooled panel of 666 observations.

Table 2. Summary statistics.

Variable	Obs	Mean	Std. Dev.	Min	Max
Firms ID	666	9.5	5.19	1	18
Year	666	2021.14	2.67	2017	2026
Quarter	666	2.46	1.13	1	4
Tobin’s Q	628	40.99	244.31	0	5196.12
Capital Expenditure (USD ‘000)	612	338,484	1,351,650	0	22,200,000
Capex (ln)	608	9.14	3.37	0	16.92
Capex/Total Assets (TA) Ratio	612	2.48	11.98	0	121.43
Total Assets (USD ‘000)	634	18,200,000	47,800,000	0	260,000,000
Total Assets (ln)	628	13.66	2.73	0	19.38
Market Value of Equity (MVE) (ln)	634	12.49	1.93	8.09	15.92
Nickel Price in USD	666	776,081.50	3,611,928	9450	23,700,000
Treat Mineral (Dummy)	666	0.44	0.50	0	1
Treat Exposed (Dummy)	666	0.17	0.37	0	1
Treat Pure (Dummy)	666	0.11	0.31	0	1
Post (Dummy)	666	0.68	0.47	0	1
DID Mineral (Treat × Post)	666	0.30	0.46	0	1
DID Exposed (Treat × Post)	666	0.11	0.32	0	1
DID Pure (Treat × Post)	666	0.08	0.26	0	1
Placebo Post	666	0.89	0.31	0	1
Placebo DID Mineral	666	0.40	0.49	0	1
Placebo DID Exposed	666	0.15	0.36	0	1
Placebo DID Pure	666	0.10	0.30	0	1
Stock Return	648	0.06	0.37	−0.76	4.32
Share Price	666	2484.86	5358.46	6	41,425
Exchange Rate	666	14,842.84	980.97	13,310	16,853

Source: Author’s calculation from IDX corporate filings and market data.

The unadjusted baseline metric for firm valuation, Tobin’s Q—computed as the sum of market value of equity and book value of total liabilities divided by the book value of total assets—displays severe positive skewness. Table 2 shows that while the sample mean is 40.99, the median is situated near conventional valuation levels, and the sample maximum reaches 5196.12.

Meanwhile, nominal quarterly capital expenditure exhibits marked lumpiness, ranging from zero during periods of exploration or project mothballing to USD 22.2 billion during phases of major industrial expansion, with a pooled mean of USD 338.48 million. Transforming capital expenditures into natural logarithms (lnCapex) and calculating investment intensity as a share of total assets (Capex/TA) stabilizes the underlying variance: the mean log-capex is 9.14 (SD = 3.37), and the mean investment ratio is 2.48% (SD = 11.98%).

On the other hand, the log market value of equity (lnMVE) is stably distributed around a mean of 12.49 (SD = 1.93; range 8.09–15.92), making it a well-behaved proxy for absolute shareholder wealth.

Over the panel horizon, macroeconomic conditions varied significantly. Currency pressures during Federal Reserve monetary tightening cycles were reflected in the Indonesian Rupiah’s range of 13,310 to 16,853 per USD (mean = 14,842.84, SD = 980.97). Global nickel market prices, indexed to quarterly settlement benchmarks, averaged USD 776,081.5 per unit across contract variations, reflecting the price spikes associated with the 2022 LME structural squeeze.

Table 3 breaks down the data across treatment and control cohorts for each operational tier. Comparing these partitions reveals structural distinctions.

Table 3. Summary statistics of key variables partitioned by treatment firms.

	Tobin’s Q	Capex (Thousands USD)	Capex (ln)	Capex/TA Ratio	Stick Return	DID	Market Value Equity (ln)
Panel A: Mineral Firms
Coal control (N = 370)	357	351	348	351	360	370	357
Mean	56.67	209,568.10	8.84	2.14	0.06	0	12.77
SD	170.35	701,803.30	3.52	8.89	0.40	0	1.64
Min	0	0	0	0	−0.76	0	8.13
Max	818.98	5,656,835.00	15.55	115.90	4.32	0	15.92
Mineral treatment (N = 296)	271	261	260	261	288	296	277
Mean	20.33	511,853.50	9.53	2.94	0.06	0.68	12.13
SD	315.58	1,891,488.00	3.10	15.18	0.33	0.47	2.21
Min	0	0	0	0	−0.67	0	8.09
Max	5196.12	22,200,000	16.92	121.43	1.94	1	15.77
Panel B: Nickel Exposed
Non-exposed control (N = 555)	518	502	498	502	540	555	524
	49.58	398,840.70	8.92	3.01	0.06	0	12.29
	268.26	1,484,720	3.63	13.17	0.38	0	1.99
	0	0	0	0	−0.76	0	8.09
	5196.12	22,200,000	16.92	121.43	4.32	0	15.92
Exposed treatment (N = 111)	110	110	110	110	108	111	110
	0.54	63,037.85	10.13	0.05	0.08	0.68	13.46
	0.49	124,709.30	1.35	0.23	0.34	0.47	1.23
	0.07	1133	7.03	0	−0.46	0	10.60
	1.86	929,443	13.74	2.14	1.74	1	15.42
Panel C: Direct Nickel
Non-direct control (N = 592)	555	539	535	539	576	592	561
	46.28	375,054	9.03	2.81	0.06	0	12.36
	259.44	1,435,477	3.53	12.73	0.38	0	1.95
	0	0	0	0	−0.76	0	8.09
	5196.12	22,200,000	16.92	121.43	4.32	0	15.92
Direct treatment (N = 74)	73	73	73	73	72	74	73
	0.71	68,466.92	9.94	0.07	0.07	0.68	13.54
	0.52	150,403.40	1.49	0.28	0.32	0.47	1.48
	0.09	1133	7.03	0	−0.46	0	10.60
	1.86	929,443	13.74	2.14	1.74	1	15.42

Source: Author’s calculation. Number of observations reflects available reporting.

