From Exposure to Action? Natural Disasters and the Environmental Proactivity of Chilean Micro-Enterprises

Abstract

As climate-driven disasters intensify globally, this study investigates how environmental volatility influences the pro-environmental initiatives of micro-entrepreneurs in Chile. While Chile possesses world-class seismic resilience, the 2020–2025 period marked a dramatic shift toward hydro-climatological extremes, including mega-fires and catastrophic flooding. Integrating construal level theory, protection motivation theory, and the concept of focusing events, this research examines the psychological and structural drivers of business adaptation. Results indicate that residing in disaster-prone regions is insufficient to trigger proactivity; instead, a stark distinction exists between abstract geographic proximity and the behavior triggered by personal exposure. Furthermore, mediation analysis provides mixed support for the role of business profit; while profit loss negatively mediated equipment efficiency and recycling, the magnitude was marginal. This coping gap suggests that resource-constrained actors favor low-cost survivalist tactics over systemic shifts due to depleted organizational slack. Ultimately, the study highlights that disasters are powerful but inefficient teachers; without addressing technical and financial barriers to mitigation, global supply chains remain fragile despite localized disaster experiences.

1. Introduction

In recent years, the intersection of climate change and disaster risk has forced a global re-evaluation of how small-scale economic actors adapt to environmental volatility (e.g., [1,2,3,4,5,6,7,8,9]). While large corporations often possess the slack resources necessary for sophisticated sustainability transitions, micro-entrepreneurs operate under significant constraints, where the decision to adopt environmentally friendly initiatives is often a matter of immediate survival rather than long-term strategy (e.g., [10,11]). This article explores this dynamic within the unique context of Chile, a nation currently navigating a profound transition in its multi-hazard profile.

Historically, Chile’s national risk landscape has been dominated by seismic events, leading to world-class earthquake resilience (e.g., [12]). However, the period between 2020 and 2025 has signaled a dramatic shift toward climatological and hydrological extremes. Driven by a 15-year mega-drought and the intensification of the El Niño-Southern Oscillation (ENSO), the country has faced a series of catastrophic events: urban-forest interface fires, hydrological extremes, and mass movement events (e.g., [13,14,15]).

Research into pro-environmental behavior indicates that addressing the psychological and structural drivers of action can foster change (e.g., [16]). However, existing evidence suggests that business entities globally, including those in developed economies, exhibit low levels of readiness, frequently delegating the burden of mitigation to the state (e.g., [11]). To understand why some micro-entrepreneurs adopt green initiatives while others remain reactive, this study employs three complementary theoretical lenses. First, construal level theory suggests that risk is often perceived through the lens of psychological distance (e.g., [17,18]). For many, regional vulnerability is a high-level, abstract construal. It is only when a disaster becomes a lived experience—a low-level, concrete construal—that the mental anchor required to prioritize a response is established (e.g., [19]). Second, protection motivation theory clarifies the internal decision-making process (e.g., [20,21]). Even when a threat is acknowledged, an entrepreneur’s coping appraisal often dictates the outcome. Finally, these disasters can be viewed as focusing events. These sudden, high-impact occurrences disrupt the status quo and create a window for rapid learning (e.g., [22,23]). However, the pressure of these events often favors visible, direct adaptations over the systemic, long-term structural shifts necessary for true climate resilience.

Despite extensive work on small-to-medium enterprise (SME) resilience (e.g., [24,25,26]), empirical evidence remains limited on whether direct disaster experience triggers pro-environmental action differently from merely operating in high-risk territories; this distinction matters because it separates abstract awareness from behaviorally consequential shocks. Evidence from the present study confirms that while exposure to natural disasters influences initiatives such as emission reductions, equipment efficiency, and recycling, the depth of this influence is notably uneven. Specifically, while recycling and immediate mitigation measures demonstrate significant increases following a disaster, energy savings and systemic emissions reduction continue to lag. Most notably, the percentage of micro-entrepreneurs engaging in comprehensive mitigation remains small, even among those in regions severely affected within the last five years. This suggests that residing in a high-risk zone does not inherently translate into environmental proactivity (e.g., [27]). Instead, there is a stark cognitive and behavioral divergence between the abstract awareness generated by geographic proximity and the concrete, low-level construal triggered by direct, personal exposure (e.g., [28]).

Altogether, these findings highlight that while disasters serve as powerful focusing events for learning, the resulting proactivity remains largely superficial and reactive. Such responses often fail to translate into the systemic, long-term structural shifts required for true climate resilience in emerging economies (e.g., [29]). Furthermore, this evidence offers a critical template for understanding the systemic resilience gap that plagues both emerging and developed economies. It illustrates that the shift toward environmental proactivity is hindered less by a lack of awareness and more by a universal survivalist mindset common among resource-constrained actors (e.g., [11,30]). Indeed, across the global landscape, micro-entrepreneurs and SMEs often lack the necessary organizational slack—surplus financial or human capital—to move beyond low-cost, visible adaptations (e.g., [31]). This leaves global supply chains inherently fragile, as the will to act generated by localized disaster experiences is consistently stifled by a structural inability to act effectively.

This article is organized as follows. Section 2 presents a summary of the main climatological events that occurred during the period 2020–2025. Section 3 presents a theoretical framework and research hypotheses. Section 4 presents the data, analyzes some descriptive statistics, and describes methodological aspects. Section 5 presents the statistical results, while Section 6 analyzes them from the perspective of the theoretical framework. Section 7 concludes by summarizing the main findings and proposing some policy implications and themes for future research.

2. Recent Natural Disasters in Chile: 2020–2025

Since 2020, Chile has undergone a significant transition in its multi-hazard profile. While seismic events historically dominated the national risk landscape, the 2020–2025 period has been characterized by climatological and hydrological extremes. This shift is primarily driven by a 15-year mega-drought and the intensification of the ENSO, leading to record-breaking wildfires, catastrophic flooding, and mass movement events (e.g., [14,15,23,32,33,34]). These events can be summarized as follows.

(i)

Urban-forest interface fires. The 2024 Valparaíso event: In February 2024, extreme high temperatures (exceeding 40 °C), coupled with Puelche winds (warm, dry downslope winds), created the conditions for a mega-fire. In cities like Viña del Mar and Quilpué, urban sprawl into pine and eucalyptus plantations created a continuous fuel bed into residential areas. This resulted in over 131 fatalities and the destruction of over 6000 homes.

(ii)

Hydrological extremes. The 2023 central-south floods: Two distinct atmospheric rivers hit the central-south zone in June and August 2023. These events featured an abnormally high 0 °C isotherm, causing warm rain to fall at high altitudes where snow typically accumulates. This resulted in instantaneous runoff and flash flooding of the Maule and Biobío watersheds, and a complete flooding of the Licantén urban center. Agricultural losses were estimated at USD 720 million.

(iii)

Forest fires and ecosystem degradation (2023–2025): Major outbreaks took place in February 2023 and early 2025. Their main drivers were long-term soil moisture deficit (mega-drought) and the prevalence of monoculture timber plantations. In 2023 alone, over 430,000 hectares were consumed. The 2025 season continued to threaten high-biodiversity areas in La Araucanía, impacting indigenous Mapuche communities.

(iv)

Mass movement events. The 2021 summer mudslides: A rare warm-core summer storm in January 2021 triggered mass movements (aluviones) in the Cajón del Maipo, Región Metropolitana. Debris flows destroyed infrastructure in San Alfonso and increased river turbidity to levels that forced the suspension of potable water for 6 million residents in Santiago.

The Chilean experience between 2020 and 2025 underscores a growing climate risk gap. While the country has world-class seismic resilience, its urban planning and water management systems must now rapidly adapt to the increased frequency of hydro-meteorological disasters (e.g., [23]).

3. Theoretical Frameworks

Construal level theory (CLT) posits that the way individuals perceive an event or a risk is dictated by its psychological distance from it across dimensions of time, space, and personal experience. When a risk is perceived as a distant possibility or a statistical likelihood associated with a geographic area, it is processed as a high-level construal—an abstract, theoretical concept that rarely triggers immediate behavioral change. However, once that risk becomes a lived experience, it shifts into a low-level construal, characterized by concrete, proximal, and urgent details (e.g., [31,35,36,37]). This cognitive transition is essential for bridging the gap between general awareness and localized action, as the shift from the abstract to the concrete provides the necessary mental anchor to prioritize a response.