In Panel A, mineral-focused treatment companies display lower average Tobin’s Q values (20.33 vs. 56.67) but higher average physical capital outlays (USD 511.85 million vs. USD 209.57 million) and higher log-capex (9.53 vs. 8.84) than coal counterparts. Both groups display identical historical average quarterly stock returns (6.0%).

Moving to Panels B and C, isolating nickel-exposed and direct nickel firms reveals that these pure-play and integrated entities are substantially larger, more mature industrial corporations. Their average log market value of equity is 13.46 and 13.54, respectively, while the larger control groups’ average is about 12.3. Importantly, their Tobin’s Q measurements are closely spaced around economically reasonable replacement-cost multiples: direct nickel firms have an average Tobin’s Q of 0.71 (SD = 0.52; minimum 0.09, maximum 1.86), whereas exposed firms have an average of 0.54 (SD = 0.49; minimum 0.07, highest 1.86). The extreme valuation outliers that inflate the broader control samples are entirely absent from the nickel cohorts.

Furthermore, both nickel groups sustain higher baseline log-capex levels (10.13 for exposed, 9.94 for direct nickel), while their lower Capex-to-Total-Assets ratios (0.05% to 0.07%) reflect their substantial asset scale. Because unscaled Tobin’s Q contains cross-cohort heterogeneity in the control groups, empirical specifications must incorporate firm fixed effects, log-transformed alternatives, and outlier-robust sensitivity checks.

3.2. Econometric Framework

3.2.1. Theoretical Setup and Derivation of the Empirical Model

To microfound the empirical specification, we adapt Hayashi’s (1982) neoclassical investment framework to an extractive setting where the state enforces a domestic downstreaming constraint. Consider a price-taking mining enterprise i operating in quarter t, maximizing the expected present discounted value of free cash flows to equity:

$$V_{\mathit{it}}\left(K_{\mathit{it}}\right)=\underset{{\left\{I_{i,t+s}\right\}}_{s=0}^{\infty }}{\max }{\mathbb{E}}_t{\displaystyle {\sum }_{s=0}^{\infty }}{\beta }^s\mathsf{\Pi }\left(K_{i,t+s},I_{i,t+s};{\mathsf{\Omega }}_{i,t+s}\right)$$

(1)

where β ∈ (0, 1) is the quarterly discount factor, K_it denotes the firm’s productive capital stock, I_it represents gross capital expenditure, and Ω_it represents the information set containing regulatory regimes, world commodity prices ($P_t^{\mathit{Ni}}$), and exchange rates (E_t). The net cash flow function Π(⋅) is given by

Π(K_it, I_it; Ω_it) = R(K_it; Ω_it) − C(I_it, K_it) − I_it

(2)

where R(K_it;Ω_it) is the operating revenue generated from ore extraction and processing, and C(I_it,K_it) denotes internal capital adjustment costs, assumed strictly convex, homogeneous of degree one in (I,K), and parameterized as

$$C\left(I_{\mathit{it}},K_{\mathit{it}}\right)={\displaystyle \frac{\psi }{2}}{\left({\displaystyle \frac{I_{\mathit{it}}}{K_{\mathit{it}}}}-c_0\right)}^2{K}_{\mathit{it}}$$

(3)

The regulatory intervention of 1 January 2020 introduces a structural regime shift in the revenue and adjustment cost equations. Prior to the intervention (Post_t = 0), upstream firms sell raw nickel ore directly on global markets at prevailing international prices:

$$R_{\mathit{it}}^{\mathit{Pre}}=P_t^{\mathit{Ni}}\cdot Q\left(K_{\mathit{it}}\right)$$

(4)

Under the downstreaming mandate (Post_t = 1), exporting unprocessed ore is legally prohibited (Q_export = 0). For affected firms (Treat_i = 1), operating revenues depend on the construction of domestic smelting capacity:

$$R_{\mathit{it}}^{\mathit{Post}}=P_t^{\mathit{Processed}}\cdot Q_{\mathit{smelted}}\left(K_{\mathit{it}}^{\mathit{smelter}}\right)-{\mathsf{\Phi }}_{\mathit{mandate}}$$

(5)

where Φ_mandate represents the fixed compliance and administrative cost imposed by sovereign enforcement, and constructing smelting capital requires large upfront investment outlays (I_it) subject to installation frictions.

Solving the firm’s dynamic optimization problem yields the first-order condition with respect to investment:

$${\displaystyle \frac{\mathsf{\partial }{\mathsf{\Pi }}_{\mathit{it}}}{\mathsf{\partial }{I}_{\mathit{it}}}}=-1-\psi \left({\displaystyle \frac{I_{\mathit{it}}}{K_{\mathit{it}}}}-c_0\right)+q_{\mathit{it}}=0⟹{\displaystyle \frac{I_{\mathit{it}}}{K_{\mathit{it}}}}=c_0+{\displaystyle \frac{1}{\psi }}\left(q_{\mathit{it}}-1\right)$$

(6)

where q_it is the marginal shadow value of capital (marginal q), defined as the co-state variable on the capital accumulation constraint:

$$q_{\mathit{it}}={\displaystyle \frac{\mathsf{\partial }{V}_{\mathit{it}}\left(K_{\mathit{it}}\right)}{\mathsf{\partial }{K}_{\mathit{it}}}}$$

(7)

Under the assumptions of linearly homogeneous production and adjustment cost technologies, Hayashi (1982) demonstrates that marginal q equals average q (Tobin’s Q, Q_it ≡ V_it/K_it). Taking the total differential of the valuation function V_it with respect to the downstreaming mandate, the differential valuation response of a policy-exposed firm relative to an unconstrained counterpart can be expressed as

$$\Delta {\ln V}_{\mathit{it}}=\underset{\mathrm{Value}-\mathrm{Added\; Rent\; Capture}(+)}{\underbrace{{\displaystyle {\sum }_{s=0}^{\infty }}{\beta }^s{\displaystyle \frac{\mathsf{\partial }{R}_{i,t+s}}{\mathsf{\partial }\mathrm{Mandate}}}}}-\underset{\mathrm{Sunk\; Capital\; Expenditure\; and\; Sovereign\; Risk}(-)}{\underbrace{{\displaystyle {\sum }_{s=0}^{\infty }}{\beta }^s\left({\displaystyle \frac{\mathsf{\partial }{I}_{i,t+s}}{\mathsf{\partial }\mathrm{Mandate}}}+{\displaystyle \frac{\mathsf{\partial }{C}_{i,t+s}}{\mathsf{\partial }\mathrm{Mandate}}}+{\displaystyle \frac{\mathsf{\partial }{\mathsf{\Phi }}_{\mathit{mandate}}}{\mathsf{\partial }\mathrm{Mandate}}}\right)}}$$

(8)

The net sign of ΔlnV_it is theoretically ambiguous:

• If capital markets perceive that captive downstream smelting generates sustainable operational rents and insulates the firm from trade restrictions, the first term dominates, driving a positive valuation response (β₁ > 0).
• If investors view the required capital deployment as value-destroying free cash flow expropriation—compounded by smelter construction delays, cost overruns, and sovereign hold-up risk—the second term dominates, lowering relative firm valuation (β₁ < 0).

3.2.2. Baseline Difference-in-Differences Specification

To empirically estimate this net effect, we implement a two-way fixed effects (TWFE) Difference-in-Differences estimator. Let i ∈ {1,…,18} index listed mining firms and t ∈ {1,…,37} index observation quarters from 2017:Q1 through 2026:Q1. The baseline structural econometric model is specified as

Y_it = β₀ + β₁(Treat_i × Post_t) + X_it′γ + α_i + δ_t + ε_it

(9)

where Y_it is the primary dependent variable of interest, representing firm valuation (Tobin’s ${\mathrm{Q}}_{\mathit{it}}$), log physical investment (lnCapex_it), investment intensity (Capex_it/TA_it), log market value of equity (lnMVE_it), or quarterly stock return (Return_it). Treat_i is a time-invariant binary indicator equal to 1 if firm i belongs to the treated cohort and 0 if it belongs to the counterfactual control group. The model is estimated across the three operational definitions: ${\mathit{Treat}}_i^{\mathit{pure}}$, ${\mathit{Treat}}_i^{\mathit{exposed}}$, and ${\mathit{Treat}}_i^{\mathit{mineral}}$. Post_t is a time-series indicator equal to 1 for all quarters subsequent to the enforcement of the export ban (t ≥ 2020:Q1) and 0 for pre-ban baseline quarters (t ≤ 2019:Q4). The interaction term Treat_i × Post_t represents the Difference-in-Differences regressor, designated in our empirical estimations as DID pure, DID exposed, and DID mineral. X_it is a vector of time-varying firm-level covariates, centered primarily on firm scale, proxied by the natural logarithm of total assets (lnTA_it). Controlling for firm size isolates valuation shifts from pure balance sheet expansion driven by smelter asset consolidation. α_i denotes firm fixed effects, absorbing all unobserved, time-invariant firm heterogeneity (such as geological asset quality, historical corporate governance structures, concession tenure types, and baseline operational efficiency). Note that because α_i fully subsumes time-invariant cross-sectional characteristics, the constitutive dummy Treat_i is absorbed. δ_t denotes quarter fixed effects, capturing all macroeconomic, political, and systemic capital market shocks common to all IDX mining firms in quarter t (such as global financial cycles, shifts in sovereign debt ratings, COVID-19 macroeconomic contractions, and domestic general elections). The quarter fixed effects similarly subsume the constitutive time dummy Post_t. Finally, ε_it is the idiosyncratic error term.

The coefficient β₁ is the parameter of interest, identifying the average treatment effect on the treated (ATT):

β₁ = 𝔼[Y_it(1) − Y_it(0)∣Treat_i = 1, Post_t = 1]

(10)

A statistically significant negative estimate (${\widehat{\beta }}_1<0$) indicates that the downstreaming mandate reduced the valuation or financial performance of exposed mining firms relative to unconstrained extractive peers. Conversely, ${\widehat{\beta }}_1>0$ indicates that equity markets priced the policy as net value enhancing over the medium-to-long term.

Because the January 2020 regulatory enforcement happened at the same time as the global start of the COVID-19 pandemic, it is very important to keep the empirical strategy away from big economic noise. Using financial data would mix these two events together because the export ban and the first COVID-19 lockdown would be seen as one 2020 observation. By using a quarterly panel, the empirical strategy can see fine time differences and separate the regulatory enforcement from the crisis. The quarterly data let the model see when the ban started in early Q1 2020 while the quarterly time fixed-effects take care of the very specific non-linear path of the COVID-19 pandemic—such as the sharp economic drop in Q2 2020 and the uneven recovery in later quarters. The model also adds macro-economic controls, such as regional GDP changes the IDR/USD exchange rate and global LME nickel spot prices. Because the big demand shocks of the COVID-19 pandemic hit both the nickel firms that were affected and the coal and non-nickel control firms that were not the Difference-in-Differences estimator removes the macro-economic noise. This quarterly design isolates the capital change caused by the ban so the analysis focuses on the policy change rather than temporary crisis moves.

3.2.3. Identifying Assumptions and Dynamic Event-Study Specification

The causal interpretability of ${\widehat{\beta }}_1$ under the TWFE framework rests on two standard identifying assumptions:

Assumption 1. 

Strict Exogeneity and Non-Anticipation.

Conditional on firm fixed effects, time fixed effects, and time-varying covariates, the error term has conditional mean zero:

𝔼[ε_it∣Treat_i, Post_t, X_it, α_i, δ_t] = 0

(11)

Furthermore, we require that firms did not strategically reallocate capital or alter their market valuations in anticipation of the ban prior to its official gazetting. While Law No. 4/2009 outlined long-term processing ambitions, the implementation date was brought forward to 1 January 2020 through ESDM Regulation No. 11/2019, promulgated late in the third quarter of 2019. This abrupt acceleration limited the scope for advance balance-sheet restructuring.

Assumption 2. 