Once a threat is perceived as concrete, protection motivation theory (PMT) explains the internal decision-making process that leads to action. This process involves two distinct appraisals: threat appraisal, where the individual evaluates the severity of the danger and their personal vulnerability to it; and coping appraisal, where an individual assesses his/her own ability to respond (self-efficacy) and the effectiveness of the proposed solution. Even when the perceived threat is high, a coping gap can occur if the individual feels they lack the resources or expertise to implement a complex solution. Consequently, people often gravitate toward actions that offer a high sense of self-efficacy—those that feel manageable and immediate—while avoiding more technically demanding or costly interventions (e.g., [20,21]).

These psychological processes are complemented by the concept of focusing events, which describes sudden, high-impact occurrences that serve as powerful catalysts for change (e.g., [22,23]). Unlike chronic or background risks that can be easily ignored or normalized over time, a focusing event commands total attention and forces a rapid re-prioritization of goals. These events disrupt the status quo and create a unique window for learning and adaptation. However, because the pressure of such events often demands a swift response, the resulting actions tend to favor visible and direct adaptations rather than complex, long-term structural shifts, as the immediate need to restore stability often outweighs the drive for systemic transformation (e.g., [38]).

Based on this theoretical framework, the first hypothesis is:

H₁: 

Direct personal exposure to natural disasters exerts a significantly stronger positive influence on the adoption of environmentally friendly initiatives among micro-entrepreneurs than the mere residential location in a disaster-prone region.

Disasters typically lead to direct physical damage, inventory loss, and supply chain disruptions, which immediately slash business profits. Because micro-entrepreneurs operate with virtually no organizational slack (i.e., surplus cash or human capital), this profit loss restricts their coping appraisal (e.g., [39]). Even if they are aware of the risk, they lack the capital to implement systemic environmentally friendly initiatives. Conversely, if an environmentally friendly initiative is perceived to restore or protect profits, the entrepreneur is more likely to adopt it. In this case, profit acts as a positive mediator that rationalizes green behavior.

From the perspective of CLT, business profit acts as a mental anchor. Pre-disaster high profit allows for high-level construal. The entrepreneur can think abstractly about climate change and long-term sustainability. In contrast, post-disaster low profit forces low-level construal. The entrepreneur’s focus narrows to immediate, concrete survival. Environmental initiatives are only considered if they solve a concrete, immediate problem. On the other hand, according to PMT, profit is the engine for coping appraisal. If the disaster destroys profit, the entrepreneur’s perceived self-efficacy plummets (e.g., [11]).

Based on this discussion, the second hypothesis is:

H₂: 

Direct personal exposure to natural disasters positively influences the adoption of reactive environmental initiatives, but this relationship is negatively mediated by business profit for systemic mitigation measures, such as emission reductions and structural efficiency.

The two hypotheses are illustrated in Figure 1.

A methodological caveat regarding the testing of H₂ warrants consideration. Because this study primary dataset lacks direct psychometric scales for latent variables—specifically psychological distance from CLT or perceived self-efficacy from PMT—monthly profit is utilized as an observable proxy for the material constraints inherent in coping appraisal. While this is an indirect measure, it provides a testable mechanism for capturing the resource channel through which financial stability dictates a firm’s capacity for complex, systemic adaptation. This approach aligns with contemporary research suggesting that for resource-scarce micro-enterprises, financial performance often serves as a quantifiable indicator of the objective feasibility of protective actions, effectively operationalizing the gap between the will and the way to mitigate risk (e.g., [39,40]).

This figure outlines the two primary hypotheses (H₁ and H₂) governing the relationship between natural disaster exposure and the adoption of environmental initiatives among micro-entrepreneurs. H₁: Experiential knowledge as a driver of proactivity. The upper portion of the diagram illustrates the psychological distance gap inherent in risk perception. While residing in a disaster-prone region exerts a baseline influence on behavior, the figure highlights that direct personal disaster exposure provides a significantly stronger catalyst for action. This concrete experience collapses the abstract nature of climate risk, leading primarily to reactive green initiatives such as recycling and basic energy efficiency measures. H₂: Business profit as a mediator of systemic action. The lower portion of the figure details the mediation mechanism (coping gap) that dictates the depth of environmental response. Direct personal exposure is shown to have a negative relationship with business profit. The destruction of organizational slack and financial liquidity following a disaster creates a significant barrier to action. Business profit is positively correlated with the adoption of systemic green initiatives. The indirect effect highlighted in the blue box represents the coping gap. Even when the will to act is present due to threat appraisal, the mediation of profit loss suggests that micro-entrepreneurs are often financially unable to move beyond low-cost reactive behaviors toward long-term, structural mitigation.

4. Resources and Methods

4.1. Data

This study is based on the 8th wave of the Micro-entrepreneurship Survey (EME), which is publicly available at http://www.economia.gob.cl/category/estudios-encuestas/encuestas-y-bases-de-datos (accessed on 15 December 2025). The EME has been jointly conducted biannually by the Chile Ministry of Economics, Development and Tourism and the National Institute of Statistics (INE) since 2013. To date, it is the main instrument for characterizing formal and informal micro-enterprises in the country, providing relevant data for the development and monitoring of public policies in this area. The EME is based on the National Employment Survey (ENE), is nationally and regionally representative, and analyses individuals who are self-employed or own a micro-enterprise with up to 10 employees. Fieldwork for the eighth edition of the EME was conducted in May–August 2025.

4.2. Descriptive Statistics

4.2.1. Micro-Entrepreneur and Business Characteristics

According to the EME 8, the typical respondent profile was predominantly male (59%) and relatively young, with 44% under the age of 45. In terms of human capital and motivation, 34% of micro-entrepreneurs had completed tertiary education, while 35% were classified as opportunity driven.

Financial independence was a hallmark of this group: 67% of respondents identified personal savings and their own resources as their most critical start-up funding sources, as compared with 2% of commercial bank credit and 8.8% of loans from family and friends. In turn, 90% of respondents stated that regular operational expenses, such as production inputs and salaries, were exclusively financed by business income in the past year. Regarding business maturity and structure, nearly half (46%) were formally registered at Chile Internal Revenue Service (SII), yet only 27% employed additional personnel. Furthermore, 29% of these businesses were considered established ventures (10 years of age or more), 24% operated on a national scope, and 28% were concentrated within the retail sector.

See

Table A1. Descriptive statistics of some study variables.

Variable	Obs.	Mean	Std. Dev.	Min	Max
Male	2,114,579	0.594	0.491	0	1
Married	2,114,579	0.315	0.465	0	1
Under 45 years of age	2,114,579	0.443	0.497	0	1
Graduate studies	2,114,579	0.344	0.475	0	1
Start-up financing: savings/own resources	1,998,189	0.668	0.471	0	1
Start-up financing: Bank credit	1,998,189	0.017	0.128	0	1
Start-up financing: Loan from family & friends	1,998,189	0.088	0.284	0	1
Expenses exclusively financed by business income	1,979,247	0.898	0.302	0	1
Family tradition driven	1,960,342	0.090	0.287	0	1
Necessity driven	1,960,342	0.503	0.500	0	1
Opportunity driven	1,960,342	0.354	0.478	0	1
Formal business	1,998,189	0.458	0.498	0	1
Municipal permit	1,994,224	0.257	0.437	0	1
Personnel	1,998,189	0.266	0.442	0	1
Established business	1,994,980	0.286	0.452	0	1
Nationally oriented	1,989,963	0.236	0.425	0	1
Retail	2,114,579	0.279	0.448	0	1

for further details and

Table A2. List of variables in alphabetical order.