Parallel Trends Counterfactual.

In the absence of the raw export prohibition, the average path of the outcome variable for the treated mining companies would have run parallel to that of the control cohort:

𝔼[Y_it(0) − Y_i,t−1(0)∣Treat_i = 1] = 𝔼[Y_it(0) − Y_i,t−1(0)∣Treat_i = 0] ∀t

(12)

Because the counterfactual trajectory 𝔼[Y_it(0)∣Treat_i = 1, Post_t = 1] is unobservable post-2020, we empirically evaluate this assumption by estimating a dynamic event-study specification:

$$Y_{\mathit{it}}={\beta }_0+{\displaystyle {\sum }_{\begin{array}{c}\tau =-12\\ \tau \ne -1\end{array}}^{24}}{\beta }_{\tau }\left({\mathit{Treat}}_i\times 1_{\left\{t=\tau \right\}}\right)+{\mathrm{X}}_{\mathit{it}}\prime \mathit{\gamma }+{\alpha }_i+{\delta }_t+{\epsilon }_{\mathit{it}}$$

(13)

where τ = t − T_event indexes event time in quarters relative to the policy enforcement date (T_event = 2020:Q1), spanning τ ∈ {−12,…,+24}. The quarter immediately preceding enforcement (τ = −1, 2019:Q4) is omitted as the reference baseline.

This leads to two explicit testable hypotheses:

• Test of Pre-Trends: For all τ < −1, we test the joint null hypothesis:

H₀:β₋₁₂ = β₋₁₁ = ⋯ = β₋₂ = 0

(14)

Failure to reject this joint null indicates that treatment and control firms followed statistically indistinguishable trends prior to the intervention, supporting the parallel-trends assumption.

b.

Dynamic Treatment Trajectories: For τ ≥ 0, the coefficients β_τ map the dynamic transmission of the policy shock over the subsequent 25 quarters. This enables us to distinguish between transitory adjustment costs (e.g., an immediate drop in Tobin’s Q during the initial capital-commitment phase) and subsequent valuation recovery as processing facilities come online.

3.2.4. Falsification Tests: Temporal and Spatial Placebos

To confirm that the baseline estimates are not artifacts of spurious correlation or unmodeled regime shifts, we formulate a diagnostic falsification exercise, namely, temporal placebo test. Particularly, we restrict the sample strictly to the pre-policy regime (t ≤ 2020:Q1) and introduce an artificial policy shock set mid-way through the pre-treatment window (for instance, setting PlaceboPost_t = 1 for quarters from 2018:Q1 onward, and 0 otherwise):

Y_it = β₀ + β_placebo(Treat_i × PlaceboPost_t) + X_it′γ + α_i + δ_t + ε_it ∀t ≤ 2020:Q1

(15)

Because no export ban occurred during this period, a valid identification design requires ${\widehat{\beta }}_{\mathit{placebo}}$ to be statistically indistinguishable from zero. Rejection of the null would signal underlying pre-existing divergence between cohorts prior to regulatory intervention.

3.2.5. Robust Inference with Few Clusters: The Wild Cluster Bootstrap

A central econometric challenge in this study concerns cluster-level inference. In firm-level panel studies, standard errors are conventionally clustered at the firm level to account for arbitrary within-unit serial correlation and heteroskedasticity:

$$\mathsf{\Omega }={\displaystyle {\sum }_{i=1}^G}{\mathrm{X}}_i\prime {\epsilon }_i{\epsilon }_i\prime {\mathrm{X}}_i$$

(16)

However, standard asymptotic properties of the cluster-robust variance estimator (CRVE; Liang & Zeger, 1986) rely on the number of clusters G approaching infinity (G → ∞). When the number of clusters is small—as is standard in public extractive markets, where our sample comprises G = 18 listed mining companies—standard cluster-robust standard errors can be severely downward-biased, leading to over-rejection of the null hypothesis and inflated Type I error rates (Cameron et al., 2008).

To obtain reliable inference without relying on large-cluster asymptotics, we compute p-values and confidence intervals using the wild cluster bootstrap (Cameron et al., 2008; MacKinnon & Webb, 2018). We estimate the restricted model under the null hypothesis H₀:β₁ = 0:

$$Y_{\mathit{it}}={\tilde{\beta }}_0+{\mathrm{X}}_{\mathit{it}}\prime \tilde{\gamma }+{\tilde{\alpha }}_i+{\tilde{\delta }}_t+{\tilde{\epsilon }}_{\mathit{it}}$$

(17)

yielding restricted residuals ${\tilde{\epsilon }}_{\mathit{it}}$. In each bootstrap iteration b ∈ {1,…,B} (setting B = 9999):

• For each firm cluster i, we draw an auxiliary scalar weight $v_i^{*\left(b\right)}$ from the two-point Rademacher distribution:

  $$v_i^{*\left(b\right)}\in \left\{-1,+1\right\}\hspace{1em}\mathrm{with\; equal\; probability}\mathrm{Pr}\left(v_i^{*\left(b\right)}=1\right)=0.5$$

  (18)

  or alternatively from the six-point Webb (2023) distribution, which is well suited for settings where G < 15.
• The pseudo-response vector is generated by imposing the null and scaling the cluster residuals:

$$Y_{\mathit{it}}^{*\left(b\right)}={\tilde{Y}}_{\mathit{it}}+v_i^{*\left(b\right)}\cdot {\tilde{\epsilon }}_{\mathit{it}}$$

(19)

3.

The full model is re-estimated on the pseudo-sample to obtain the bootstrap test statistic:

$$t^{*\left(b\right)}={\displaystyle \frac{{\widehat{\beta }}_1^{*\left(b\right)}}{\mathrm{se}\left({\widehat{\beta }}_1^{*\left(b\right)}\right)}}$$

(20)

The wild cluster bootstrap p-value is then computed as the empirical proportion of bootstrap test statistics that exceed the observed sample t-statistic in absolute value:

$$p_{\mathit{WCB}}={\displaystyle \frac{1}{B}}{\displaystyle {\sum }_{b=1}^B}{1}_{\left\{\left|t^{*\left(b\right)}\right|\ge \left|t_{\mathit{sample}}\right|\right\}}$$

(21)

Reporting wild cluster bootstrap p-values alongside conventional standard errors ensures that statistical significance assessments are robust to the small number of cross-sectional clusters.