Variable	Description	Type	Source
Capital wealth	Monetary value of capital goods—computers, equipment, tools, utensils, vehicles—recorded into thirds.	Categorical	EME8 h3
Disaster-prone region	=1 if region was subjected to the most natural disasters in 2020–2025: 5 to 9 (Valparaíso, O’Higgins, Maule, Biobío, Araucanía), 13 (Metropolitana) and 16 (Ñuble).	Binary	EME8 region
Education	=1 if they have tertiary education (i.e., technical higher education, university education, master’s degree, doctorate).	Binary	EME6 cine_eme; EME7 cine_eme_re
Efficiency	=1 if micro-entrepreneur bought more efficient equipment and/or reduced fuel consumption in the past year.	Binary	EME8 m1_1, m1_2
Energy/Emission	=1 if micro-entrepreneur implemented actions aimed at energy savings and/or polluting gas emission reduction in the past year.	Binary	EME8 m1_3, m1_4
Established business	=1 if business is more than 10 years old.	Binary	EME 8 b2
Environmental	=1 if micro-entrepreneurs carried out at least one of the following actions: energy saving, polluting gas emission reduction, equipment efficiency improvement, fuel consumption reduction, recycling in the past year.	Binary	EME8 m1_6
Formal business	=1 if they registered entrepreneurial activity at Chile Internal Revenue Service (SII).	Binary	EME 8 Informalidad
Male	=1 if male.	Binary	EME8 sexo
Married	=1 if married.	Binary	EME8 est_conyugal
Mitigation	=1 if micro-entrepreneur has implemented any action to prevent or reduce the impact of natural disasters on his/her business or self-employment activity.	Binary	EME8 m3
Municipal permit	=1 if they got a municipal patent or permit to operate business or self-employment activity.	Binary	EME8 e10
National scope	=1 if sales products or services to other Chile regions.	Binary	EME 8 k10
Natural disasters	=1 if during the past 5 years a micro-entrepreneur experienced one or more of the following events: drought/frost, flooding/landslide, forest fire, storm.	Binary	EME8 m2
Opportunity	=1 if motivation for business was a market opportunity.	Binary	EME8 motivacion
Personnel	=1 if they had workers employed or hired for at least one hour a week, including unpaid family members and working partners, during the previous month.	Binary	EME 8 f1
Profit	Monthly profit according to depuration methodology	Continuous	EME 8 ganancia_final
Recycling	=1 if micro-entrepreneurs recycled.	Binary	EME8 m1_5
Retail	=1 if economic sector corresponds to retail.	Binary	EME 8 c1_caenes_1d_red
Under 45	=1 if they are under 45 years of age.	Binary	EME 8 tramo_etario

for a description of the study variables.

4.2.2. Natural Disasters

Chile is divided into sixteen geographic regions. Figure 2 depicts the percentages of micro-entrepreneurs who experienced one or more natural disasters in the five years prior to the survey in each of these regions. The events considered are droughts/frosts, floods/landslides, forest fires, and storms. As can be seen, the five regions with the highest percentages of micro-entrepreneurs affected by at least one of these natural disasters were O’Higgins (42.5%), Maule (50.2%), Biobío (50.8%), Araucanía (38.7%), and Los Ríos (42.0%). Further down the list were Coquimbo (36.1%), Valparaíso (24.6%), Los Lagos (31.5%), Región Metropolitana (31.4%), and Ñuble (29.4%). These figures somehow reflect the characterization of Section 2.

Altogether, 32.1% of the survey participants (639,388) experienced at least one natural disaster. In this regard, among those affected, about 97% experienced one (85%) or two events (12%).

Table 1. Percentage of micro-entrepreneurs subject to specific natural disasters.

	Droughts/Frosts		Floods/Landslides		Forest Fires		Storms
Region	No	Yes	No	Yes	No	Yes	No	Yes
1. Tarapacá	46,480	315	45,182	1613	46,460	335	44,718	2077
%	99.33	0.67	96.55	3.45	99.28	0.72	95.56	4.44
2. Antofagasta	58,985	46	57,438	1593	58,784	247	57,594	1437
%	99.92	0.08	97.3	2.7	99.58	0.42	97.57	2.43
3. Atacama	34,460	704	33,973	1191	35,115	49	34,372	792
%	98	2	96.61	3.39	99.86	0.14	97.75	2.25
4. Coquimbo	66,000	17,854	80,824	3030	83,309	545	70,061	13,793
%	78.71	21.29	96.39	3.61	99.35	0.65	83.55	16.45
5. Valparaíso	179,133	8937	181,834	6236	176,672	11,398	159,440	28,630
%	95.25	4.75	96.68	3.32	93.94	6.06	84.78	15.22
6. O’Higgins	84,889	7207	81,593	10,503	90,811	1285	61,067	31,029
%	92.17	7.83	88.6	11.4	98.6	1.4	66.31	33.69
7. Maule	104,244	9446	106,543	7147	112,400	1290	65,901	47,789
%	91.69	8.31	93.71	6.29	98.87	1.13	57.97	42.03
8. Biobío	112,247	15,544	113,714	14,077	114,176	13,615	81,601	46,190
%	87.84	12.16	88.98	11.02	89.35	10.65	63.86	36.14
9. Araucanía	114,935	19,460	130,487	3908	124,718	9677	103,385	31,010
%	85.52	14.48	97.09	2.91	92.8	7.2	76.93	23.07
10. Los Lagos	95,505	7125	99,287	3343	102,115	515	74,860	27,770
%	93.06	6.94	96.74	3.26	99.5	0.5	72.94	27.06
11. Aysén	15,238	1155	16,180	213	16,225	168	14,315	2078
%	92.95	7.05	98.7	1.3	98.98	1.02	87.32	12.68
12. Magallanes	17,273	1463	18,736	0	18,736	0	17,292	1444
	92.19	7.81	100	0	100	0	92.29	7.71
13. Metropolitana	821,335	26,835	813,884	34,286	837,020	11,150	621,219	226,951
	96.84	3.16	95.96	4.04	98.69	1.31	73.24	26.76
14. Los Ríos	39,567	4883	43,318	1132	43,598	852	29,976	14,474
	89.01	10.99	97.45	2.55	98.08	1.92	67.44	32.56
15. Arica y Parinacota	30,874	1247	31,876	245	31,919	202	30,953	1168
	96.12	3.88	99.24	0.76	99.37	0.63	96.36	3.64
16. Ñuble	43,082	6351	46,564	2869	47,585	1848	40,816	8617
	87.15	12.85	94.2	5.8	96.26	3.74	82.57	17.43
Total	1,864,247	128,572	1,901,433	91,386	1,939,643	53,176	1,507,570	485,249
%	93.55	6.45	95.41	4.59	97.33	2.67	75.65	24.35

Source: own elaboration based on EME 8.

provides details on the type of natural disaster by region. For example, forest fires affected a higher percentage of survey participants in Valparaíso (6.1%), Biobío (10.7%), and Araucanía (7.2%). The bottom of the table shows that, nationally, the most common natural disasters were storms (485,249 people), followed by droughts/frosts (128,572 people), floods/landslides (91,386 people), and forest fires (53,176 people).

Survey participants were asked whether in the past year they had carried out one or more of the following initiatives: saving energy, reducing polluting gas emissions, buying more efficient equipment, reducing fuel consumption, and recycling. For simplicity, saving energy and reducing polluting gas emissions are grouped as Energy/Emissions, while buying more efficient equipment and reducing fuel consumption are grouped as Efficiency.

Table 2. Environmentally friendly initiatives versus experiencing natural disasters.

(a)		Natural Disasters
				No		Yes		Total
		No		818,605		312,216		1,130,821
Environmental		%		60.48		48.83		56.74
		Yes		534,826		327,172		861,998
		%		39.52		51.17		43.26
		Total		1,353,431		639,388		1,992,819
		%		100		100		100
(b)	Natural disasters					(c)	Natural disasters
		No	Yes	Total			No	Yes	Total
	No	1,163,689	525,289	1,688,978		No	1,237,217	571,718	1,808,935
Energy/Emissions	%	85.98	82.15	84.75	Efficiency	%	91.41	89.42	90.77
	Yes	189,742	114,099	303,841		Yes	116,214	67,670	183,884
	%	14.02	17.85	15.25		%	8.59	10.58	9.23
	Total	1,353,431	639,388	1,992,819		Total	1,353,431	639,388	1,992,819
	%	100	100	100		%	100	100	100
(d)	Natural disasters					(e)	Natural disasters
		No	Yes	Total			No	Yes	Total
	No	958,454	389,237	1,347,691		No	1,288,172	488,433	1,776,605
Recycling	%	70.82	60.88	67.63	Mitigation	%	95.3	76.48	89.26
	Yes	394,977	250,151	645,128		Yes	63,573	150,227	213,800
	%	29.18	39.12	32.37		%	4.7	23.52	10.74
	Total	1,353,431	639,388	1,992,819		Total	1,351,745	638,660	1,990,405
	%	100	100	100		%	100	100	100

Source: own elaboration based on EME 8. Environmental indicates whether a survey participant fulfilled one or more of the following environmental-friendly actions: energy saving, emission reduction, improved equipment efficiency, fuel consumption reduction, and/or recycling in the past year.

presents two-way panels where the columns are having experienced natural disasters or not. In Table 2a, Environmental indicates whether a survey participant fulfilled one or more of the above-mentioned environmentally friendly actions (i.e., energy saving, emissions reduction, improved equipment efficiency, fuel consumption reduction, recycling) in the past year. As shown, among those who had experienced a natural disaster in the past 5 years, 51.2% carried out these types of initiatives as compared with 39.5% of those who had not. This pattern held true for any of the categories considered under Environmental: Energy/Emissions, (17.9% versus 14.0%), Efficiency (10.6% versus 8.6%), and Recycling (39.1% versus 29.2%), as shown in Table 2b–d.

Table 2e refers to mitigation. Specifically, survey participants were asked whether they had implemented any measure to prevent or reduce the impact of natural disasters on his/her business or self-employment activity. The measures under consideration were switching products, conducting technological innovations, changing the months of business operation, purchasing insurance against disasters (frosts, droughts, floods, earthquakes), and performing waste management. As can be seen, only 10.7% of all survey participants (213,800) implemented any mitigation measure. In contrast, among those who had experienced natural disasters, such percentage reached 23.5%.

4.3. Methodology

Statistical analysis is based on logistic regressions (e.g., [41], chapter 3), treatment effects computed from propensity score matching (e.g., [42], chapter 11), and mediation analysis (e.g., [42], chapter 27).

4.3.1. Logistic Regression

A logistic regression model for a binary response variable can be expressed as

$$\ln \left({\displaystyle \frac{{\mathrm{p}}_{\mathrm{i}}}{1-{\mathrm{p}}_{\mathrm{i}}}}\right)=\mathsf{\alpha }+{{\mathbf{x}}_{\mathrm{i}}}^{\mathbf{\prime }}\mathsf{\beta }+{\mathsf{\epsilon }}_{\mathrm{i}}$$

(1)

where $\left({\displaystyle \frac{{\mathrm{p}}_{\mathrm{i}}}{1-{\mathrm{p}}_{\mathrm{i}}}}\right)$ is the odds ratio between the probability of success, ${\mathrm{p}}_{\mathrm{i}}$ (e.g., carrying out a environmentally friendly initiative in the past year) and the probability of failure, ($1-{\mathrm{p}}_{\mathrm{i}}$), for a given individual, $\mathsf{\alpha }$ and $\mathsf{\beta }$ are parameters, ${\mathbf{x}}_{\mathrm{i}}$ is a vector of covariates, and ${\mathsf{\epsilon }}_{\mathrm{i}}$ is a random error.

4.3.2. Average Treatment Effect

For a binary treatment (e.g., experiencing natural disasters), T (=0, 1), which is orthogonal to the outcome variable, Y, (e.g., an environmental-friendly initiative) conditional on a set of covariates, $\mathbf{X}$, the average treatment effect (ATE) under propensity score weighting, $\mathrm{p}\left(\mathbf{X}\right)=\mathrm{P}(\mathrm{T}=1|\mathbf{X})$, is given by

$$\mathsf{\tau }=\mathrm{E}[\mathrm{Y}\left(1\right)-\mathrm{Y}\left(0\right)]=\mathrm{E}\left({\displaystyle \frac{\mathrm{T}\mathrm{Y}}{\mathrm{p}\left(\mathbf{X}\right)}}-{\displaystyle \frac{\left(1-\mathrm{T}\right)\mathrm{Y}}{1-\mathrm{p}\left(\mathbf{X}\right)}}\right)$$

(2)

where $\mathrm{p}\left(\mathbf{X}\right)$ satisfies the overlap condition, 0 < $\mathrm{p}\left(\mathbf{X}\right)$ < 1.

4.3.3. Mediation

In the presence of a mediator M, the total effect of the treatment or ATE is given by

$$\mathsf{\tau }=\mathrm{E}[(\mathrm{Y}\left(1\right)-\mathrm{Y}\left(0\right)]=\mathrm{E}[\mathrm{Y}\left(1, {\mathrm{M}}_1\right)-\mathrm{Y}\left(0, {\mathrm{M}}_0\right)]$$

(3)

where M₁ and M₀ denote the levels of the mediator in the presence or absence of the treatment, respectively.

The effect of the treatment on the outcome, Y, through the mediator is called the natural indirect effect (NIE). This measures the effect of the treatment through changing the mediator if the treatment were set at T = 1:

$$\mathrm{N}\mathrm{I}\mathrm{E}\equiv \mathsf{\delta }=\mathrm{E}[\mathrm{Y}\left(1,{\mathrm{M}}_1\right)-\mathrm{Y}\left(1,{\mathrm{M}}_0\right)]$$

(4)

In turn, the direct effect of the treatment on the outcome is called the natural direct effect (NDE). This measures the effect of the treatment on the outcome if the mediator were set at the natural level value M₀:

$$\mathrm{N}\mathrm{D}\mathrm{E}\equiv \mathsf{\eta }=\mathrm{E}[\mathrm{Y}\left(1,{\mathrm{M}}_0\right)-\mathrm{Y}\left(0,{\mathrm{M}}_0\right)]$$

(5)

where Y(1, M₀) is the hypothetical outcome if the individual received treatment 1 but his/her mediator was set at its natural level M₀ without the treatment.

Under the composition assumption, Y(1, M₁) = Y(1) and Y(0, M₀) = Y(0). Hence,

$$\mathsf{\delta }=\mathrm{E}\left[\mathrm{Y}\left(1\right)-\mathrm{Y}\left(1,{\mathrm{M}}_0\right)\right],\hspace{1em}\mathsf{\eta }=\mathrm{E}[\mathrm{Y}\left(1,{\mathrm{M}}_0\right)-\mathrm{Y}\left(0\right)]$$

It holds that the sum of the indirect and direct natural effects equals the total effect. That is, τ = δ + η.

To illustrate, let us assume that

$${\mathrm{Y}}_{\mathrm{i}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{M}}_{\mathrm{i}}+{\mathsf{\beta }}_2{\mathrm{T}}_{\mathrm{i}}+{\mathsf{\beta }}_3{\mathrm{M}}_{\mathrm{i}}{\mathrm{T}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{i}}$$

(6)

$${\mathrm{M}}_{\mathrm{i}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{T}}_{\mathrm{i}}+{\mathsf{\omega }}_{\mathrm{i}}$$

(7)

After substituting M_i into Y_i and taking expectation, one obtains:

$$\begin{array}{cc}\mathrm{E}\left({\mathrm{Y}}_{\mathrm{i}}\right|{\mathrm{M}}_{\mathrm{i}},{\mathrm{T}}_{\mathrm{i}})& ={\mathsf{\beta }}_0+{\mathsf{\beta }}_1({\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{T}}_{\mathrm{i}})+{\mathsf{\beta }}_2{\mathrm{T}}_{\mathrm{i}}+{\mathsf{\beta }}_3{\mathrm{T}}_{\mathrm{i}}({\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{T}}_{\mathrm{i}})\\ & ={\mathsf{\beta }}_0+{\mathsf{\beta }}_2{\mathrm{T}}_{\mathrm{i}}+\left({{\mathsf{\beta }}_1+\mathsf{\beta }}_3{\mathrm{T}}_{\mathrm{i}}\right)({\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{T}}_{\mathrm{i}})\hfill \end{array}$$