4. Results and Discussion

4.1. Firm Valuation and Corporate Investment Response

Evaluating the market’s response to the mineral export prohibition requires disentangling immediate valuation shifts from structural changes in physical capital investment. The empirical analysis begins by estimating baseline Difference-in-Differences (DID) models for firm valuation, proxied by Tobin’s Q, alongside physical capital expenditure (Capex) across three classification tiers: all mineral firms, nickel-exposed conglomerates, and pure-play nickel extractors.

Table 4 reports the two-way fixed effects Difference-in-Differences (DID) estimates across our three exposure tiers: all metallic mineral producers, integrated holding firms and subsidiaries, and pure-play nickel extractors.

Table 4. Tobin’s Q vs. Capex models.

Panel A: Tobin’s Q Models: Witdout Firm Size (Baseline) vs. Witd Firm Size (Total Assets)
	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES	Baseline DID Mineral	Baseline DID Exposed	Baseline DID Pure	Firm Size DID Mineral	Firm Size DID Exposed	Firm Size DID Pure
DID Mineral	−68.42			−14.31
	−71.24			−23.89
DID Exposed		39.17			18.97
		−39.43			−27.78
DID Pure			37.33			32.24
			−37.28			−39.36
Total Assets (ln)				−66.27	−66.66	−66.78
				−67.46	−67.88	−68
Constant	61.58 **	36.31 ***	38.01 ***	950.5	949.3	950.6
	−21.44	−4.709	−2.968	−925.7	−925.4	−926.5
Observations	628	628	628	628	628	628
R-squared	0.335	0.332	0.332	0.412	0.412	0.412
Panel B: Capex Models: Without and With Natural Log (ln)
VARIABLES	Capex DID Mineral	Capex DID Exposed	Capex DID Pure	Capex (ln) DID Mineral	Capex (ln) DID Exposed	Capex (ln) DID Pure
DID Mineral	5.475			1.14
	−5.101			−0.951
Total Assets (ln)	−0.712	−0.526	−0.505	0.839 ***	0.868 ***	0.881 ***
	−0.553	−0.413	−0.401	−0.0855	−0.0827	−0.083
DID Exposed		−2.612			−1.625 ***
		−2.537			−0.547
DID Pure			−2.389			−1.642 ***
			−2.309			−0.518
Constant	10.63	10.04	9.627	−2.760 **	−2.621 **	−2.857 **
	−6.476	−5.867	−5.624	−1.146	−1.163	−1.156
Observations	612	612	612	608	608	608
R-squared	0.393	0.385	0.384	0.738	0.739	0.737

Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

While specifications (4) through (6) condition on firm scale using the natural logarithm of total assets, specifications (1) through (3) in Panel A of Table 4 give the baseline value estimates without firm-level variables. The DID interaction coefficients do not reach conventional levels of statistical significance across all six parameters. In specification (4), the estimated coefficient on DID mineral is −14.31 (p > 0.10). Similarly, the estimated coefficients for nickel-exposed firms (β = 18.97, specification 5) and pure nickel operators (β = 32.24, specification 6) remain statistically indistinguishable from zero against their clustered standard errors.

This muted valuation response suggests equity markets did not systematically price the hilirization mandate as a net value-destroying expropriation, nor did they immediately reward the anticipated rents of downstream integration.

A contrasting empirical pattern emerges when evaluating corporate investment behavior in Panel B of Table 4. When physical investment intensity is operationalized through the ratio of capital expenditures to total assets (specifications 1–3), the DID coefficients remain statistically insignificant. However, because capital intensity in the extractive sector is notoriously lumpy and scaled ratios can be distorted by sudden denominator expansions, specifications (4) through (6) model physical investment using the natural logarithm of capital expenditure (lnCapex).

The logarithmic specifications reveal an immediate, highly significant contraction in physical capital deployment among policy-exposed firms. Both nickel-exposed conglomerates (β = −1.625, p < 0.01) and pure nickel producers (β = −1.642, p < 0.01) experienced severe reductions in their physical investment trajectories relative to the control group. Rather than immediately mobilizing physical outlays into downstream smelter construction as mandated, upstream extractors curtailed capital deployment by drastically scaling back their traditional capital expenditure compared to unaffected peers.

We can interpret this as acute capital rationing; corporate balance sheets effectively froze conventional upstream investments while bracing for the massive downstream capital barriers required by the government mandate.

4.2. Secondary Market Pricing: Stock Returns and Equity Value

To examine whether secondary equity markets captured dimensions of shareholder wealth distinct from replacement-cost ratios, Table 5 evaluates quarterly stock returns alongside the market value of equity

Table 5. Market indicator models (stock return vs. market value of equity).

	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES	Stock Return DID Mineral	Stock Return DID Exposed	Stock Return DID Pure	MVE (ln) DID Mineral	MVE (ln) DID Exposed	MVE (ln) DID Pure
DID Mineral	−0.0321			0.162
	−0.0416			−0.434
DID Exposed		0.00805			0.434
		−0.0531			−0.547
DID Pure			0.0731 ***			0.957 **
			−0.0226			−0.452
Total Assets (ln)	−0.000185	−0.00132	−0.00131	0.176 **	0.186 **	0.184 **
	−0.00854	−0.00827	−0.00841	−0.0741	−0.0755	−0.0734
Constant	0.0738	0.0785	0.0732	10.07 ***	9.932 ***	9.939 ***
	−0.113	−0.114	−0.115	−1.014	−1.036	−1.001
Observations	613	613	613	628	628	628
R-squared	0.182	0.182	0.182	0.911	0.912	0.916

Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

In Table 5, specifications (1) through (3) report quarterly stock returns, while specifications (4) through (6) report the natural logarithm of market capitalization (lnMVE). For the broad mineral cohort (DID mineral) and the conglomerate-exposed cohort (DID exposed), the policy interaction generates no discernible secondary market repricing.

A pronounced divergence occurs, however, for pure-play nickel extractors. Specification (3) shows that pure nickel firms experienced an average quarterly stock return premium of 7.31 percentage points (β = 0.0731, p < 0.01) following the export prohibition. In specification (6), this market reassessment is mirrored in the absolute market value of equity: pure nickel entities realized a 95.7% expansion in equity market capitalization (β = 0.957, p < 0.05) relative to control firms.

Superficially, these parametric estimates seem to suggest that equity markets priced raw mineral bans as a net benefit for concession holders. With the Indonesian state restricting ore outflows, pure-play concessionaires held the feedstocks required by downstream smelters. Investors appeared to assign an initial scarcity premium to captive ore deposits, betting that domestic processors would compete for feedstocks. Whether these estimates represent genuine causal trajectories, however, depends on the parallel trends assumption and the asymptotic validity of the clustered standard errors.

4.3. Dynamic Event-Study Trajectories and Parallel Trends

The identifying integrity of a Difference-in-Differences design requires that, absent the regulatory intervention, treated and control cohorts would have followed comparable paths. Table 6 provides the quarterly event-study estimates spanning eight quarters prior to the shock (τ ∈ {−8,…,−2}) through eight quarters post-enforcement (τ ∈ {0,…,+8}), setting the quarter immediately prior to the ban (2019:Q4, τ = −1) as the reference baseline.

Table 6. Event study.

	(1)	(2)	(3)	(4)
VARIABLES	Tobin’s Q	Capex (ln)	Stock Return	MVE (ln)
treat_event_m8	126.9	0.283	0.00975	−0.574
	(124.2)	(1.088)	(0.0924)	(0.466)
treat_event_m7	4.858	−0.199	0.189 *	−0.776 *
	(16.08)	(1.246)	(0.106)	(0.414)
treat_event_m6	12.78	−0.144	0.273	−0.842 *
	(18.96)	(1.097)	(0.187)	(0.455)
treat_event_m5	65.30	−0.178	0.337 ***	−0.844 *
	(45.52)	(0.997)	(0.102)	(0.444)
treat_event_m4	12.44	0.859	0.283 **	−0.404
	(12.56)	(1.262)	(0.127)	(0.341)
treat_event_m3	30.14	−0.354	0.299 **	−0.332
	(27.27)	(1.059)	(0.109)	(0.283)
treat_event_m2	0.867	−0.710	0.311 ***	−0.0252
	(13.59)	(1.087)	(0.0796)	(0.233)
treat_event_0	9.206	0.798	0.141	0.0638
	(23.89)	(1.640)	(0.123)	(0.157)
treat_event_1	33.39	0.905	0.327 ***	0.191
	(31.76)	(0.935)	(0.0928)	(0.151)
treat_event_2	28.16	0.937	0.175 *	0.215
	(28.75)	(0.877)	(0.0863)	(0.141)
treat_event_3	−6.653	1.233	0.267	0.169
	(10.95)	(0.866)	(0.264)	(0.175)
treat_event_4	8.222	1.231	0.416	0.304
	(10.31)	(1.822)	(0.290)	(0.227)
treat_event_5	6.561	1.012	0.160	0.326
	(11.55)	(1.662)	(0.101)	(0.225)
treat_event_6	−13.59	1.004	−0.293 *	−0.0801
	(9.600)	(1.665)	(0.153)	(0.259)
treat_event_7	−28.98	2.633 **	0.142	−0.0890
	(19.97)	(1.084)	(0.185)	(0.339)
treat_event_8	70.19	1.168 *	0.104 *	−0.567
	(64.89)	(0.581)	(0.0591)	(0.448)
Total Asset (ln)	−68.84	0.844 ***	4.15 × 10−5	0.204 **
	(68.56)	(0.0964)	(0.00804)	(0.0758)
Constant	958.3	−2.850 *	0.00513	9.907 ***
	(920.3)	(1.380)	(0.112)	(0.981)
Observations	628	608	613	628
R-squared	0.419	0.740	0.209	0.917

Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

For Tobin’s Q (column 1), the pre-treatment interactions are statistically indistinguishable from zero across all lead quarters. The coefficient point estimates fluctuate without statistical significance, confirming parallel pre-trends between treatment and control groups before the ban. Post-policy coefficients similarly display no significant movement, corroborating the static results from Table 4.

For log capital expenditure (column 2), the pre-treatment estimates are tightly clustered around zero, ranging from −0.144 to 0.859, with none approaching statistical significance. This stability confirms that physical investment trajectories were running parallel prior to the regulatory shock. The investment contraction observed in the static DID models emerges with a lag in the post-treatment horizon, reflecting the delayed timeline required to alter capital programs.

For secondary equity market indicators, the event study reveals evidence of anticipation. Stock returns (column 3) exhibit positive and statistically significant pre-treatment coefficients in five out of seven lead quarters (treat_event_m7, treat_event_m5, treat_event_m4, treat_event_m3, and treat_event_m2). Similarly, the market value of equity (column 4) shows significant negative pre-treatment point estimates during event quarters −7 through −5.

These pre-treatment divergences indicate that secondary equity markets were already reacting to regulatory announcements during late 2018 and 2019, well before the formal January 2020 export deadline. Because market participants began adjusting positions as regulatory signals emerged, the static DID estimators for stock return and market value of equity pick up adjustments that began before the official policy date.

Because the January 2020 rules enforcement happened exactly when the COVID-19 pandemic started keeping the study away from big economic changes is very important. The model does this by using quarterly time fixed effects and also adding in economic factors that change over time. These include GDP changes, the IDR/USD exchange rate and global LME nickel prices. These factors help take in the changes that happen in the economy. Since the lockdowns from the pandemic stopped factories from working in both the nickel companies that were affected and the ones that were not, the DID method is able to take out the effects of the pandemic. This helps show the effect of the export ban.