Therefore,

$$\begin{array}{cc}\mathrm{E}\left[\mathrm{Y}\left(1,{\mathrm{M}}_1\right)\right]\hfill & ={\mathsf{\beta }}_0+{\mathsf{\beta }}_2\times 1+({{\mathsf{\beta }}_1+\mathsf{\beta }}_3\times 1)({\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1\times 1)\hfill \\ \mathrm{E}\left[\mathrm{Y}\left(1,{\mathrm{M}}_0\right)\right]\hfill & ={\mathsf{\beta }}_0+{\mathsf{\beta }}_2\times 1+({{\mathsf{\beta }}_1+\mathsf{\beta }}_3\times 1)({\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1\times 0)\hfill \end{array}$$

$$\mathsf{\delta }=\mathrm{E}\left[\mathrm{Y}\left(1,{\mathrm{M}}_1\right)-\mathrm{Y}\left(1,{\mathrm{M}}_0\right)\right]=\left({{\mathsf{\beta }}_1+\mathsf{\beta }}_3\right){\mathsf{\alpha }}_1$$

$$\mathrm{E}\left[\mathrm{Y}\left(0,{\mathrm{M}}_0\right)\right]={\mathsf{\beta }}_0+{\mathsf{\beta }}_2\times 0+\left({{\mathsf{\beta }}_1+\mathsf{\beta }}_3\times 0\right)\left({\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1\times 0\right)={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathsf{\alpha }}_0$$

$$\begin{gathered} \mathsf{\eta }=\mathrm{E}\left[\mathrm{Y}\left(1,{\mathrm{M}}_0\right)-\mathrm{Y}\left(0,{\mathrm{M}}_0\right)\right]={\mathsf{\beta }}_2+\left({{\mathsf{\beta }}_1+\mathsf{\beta }}_3\right)\left({\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1\right)-{\mathsf{\beta }}_1{\mathsf{\alpha }}_0 \\ ={\mathsf{\beta }}_1{\mathsf{\alpha }}_1+{\mathsf{\beta }}_2+{\mathsf{\beta }}_3\left({\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1\right) \end{gathered}$$

5. Results

This section focuses on the adoption of environmentally friendly and climate mitigation initiatives. The primary independent variables are direct personal exposure to natural disasters and residence within a high-risk geographic region. To isolate these effects, the model incorporates a suite of control variables categorized by both entrepreneur demographics—including gender, education, age, and wealth—and firm-level attributes, such as formal registration, personnel size, establishment tenure, economic sector, and national market scope. The selection of these covariates is consistent with established literature in the field (e.g., [43,44]). The econometric specifications detailed in Section 4.3 were estimated using Stata 19, incorporating EME sampling weights to ensure the representativeness of the findings.

5.1. Logistic Regressions of Sustainable Practices

Table 3. Logistic regressions: environmentally friendly initiatives.

	(1)	(2)	(3)	(4)	(5)
	Energy/Emission	Efficiency	Recycling	Environmental	Mitigation
Independent Variable	Odds Ratio	Odds Ratio	Odds Ratio	Odds Ratio	Odds Ratio
Natural disasters	1.498 ***	1.295 ***	1.654 ***	1.732 ***	6.981 ***
	(0.007)	(0.007)	(0.006)	(0.006)	(0.039)
Control variables
Male	0.868 ***	1.725 ***	0.662 ***	0.802 ***	1.003
	(0.004)	(0.011)	(0.002)	(0.003)	(0.006)
Married	1.061 ***	0.975 ***	1.219 ***	1.098 ***	1.137 ***
	(0.005)	(0.005)	(0.004)	(0.004)	(0.007)
Education	1.016 ***	0.891 ***	1.047 ***	0.980 ***	0.912 ***
	(0.005)	(0.006)	(0.004)	(0.004)	(0.006)
Under 45 years	0.942 ***	1.071 ***	0.955 ***	0.981 ***	1.155 ***
	(0.004)	(0.006)	(0.004)	(0.003)	(0.007)
Opportunity-driven	1.095 ***	1.091 ***	1.113 ***	1.101 ***	0.880 ***
	(0.005)	(0.006)	(0.004)	(0.004)	(0.005)
Middle 33rd pctl. capital wealth	1.073 ***	1.887 ***	0.998	1.080 ***	1.302 ***
	(0.006)	(0.016)	(0.004)	(0.004)	(0.009)
Upper 33rd pctl. capital wealth	1.077 ***	2.967 ***	0.977 ***	1.128 ***	1.870 ***
	(0.006)	(0.025)	(0.005)	(0.005)	(0.014)
Municipal permit	1.370 ***	1.256 ***	1.012 ***	1.231 ***	1.145 ***
	(0.007)	(0.008)	(0.004)	(0.005)	(0.007)
Formal business	1.288 ***	1.127 ***	0.918 ***	1.035 ***	1.103 ***
	(0.007)	(0.008)	(0.004)	(0.004)	(0.007)
Retail sector	0.750 ***	0.683 ***	0.959 ***	0.825 ***	1.274 ***
	(0.004)	(0.005)	(0.004)	(0.003)	(0.008)
Personnel	1.456 ***	1.537 ***	1.559 ***	1.551 ***	1.198 ***
	(0.007)	(0.009)	(0.006)	(0.006)	(0.007)
National scope	1.110 ***	1.219 ***	1.245 ***	1.259 ***	1.440 ***
	(0.005)	(0.007)	(0.005)	(0.005)	(0.009)
Established business	1.033 ***	0.888 ***	0.938 ***	0.868 ***	1.150 ***
	(0.005)	(0.006)	(0.004)	(0.003)	(0.007)
Constant	0.144 ***	0.023 ***	0.236 ***	0.443 ***	0.014 ***
	(0.002)	(0.000)	(0.003)	(0.005)	(0.000)
Region fixed effects	Yes	Yes	Yes	Yes	Yes
Observations	1,780,799	1,780,799	1,780,799	1,780,799	1,778,840
Pseudo R2	0.037	0.086	0.050	0.047	0.155
Marginal effect of natural disasters
	(1)	(2)	(3)	(4)	(5)
	Energy/Emission	Efficiency	Recycle	Environmental	Mitigation
Natural disasters	0.052 ***	0.019 ***	0.109 ***	0.136 ***	0.137 ***
	(5.8 × 10−4)	(4.0 × 10−4)	(7.8 × 10−4)	(8.6 × 10−4)	(3.8 × 10−4)

Robust standard errors in parentheses; *** p < 0.01; Note: (1) Environmental indicates whether a survey participant fulfilled one or more of the following environmental-friendly actions: energy saving, emission reduction, improved equipment efficiency, fuel consumption reduction, and/or recycling in the past year. (2) Marginal effects are evaluated at sample means. (3) For variable definitions, see Table A2.

, columns (1)–(5), presents logistic regressions for the variables analyzed in Table 2: energy/emissions, efficiency, recycling, environmental, and mitigation. Parameter estimates are expressed as odds ratios, so that the value below (above) one indicates an inverse (direct) association between the dependent variable and the covariate in question. As shown, natural disasters had a positive impact on each of these dependent variables. For example, column (1) shows that the odds of achieving energy savings/emissions reductions were 50% higher for someone who had experienced a natural disaster in the past 5 years. For all environmental initiatives, column (4) shows that the odds were 73% higher. Not surprisingly, the odds were particularly high in the case of mitigation. This reflects the fact reported in Table 2e: mitigation was very infrequent among micro-entrepreneurs who had not experienced natural disasters (4.7% versus 23.5% of those who had). In fact, the mitigation odds for those who experienced natural disasters, computed from Table 2e, are (23.5/76.5)/(4.7/95.3) = 6.24, a figure close to that reported in column (5) of Table 3, 6.98.

For completeness, the bottom of Table 3 reports marginal effects of natural disasters on each dependent variable, evaluated at sample means. Given that the dependent variables are binary, the marginal effect is interpreted as an increment in the likelihood that a particular dependent variable equals 1. For example, the likelihood of recycling increased by 10.9 percentage points for someone who had experienced natural disasters. In turn, the increments in the likelihood of energy/emissions, efficiency, environmental and mitigation were, respectively, 5.2, 1.9, 13.6, and 13.7 percentage points. In other words, it appears that energy saving/emissions reduction and improved equipment efficiency were not particularly sensitive to experiencing extreme climatological events.

Regarding control variables, the adoption of green initiatives and climate mitigation strategies is fundamentally shaped by a combination of organizational scale, entrepreneurial motivation, and socio-economic factors. The statistical results indicate that businesses with a national reach, workforce, and municipal permits are the most consistent in implementing sustainable practices, as these structural advantages provide the necessary framework for environmental accountability. However, a significant gap exists between intent and action based on the entrepreneur’s mindset; specifically, while opportunity-driven individuals frequently embrace green initiatives to capture market value, they often neglect the long-term defensive measures required for climate risk mitigation. Furthermore, while personal wealth serves as a vital resource for funding both green innovations and resilience efforts—except for recycling—gender remains a defining variable. Men are generally less inclined toward broad green-friendly behaviors, instead focusing their environmental efforts almost exclusively on operational efficiency, where there is a clear technical or financial return.