4.4. Falsification Diagnostics: Placebo Policy Shocks

To verify that the baseline findings are not mechanical artifacts of cyclical mining shifts, Table 7 conducts a temporal falsification exercise. The estimation sample is restricted to the pre-policy window (t < 2020:Q1), and an artificial policy shock is imposed at 2018:Q1 (placebo post).

Table 7. Placebo test.

Panel A: Tobin’s Q vs. Capex (ln)
	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES	Tobin’s Q DID Mineral	Tobin’s Q DID Exposed	Tobin’s Q DID Pure	Capex (ln) DID Mineral	Capex (ln) DID Exposed	Capex (ln) DID Pure
Placebo DID Mineral	58.59			0.139
	−73.78			−0.146
Total Asset (ln)	−475.5 ***	−470.6 ***	−469.2 ***	0.621	0.565	0.653
	−34.05	−29.41	−30.74	−0.691	−0.669	−0.676
Placebo DID Exposed		83.91			0.325
		−62.76			−0.256
Placebo DID Pure			9.323 *			0.0593
			−5.099			−0.115
Constant	6342 ***	6282 ***	6274 ***	0.176	0.925	−0.223
	−439.4	−389.5	−407	−9.241	−8.946	−9.061
Observations	194	194	194	188	188	188
R-squared	0.908	0.908	0.906	0.834	0.834	0.834
Panel B: Stock Return vs. Market Value of Equation (ln)
VARIABLES	Stock Return DID Mineral	Stock Return DID Exposed	Stock Return DID Pure	MVE (ln) DID Mineral	MVE (ln) DID Exposed	MVE (ln) DID Pure
Placebo DID Mineral	0.0282			0.153
	−0.061			−0.165
Total Asset (ln)	0.0128	0.0182	0.0155	0.413 ***	0.429 ***	0.429 ***
	−0.019	−0.0123	−0.0173	−0.0491	−0.0449	−0.0449
Placebo DID Exposed		−0.286			−0.00685
		−0.248			−0.124
Placebo DID Pure			0.0115			−0.0761
			−0.0265			−0.164
Constant	−0.158	−0.186	−0.187	6.701 ***	6.525 ***	6.530 ***
	−0.244	−0.175	−0.228	−0.628	−0.595	−0.594
Observations	179	179	179	194	194	194
R-squared	0.138	0.192	0.137	0.96	0.96	0.96

Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

In Panel A of Table 7, specifications (1) through (3) evaluate the artificial intervention on Tobin’s Q, while specifications (4) through (6) test log capital expenditure. The placebo interaction coefficients for capital expenditure are small and statistically insignificant across all three sample variants (e.g., β = 0.0593, p > 0.10 for pure nickel firms). This confirms that physical investment trajectories were structurally stable prior to 2020.

For Tobin’s Q, the mineral and exposed placebo interactions show no statistical significance, although pure nickel firms exhibit a weakly significant coefficient at the 10% level (β = 9.323, p = 0.076). Panel B demonstrates that placebo shocks for quarterly stock returns and the log market value of equity generate no statistically significant effects across any operational tier. Across eleven of the twelve placebo specifications, the interaction terms remain statistically indistinguishable from zero, supporting the identification assumption that the structural shifts identified in Table 4 and Table 5 stem from the 2020 downstreaming policy rather than baseline trends.

To validate the parallel trends assumption, we have simulated a series of temporal placebo tests by artificially backdating the policy implementation. While the majority of the pre-treatment placebo coefficients return null results, the pure-nickel valuation placebo reaches significance at the 10% level. The temporal placebo tests provide supportive, though not entirely conclusive, evidence for the identification strategy, reminding us to interpret the finite-sample estimates with appropriate caution and reinforcing the necessity of our event-study anticipation timeline.

4.5. Finite-Sample Clustered Inference: Wild Bootstrap Evaluation

In our balanced panel, observations are clustered across G = 18 firms. To address this small-cluster issue, Table 8 benchmarks the conventional clustered inference against empirical p-values calculated via the wild cluster bootstrap (MacKinnon & Webb, 2018), estimated with 9999 replications using the Webb six-point distribution.

Table 8. Wild bootstrap inference robustness check: comparing conventional regressions with firm size-to-bootstrap p-value.

	(1) Tobin’s Q	(2) Capex (ln)	(3) Stock Return	(4) MVE
Panel A: Mineral Firms
DID Mineral Coefficient	−14.31	1.14	−0.0321	0.162
Clustered SE	−23.89	−0.951	−0.0416	−0.434
p-value	>0.1	>0.1	>0.1	>0.1
Bootstrap p-value	0.4890	0.2753	0.4381	0.7496
Estimate Remark	Not statistically significant	Not statistically significant	Not statistically significant	Not statistically significant
Panel B: Exposed Firms
DID Exposed Coefficient	18.97	−1.625 ***	0.00805	0.434
Clustered SE	−27.78	−0.547	−0.0531	−0.547
p-value	>0.1	<0.01	>0.1	>0.1
Bootstrap p-value	0.5074	0.0647 **	0.8901	0.5869
Estimate Remark	Not statistically significant	Remains robust statistically significant	Not statistically significant	Not statistically significant
Panel C: Pure Firms
DID Pure Coefficient	32.24	−1.642 ***	0.0731 ***	0.957 **
Clustered SE	−39.36	−0.518	−0.0226	−0.452
p-value	>0.1	<0.01	<0.01	<0.05
Bootstrap p-value	0.4814	0.1698	0.1576	0.1702
Estimate Remark	Not statistically significant	Weak evidence	Weak evidence	Weak evidence

Source: Author’s Calculation. Standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

The bootstrap resampling results in Table 8 clarify the baseline findings. In Panel A (all metallic mineral firms), the wild bootstrap confirms the absence of policy effects across all four dependent variables. The empirical bootstrap p-values range from 0.2753 to 0.7496, corroborating the conventional test statistics.