5.2. Average Treatment Effects of Sustainable Practices

In panel (a), the ATEs of natural disasters on environmental, energy/emissions, efficiency, and recycling do not differ substantially from the marginal effects reported in Table 3: 13.7, 6.2, 2.1, and 10.2 percentage points, respectively. However, the ATE on mitigation is much larger than the marginal effect from the logistic regression: 22.4 versus 13.7 percentage points. This discrepancy highlights the importance of covariate balancing; by using propensity scores to account for the structural and socio-economic differences between exposed and non-exposed micro-entrepreneurs, the model reveals that the true behavioral shift toward mitigation is much more profound than standard estimates suggest.

Following the criteria in Section 2, regions identified as disaster-prone are those that experienced the highest frequency of natural disasters between 2020 and 2025. These include regions 5 through 9 (Valparaíso, O’Higgins, Maule, Biobío, and Araucanía), as well as regions 13 (Metropolitana) and 16 (Ñuble). Approximately 78% of all survey participants and 84% of those who personally experienced a natural disaster resided in these areas. Consequently, the focus is a group that includes most disaster-affected micro-entrepreneurs (539,712) and a significant number of others (1,013,933) who may have faced indirect impacts, such as supply chain disruptions or reduced demand for goods and services. As illustrated in panel (b), ATEs decrease significantly compared to panel (a), except for recycling. For instance, the ATE for environmental initiatives falls from 13.7 to 5.3 percentage points, while the ATEs for efficiency and mitigation approach zero.

Panel (c) further details the ATEs for environmental initiatives categorized by disaster type: droughts/frosts, floods/landslides, forest fires, and storms. The most substantial effects are observed in response to droughts/frosts (10.7 percentage points) and storms (12.6 percentage points). According to Table 1, these categories represent the disasters that affected the largest portion of the population.

As a robustness check, Figure 3 displays kernel density plots to evaluate the balance of the estimated propensity scores between the treatment and control groups. Panels (a) and (b) illustrate the balancing for Any environmental initiative and Mitigation, respectively; these are presented separately to account for the slight variations in sample size (1,780,799 versus 1,778,840). The density plots for Energy/Emissions, Efficiency, and Recycling are omitted for brevity, as they are analogous to that of Any environmental initiative due to their identical treatment and estimation samples. Panel (c) in turn presents similar plots for any environmental initiative under the different treatments of droughts/frosts, floods/landslides, forest fires, and storms. Across all specifications, the figures demonstrate that common support is achieved, with the treated and control distributions aligning closely after matching. It is worth noting that the propensity score range reflects the baseline probability of treatment. For instance, the estimated likelihood of experiencing a storm for the matched units in Panel (c) is concentrated between approximately 10% and 45%.

Figure 3. Kernel density plots for propensity score balance. Plots are based on an Epanechnikov kernel function. (a) Plots corresponding to Table 4a; (b) Plots corresponding to Table 4b; (c) Plots corresponding to Table 4c.

Figure 3. Kernel density plots for propensity score balance. Plots are based on an Epanechnikov kernel function. (a) Plots corresponding to Table 4a; (b) Plots corresponding to Table 4b; (c) Plots corresponding to Table 4c.

Table 4. Average treatment effects: natural disasters and disaster-prone regions.

(a) Having experienced a natural disaster
(i) Any environmental initiative					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Natural disaster (Yes vs. No)	0.137	5.50 × 10−4	246.03	0.000	0.136	0.138
(ii) Energy/Emissions					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Natural disaster (Yes vs. No)	0.062	4.30 × 10−4	143.04	0.000	0.061	0.063
(iii) Efficiency					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Natural disaster (Yes vs. No)	0.021	3.30 × 10−4	65.07	0.000	0.021	0.022
(iv) Recycling					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Natural disaster (Yes vs. No)	0.102	5.40 × 10−4	189.88	0.000	0.101	0.103
(v) Mitigation					Nº obs. = 1,778,840
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Natural disaster (Yes vs. No)	0.224	3.90 × 10−4	580.32	0.000	0.223	0.224
(b) Living in a disaster-prone region
(i) Any environmental initiative					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Potential exposure (Yes vs. No)	0.053	7.30 × 10−4	72.82	0.000	0.052	0.055
(ii) Energy/Emissions					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Potential exposure (Yes vs. No)	0.019	5.20 × 10−4	37.09	0.000	0.018	0.020
(iii) Efficiency					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Potential exposure (Yes vs. No)	−0.004	4.50 × 10−4	−9.09	0.000	−0.005	−0.003
(iv) Recycling					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Potential exposure (Yes vs. No)	0.082	6.70 × 10−4	122.25	0.000	0.081	0.083
(v) Mitigation					Nº obs. = 1,778,840
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Potential exposure (Yes vs. No)	0.009	4.60 × 10−4	18.81	0.000	0.008	0.009
(c) Any type of environmental initiative by type of natural disaster experienced
(i) Any environmental initiative					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Droughts/Frosts (Yes vs. No)	0.107	8.60 × 10−4	123.76	0.000	0.105	0.108
(ii) Any environmental initiative					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Floods/Landslides (Yes vs. No)	−0.001	7.45 × 10−4	−1.05	0.296	−0.002	0.001
(iii) Any environmental initiative					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Forest fires (Yes vs. No)	0.055	6.9 × 10−4	80.40	0.000	0.054	0.057
(iv) Any environmental initiative					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Storms (Yes vs. No)	0.126	6.80 × 10−4	184.90	0.000	0.125	0.128

Notes: (1) Average treatment effects (ATEs) are computed based on propensity score matching. Covariates are those reported in Table 3. (2) In panel (b), regions classified as prone to natural disasters are those subjected to the most natural disasters in 2020–2025: 5 to 9 (Valparaíso, O’Higgins, Maule, Biobío, Araucanía), 13 (Metropolitana) and 16 (Ñuble). (3) In panel (c), Any environmental initiative indicates whether a survey participant fulfilled one or more of the following environmentally friendly actions: energy saving, emission reduction, improved equipment efficiency, fuel consumption reduction, and/or recycling in the past year. (4) For variable definitions, see Table A2.

Table 4. Average treatment effects: natural disasters and disaster-prone regions.

(a) Having experienced a natural disaster
(i) Any environmental initiative					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Natural disaster (Yes vs. No)	0.137	5.50 × 10−4	246.03	0.000	0.136	0.138
(ii) Energy/Emissions					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Natural disaster (Yes vs. No)	0.062	4.30 × 10−4	143.04	0.000	0.061	0.063
(iii) Efficiency					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Natural disaster (Yes vs. No)	0.021	3.30 × 10−4	65.07	0.000	0.021	0.022
(iv) Recycling					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Natural disaster (Yes vs. No)	0.102	5.40 × 10−4	189.88	0.000	0.101	0.103
(v) Mitigation					Nº obs. = 1,778,840
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Natural disaster (Yes vs. No)	0.224	3.90 × 10−4	580.32	0.000	0.223	0.224
(b) Living in a disaster-prone region
(i) Any environmental initiative					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Potential exposure (Yes vs. No)	0.053	7.30 × 10−4	72.82	0.000	0.052	0.055
(ii) Energy/Emissions					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Potential exposure (Yes vs. No)	0.019	5.20 × 10−4	37.09	0.000	0.018	0.020
(iii) Efficiency					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Potential exposure (Yes vs. No)	−0.004	4.50 × 10−4	−9.09	0.000	−0.005	−0.003
(iv) Recycling					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Potential exposure (Yes vs. No)	0.082	6.70 × 10−4	122.25	0.000	0.081	0.083
(v) Mitigation					Nº obs. = 1,778,840
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Potential exposure (Yes vs. No)	0.009	4.60 × 10−4	18.81	0.000	0.008	0.009
(c) Any type of environmental initiative by type of natural disaster experienced
(i) Any environmental initiative					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Droughts/Frosts (Yes vs. No)	0.107	8.60 × 10−4	123.76	0.000	0.105	0.108
(ii) Any environmental initiative					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Floods/Landslides (Yes vs. No)	−0.001	7.45 × 10−4	−1.05	0.296	−0.002	0.001
(iii) Any environmental initiative					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Forest fires (Yes vs. No)	0.055	6.9 × 10−4	80.40	0.000	0.054	0.057
(iv) Any environmental initiative					Nº obs. = 1,780,799
Average treatment effect	Coefficient	Robust s.e	z	P > |z|	95% conf. interval
Storms (Yes vs. No)	0.126	6.80 × 10−4	184.90	0.000	0.125	0.128

Notes: (1) Average treatment effects (ATEs) are computed based on propensity score matching. Covariates are those reported in Table 3. (2) In panel (b), regions classified as prone to natural disasters are those subjected to the most natural disasters in 2020–2025: 5 to 9 (Valparaíso, O’Higgins, Maule, Biobío, Araucanía), 13 (Metropolitana) and 16 (Ñuble). (3) In panel (c), Any environmental initiative indicates whether a survey participant fulfilled one or more of the following environmentally friendly actions: energy saving, emission reduction, improved equipment efficiency, fuel consumption reduction, and/or recycling in the past year. (4) For variable definitions, see Table A2.

For completeness,

Table A3. Covariate balance summary for Table 4a.

(a) Any Environmental Initiative
	Raw		Matched
Number of obs.	1,780,799		3,561,598
Treated obs.	564,076		1,780,799
Control obs.	1,216,723		1,780,799
	Standardized differences		Variance ratio
	Raw	Matched	Raw	Matched
Male	0.160	0.025	0.922	0.989
Married	−0.055	0.020	0.959	1.015
Education	−0.147	−0.100	0.915	0.935
Under 45	−0.019	−0.081	0.996	0.972
Opportunity driven	−0.141	0.029	0.918	1.016
Middle 33rd pctl. capital wealth	−0.035	−0.016	0.976	0.989
Upper 33rd pctl. capital wealth	0.114	0.008	1.080	1.006
Municipal permit	−0.021	−0.027	0.978	0.970
Formal business	−0.148	−0.038	0.986	0.996
Retail sector	0.092	−0.017	1.095	0.981
Personnel	0.117	−0.023	1.113	0.976
National scope	0.010	−0.009	1.012	0.989
Established business	0.091	0.056	1.086	1.051
(b) Mitigation
	Raw		Matched
Number of obs.	1,778,840		3,557,680
Treated obs.	563,639		1,778,840
Control obs.	1,215,201		1,778,840
	Standardized differences		Variance ratio
	Raw	Matched	Raw	Matched
Male	0.161	0.025	0.922	0.989
Married	−0.055	0.045	0.959	1.030
Education	−0.145	−0.068	0.915	0.956
Under 45	−0.017	−0.088	0.996	0.969
Opportunity driven	−0.141	0.036	0.918	1.020
Middle 33rd pctl. capital wealth	−0.034	−0.026	0.977	0.983
Upper 33rd pctl. capital wealth	0.113	0.037	1.079	1.028
Municipal permit	−0.022	−0.045	0.977	0.951
Formal business	−0.149	−0.040	0.986	0.996
Retail sector	0.091	−0.012	1.095	0.987
Personnel	0.117	−0.010	1.112	0.990
National scope	0.011	−0.001	1.012	0.998
Established business	0.089	0.062	1.085	1.056

provides detailed covariate balance summaries for two representative cases. Following the matching procedure, all standardized mean differences (SMD) are reduced to values less than or equal to 0.1 in absolute magnitude. This threshold is widely accepted in empirical literature as an indicator of a well-specified model, suggesting that the matching process successfully eliminated observable systematic differences between the treatment and control groups. Consequently, the resulting estimates are not driven by imbalances in the underlying demographic or firm-level characteristics.

5.3. Business Profit as a Mediator

Mediation analysis involves estimating separate equations for the outcome and mediator variables. Three outcome variables are considered in this case: energy/emissions, efficiency and recycling. Each of these variables is modeled on the treatment (exposure to natural disasters), an interaction between the treatment and the mediator (monthly profit) and on the control variables previously considered (Table 3). Given their binary nature, each outcome is estimated using a probit model. The mediator, monthly profit, in turn is modeled on the treatment, entrepreneur’s wealth, municipal patent, business registration, economic sector, personnel, national scope, and geographic region through a linear regression. Considering the large sample size, the analysis prioritizes marginal effects (reported as percentage-point changes) to facilitate a direct comparison across initiative types. Statistical significance is interpreted with caution; instead, the relevance of the results is assessed by evaluating the magnitude of the effect sizes relative to the baseline prevalence of each initiative.

Table 5. Profits as a mediator of natural disasters.

(a)						Nº obs. = 1,780,799
Energy/Emissions		Coefficient	Robust s.e	z	P > |z|	95% conf. interval
NIE	Natural disaster (Yes vs. No)	3.41 × 10−4	5.74 × 10−4	5.94	0.000	2.3 × 10−4	4.5 × 10−4
NDE	Natural disaster (Yes vs. No)	0.055	6.45 × 10−4	85.10	0.000	0.054	0.056
ATE	Natural disaster (Yes vs. No)	0.055	6.42 × 10−4	85.99	0.000	0.054	0.057
(b)						Nº obs. = 1,780,799
Efficiency		Coefficient	Robust s.e	z	P > |z|	95% conf. interval
NIE	Natural disaster (Yes vs. No)	−0.001	5.09 × 10−4	−20.01	0.000	−1.1 × 10−3	−9.2 × 10−4
NDE	Natural disaster (Yes vs. No)	0.024	5.19 × 10−4	46.22	0.000	0.023	0.025
ATE	Natural disaster (Yes vs. No)	0.023	5.10 × 10−4	45.14	0.000	0.022	0.024
(c)						Nº obs. = 1,780,799
Recycling		Coefficient	Robust s.e	z	P > |z|	95% conf. interval
NIE	Natural disaster (Yes vs. No)	−0.001	6.66 × 10−4	−18.19	0.000	−1.3 × 10−3	−1.1 × 10−3
NDE	Natural disaster (Yes vs. No)	0.108	7.83 × 10−4	137.38	0.000	0.106	0.109
ATE	Natural disaster (Yes vs. No)	0.106	7.80 × 10−4	136.51	0.000	0.105	0.108

Notes: (1) The mediator is monthly profit. (2) The outcome equation includes an interaction between the treatment (natural disasters) and the mediator (monthly profit) and the control variables previously considered (Table 3). Given their binary nature, the outcome variables are fitted probit models. The mediator in turn is modeled on the treatment, entrepreneur’s wealth, municipal patent, business registration, economic sector, personnel, national scope, and geographic region through a linear regression. (3) NIE: natural indirect effect, NDE: natural direct effect, ATE: Average treatment effect. (4) For variable definitions, see Table A2.

presents statistical results of mediation analysis for the above-mentioned outcome variables. As can be seen, there is evidence of a negative mediation effect for efficiency (b) and recycling (c) but not for energy/emissions (a), demonstrating mixed support for H₂. Nevertheless, the magnitude of the negative mediating effects is small: −0.1 percentage points in each case. Therefore, the ATEs obtained do not substantially differ from those reported in Table 4a: 5.5 percentage points for energy/emissions (versus 6.2 percentage points), 2.3 percentage points for efficiency (versus 2.1 percentage points), and 10.6 percentage points (versus 10.2 percentage points). In interpreting these results, it is important to note that the profit measure used reflects the preceding year’s performance, which may not fully capture the acute, immediate financial impact of the disaster event.

6. Discussion

The results provide robust support for H₁, demonstrating that direct personal exposure is a far more potent catalyst for change than mere proximity to risk:

• The psychological distance gap: While 78% of participants lived in disaster-prone regions, the ATEs for these individuals dropped significantly compared to those with direct exposure. For example, the likelihood of adopting environmental initiatives dropped from 13.7 percentage points (direct exposure) to only 5.3 percentage points (geographic residence).
• The concrete anchor: This confirms the core tenet of CLT. Regional vulnerability remains a high-level, abstract construal. It is only when the disaster disrupts the micro-entrepreneur’s own inventory, supply chain, or physical assets that the risk shifts to a low-level concrete construal. This lived experience provides the mental anchor required to shift from passive awareness to active mitigation.