The key adjustments occur in Panel B and Panel C. For pure nickel firms (Panel C), the conventional regressions indicated strong statistical significance for capital expenditure reductions (β = −1.642, p < 0.01), quarterly stock return premia (β = 0.0731, p < 0.01), and equity market expansions (β = 0.957, p < 0.05). Under the wild cluster bootstrap, however, all three effects lose significance at conventional thresholds. The empirical bootstrap p-value for stock returns rises to 0.1576, the p-value for market equity value increases to 0.1702, and the p-value for log capital expenditure rises to 0.1698.

Because the pure nickel cohort contains a small number of cross-sectional units (N₁ = 2 firms; 74 observations), conventional clustered standard errors substantially underestimate sample variability. What appeared to be a statistically significant equity surge for pure nickel miners was an artifact of small-cluster asymptotic distortion.

For the broader nickel-exposed group in Panel B (N₁ = 3 firms; 111 observations), the capital expenditure contraction remains statistically robust. The bootstrap p-value for log Capex is 0.0647, maintaining statistical significance at the 10% level. The capital expenditure freeze observed among integrated nickel holding groups survives small-cluster finite-sample penalization.

The results show that there is a need to balance numbers with real-world situations. Even though the data show that the drop in spending on projects is significant at the 10% level for some big companies, the actual amount lost is very large. Changing the logarithm number into a percentage compared to the average before 2019 shows that there is a change. For these companies, the number means that they are spending tens of millions each quarter on regular oil and gas projects. This number shows how much money is being taken away from projects to pay for new smelters as required by the government. It shows how the rules make older companies give up their projects to pay for new ones.

4.6. Synthesis and Policy Implications

These empirical patterns nuance the standard economic narrative surrounding contemporary resource nationalism. Mandatory mineral processing policies are frequently characterized as either destructive sovereign interventions that impair firm value or as rent-capturing structural reforms that benefit resource holders.

The evidence from the Indonesia Stock Exchange indicates that neither characterization fully captures the corporate outcome. Aggregate replacement-cost valuations (Tobin’s Q) remained largely unchanged, as the prospect of downstream industrial rents was counterbalanced by high capital expenditure requirements and sovereign risk. Furthermore, while unadjusted secondary market metrics suggested a value expansion for pure extractors, that premium fails to hold under small-cluster bootstrap corrections.

The primary, statistically robust consequence of the hilirisasi mandate within listed extractive corporations was an investment contraction: corporate balance sheets curtailed physical capital expenditures (β = −1.625, bootstrap p = 0.0647). Confronted with the capital intensity of smelter development, domestic listed miners held back upstream physical commitments, and downstream processing capacity was ultimately driven by international capital joint ventures and off-balance-sheet financing arrangements.

The results from the data need matching with what we already know from theory. Hypothesis 1 which said there would be an effect on company values is mostly not supported by the overall numbers. This goes against models that expect big drops in stock prices when governments step in (like the work by Pástor & Veronesi, 2012). Instead, the overall value of companies did not change much. That suggests investors quickly factored in the logic of the state: even though building smelters is risky in the short term the long-term profits from controlling nickel supply for EV batteries made it worth it. Hypothesis 2 on the other hand is backed by the data though the numbers are a bit shaky when one looks at small sample sizes. The findings show a drop in spending on physical assets upstream. The estimated impact means each affected company spent tens of millions less per quarter on upstream projects. That gives a sense of how much capital was pulled away to pay for the required smelter construction.

5. Conclusion and Policy Implications

5.1. Summary of Findings and Empirical Contribution

This study investigated the microeconomic transmission of resource nationalism by evaluating the corporate financial response to Indonesia’s 2020 raw nickel export prohibition. Utilizing a dynamic Difference-in-Differences framework on a balanced quarterly panel, the empirical results challenge the polarized narratives surrounding state-mandated downstreaming. Aggregate firm valuation (Tobin’s Q) remained statistically invariant to the regulatory shock, rejecting traditional expropriation models that predict immediate value destruction. While pure-play extractors briefly signaled a secondary market premium, this effect dissipated under rigorous finite-sample bootstrap corrections. The definitive legacy of the mandate centers on corporate capital allocation. The evidence suggests a reduction in capital expenditure among nickel-exposed firms, although the magnitude and statistical strength of this effect are sensitive to finite-sample inference. Confronted with the massive capital requirements of mandated smelter infrastructure, exposed firms engaged in acute capital rationing, fundamentally altering their upstream investment trajectories to accommodate the state’s industrialization agenda.

5.2. Policy Implications for Emerging Market Resource Governance

Policymakers who are engineering supply chain transformations should look closely at the experience because the Indonesian experience offers a clear fiscal blueprint. When policymakers enforce processing, domestic processing stops raw outflows and creates industrial agglomeration, but domestic processing also brings immediate liquidity crises for domestic extractive firms. If financial hurdle rates for infrastructure are higher than the internal balance-sheet capacities of domestic miners, the state might unintentionally push domestic miners out of the value chain. As domestic extractive firms cut back on capital expenditure because of the domestic processing mandate, the downstream sector will inevitably depend on foreign direct investment to fill the capital deficit. Therefore, governments who pursue downstreaming must pair export prohibitions with specific domestic capital market interventions, such, as state-backed syndicated project finance or phased compliance timelines so that local entities can compete in the forced industrialization.

5.3. Limitations and Future Research Directions

The empirical boundary conditions of this study focus on the structure of the Indonesia Stock Exchange. Using a panel of publicly listed companies automatically leaves out state-owned non-listed firms and privately owned foreign subsidiaries. The capital structures of these firms, their internal financing limits, and their political risk profiles probably react differently to sudden policy shocks, creating a big gap in the current literature. In addition, the small number of firms that were fully treated in the market creates finite-sample limits that need careful interpretation. Finally, industrial capital allocation follows multi-decade cycles. While this study captures the shock to structure and the period of capital rationing, it does not resolve the final profitability of the mandated smelter investments. Future research must follow the long-term payoff of these sunk costs as battery chemistry changes, especially to see whether captive infrastructure gives lasting economic returns when geopolitical trade realignments shift.