The study offers more nuanced, mixed support for H₂. While natural disasters increased the odds of all types of initiatives, there is a clear hierarchy in the sensitivity of these responses:

• Mitigation odds: The odds of engaging in mitigation were 6.98 times higher for those who experienced a disaster. However, the marginal effects show that energy savings and equipment efficiency—measures that often require significant upfront capital—were the least sensitive to these events.
• The profit mediation paradox: Mediation analysis showed a negative mediation effect for efficiency and recycling, but the magnitude was small (−0.1 percentage points). This suggests that while profit loss (the destruction of organizational slack) restricts an entrepreneur’s coping appraisal, the barrier may be more structural than purely financial.
• Rationalizing green behavior: The fact that recycling and general environmental initiatives saw the highest likelihood of adoption (increments of 10.9 and 13.6 percentage points, respectively) suggests that micro-entrepreneurs gravitate toward low-hanging fruit. These actions offer a high sense of self-efficacy (as per PMT) because they are manageable and immediate, often perceived as a way to protect shrinking profits rather than as a costly long-term investment.

The results also highlight that not all disasters teach equally. Droughts, frosts, and storms showed the largest ATEs on environmental initiatives.

• These events, which affected the largest number of people in the 2020–2025 period, served as effective focusing events.
• They disrupted the status quo so profoundly that they forced a rapid re-prioritization of goals. However, as predicted by the theory of focusing events, the response remains largely reactive. The pressure for a swift response favors visible, direct adaptations over the systemic, long-term structural shifts required for true climate resilience.

This reactive behavior suggests that, in the absence of a structured anticipatory framework, experiential shocks serve as a traumatic substitute for anticipatory capacity (e.g., [45]). Systematic syntheses of strategic foresight demonstrate that these practices can catalyze responsible environmental action by sharpening anticipatory thinking and aligning immediate decisions with long-term sustainability objectives (e.g., [46]). For micro-enterprise ecosystems, this implies that policy should move away from treating disasters as involuntary instructors. Instead, interventions should institutionalize low-cost foresight routines—such as scenario-based peer workshops and forward-looking risk budgeting—specifically designed to collapse the psychological distance of climate threats before physical and financial losses occur.

A key takeaway is that the coping gap is not just a result of a lack of awareness; it is a result of a survivalist mindset (e.g., [11,27]). Micro-entrepreneurs are trapped in a cycle where disasters provide the will to act (threat appraisal) but simultaneously destroy the means to act effectively (coping appraisal). This structural fragility is particularly evident in the Chilean context. As detailed in Table A1, most micro-entrepreneurs in Chile operate with zero financial buffer, as their annual business income is consumed entirely by critical operational expenses, including essential production inputs and employee salaries. When a disaster strikes, the capital required for effective coping, such as retrofitting infrastructure or purchasing insurance, is nonexistent because all available liquidity is locked into maintaining the immediate viability of the business.

7. Conclusions

7.1. Summary of Findings

The study demonstrates that for micro-entrepreneurs in Chile, environmental proactivity is fundamentally driven by direct personal experience rather than abstract geographic risk. While most participants reside in high-risk zones, this proximity fails to act as a significant behavioral trigger. In contrast, direct exposure to disasters serves as a powerful mental anchor, collapsing the psychological distance of climate threats and shifting them from high-level, abstract concepts to low-level, concrete realities. This experiential shift significantly increases the odds of adopting environmental initiatives, particularly in categories like recycling and immediate mitigation, which showed a dramatic rise in likelihood following a disaster.

However, the nature of this proactivity remains largely survivalist and reactive (e.g., [30]). While the will to act increases after exposure, the capacity to implement systemic, high-cost changes—such as energy efficiency or emission reductions—remains severely constrained. The results provide mixed but revealing evidence of a coping gap, where the destruction of business profit and organizational slack following a disaster prevents micro-entrepreneurs from pursuing long-term structural shifts. Instead, they gravitate toward low-barrier initiatives that offer a high sense of self-efficacy and immediate cost-protection, leaving a persistent resilience gap in the face of Chile’s evolving hydro-climatological risk profile.

Methodologically, while the estimates rely on the selection on observables assumption, the potential for omitted variable bias arising from unmeasured factors warrants a cautious interpretation. Idiosyncratic business resilience—such as an entrepreneur’s latent psychological grit or a firm’s undocumented access to informal credit—could simultaneously influence both the survival of a disaster and the subsequent capacity for environmental proactivity. Similarly, specific local vulnerabilities, including variations in the quality of neighborhood-level hydrological infrastructure or disparities in municipal response capacity, may introduce localized effects that influence the reported magnitudes. Consequently, the results should be interpreted as robust evidence consistent with the proposed CLT mechanism, rather than definitive causal proof. While statistical models account for a wide array of demographic and firm-level covariates, the concrete anchor effect remains a behavioral inference.

7.2. Policy Implications

First, policymakers—led by Chile National Service for Disaster Prevention and Response (SENAPRED) in coordination with the Technical Cooperation Service (SERCOTEC)—must recognize that traditional awareness-raising campaigns based on geographic vulnerability are largely ineffective for micro-entrepreneurs. Since behavior is triggered by concrete experience rather than abstract risk, the state should shift from general warnings toward programs that simulate the concrete necessity of adaptation through local peer-learning networks (e.g., [27]). SERCOTEC’s network of Business Centers (Centros de Negocios) could serve as the ideal platform for this, facilitating workshops where resilient peers showcase how they successfully mitigated direct losses. By utilizing social proof to bridge the psychological distance for those not yet impacted, authorities can cultivate a proactive culture without waiting for a catastrophic focusing event to serve as a teacher.

Second, the findings highlight a critical need for financial mechanisms that address the profit-mediation barrier. Because micro-entrepreneurs lack the organizational slack to move beyond reactive measures, SERCOTEC should introduce targeted green resilience subsidies or technical assistance funds specifically for climate-proofing (e.g., [7]). These instruments should move beyond general sustainability and specifically target systemic improvements—like water-efficient technologies for drought-prone zones or fire-resistant infrastructure in the central-south regions—that micro-entrepreneurs currently bypass due to low coping appraisals. By reducing the upfront capital burden through non-reimbursable grants, the state can transform survivalist green behavior into long-term climate resilience, effectively bypassing the liquidity constraints that currently stifle action.

Finally, given that micro-entrepreneurs frequently delegate the burden of large-scale mitigation to the state, SENAPRED must lead an urgent revision of urban-forest interface planning and hydrological infrastructure standards (e.g., [23]). The Chilean experience shows that individual micro-level efforts in recycling or waste management cannot compensate for systemic failures in land-use regulation. State-led initiatives must prioritize the protection of the economic ecosystems where micro-entrepreneurs operate. This requires SENAPRED and the Ministry of Housing and Urban Planning (MINVU) to integrate hydro-climatological risks into regional land-use plans with the same rigor once reserved for seismic resilience. Ensuring that the physical environments of retail and small-scale production are protected from wildfires and floods is the only way to safeguard the country’s new climatological reality and prevent the survivalist cycle from repeating.

7.3. Future Research

Future studies should investigate the longitudinal persistence of the mental anchor created by disaster exposure (e.g., [18]). It remains unclear how long the behavioral shift triggered by a focusing event lasts before the entrepreneur reverts to the status quo, or if psychological distance begins to expand again as time passes. Given the increasing frequency of compound disasters in the Southern Cone—such as the escalating drought and wildfire trends in South-Central Chile (e.g., [15])—tracking a cohort of micro-entrepreneurs over a decade would reveal whether repeated minor exposures build cumulative resilience (e.g., [1]) or lead to habituation and cognitive fatigue. This is particularly relevant when considering how social representations and media narratives of mega-events, like the 2024 Valparaíso fires, shape long-term collective memory and risk perception (e.g., [23]).

Additionally, research should explore the qualitative rationalization process through which micro-entrepreneurs choose specific green initiatives. While this study identifies that profit mediates the type of action taken, deeper ethnographic research could uncover the specific trade-offs entrepreneurs make when under extreme pressure (e.g., [5]). For instance, the decision-making of tourism entrepreneurs in emerging economies often hinges on a delicate balance between immediate economic survival and the preservation of the natural capital their business depends on (e.g., [29]).

Future inquiries should also examine the role of social capital and value chain networks, as seen in Caribbean agribusinesses, where adaptive capacity is often a collective rather than individual attribute (e.g., [27]). Understanding why some actors in coastal communities view pro-environmental behaviors as profit-protecting while others see them as a burden (e.g., [30]) could help in designing context-sensitive nudges. Finally, comparing individual responses to sudden-onset events, such as hail-debris flows (e.g., [14]), against institutional improvements in disaster response (e.g., [12]) would clarify whether state-led interventions empower or inadvertently crowd out private proactive adaptation.