Determinants of Staying Home During the COVID-19 Pandemic–Model Inventory

Abstract

The Determinants of Staying Home During the COVID-19 Pandemic–Model Inventory (Matsunaga et al., 2023) is an advanced social cognitive psychometric instrument designed to evaluate the psychological, normative, and cognitive factors governing individuals’ compliance with stay-at-home non-pharmaceutical interventions (NPIs) during the SARS-CoV-2 pandemic. Grounded in an extended formulation of the Theory of Planned Behavior (TPB; Ajzen, 1991), the instrument augments classic reasoned action constructs—namely, behavioral attitudes, injunctive norms, perceived behavioral control, and behavioral intentions—by integrating personal moral norms and a tripartite operationalization of risk perception. This risk perception component captures cognitive evaluations of infection likelihood and severity, affective fear of contagion, and information exposure frequency across broadcast, digital, and interpersonal communication channels.

Comprising 19 standardized items, the inventory uses 5-point Likert-type scales for 18 social cognitive indicators and a dedicated 7-point frequency metric assessing self-reported weekly shelter-in-place adherence. Developed simultaneously in English and translated into Portuguese, Mandarin Chinese, and Japanese via rigorous committee and back-translation protocols conforming to International Test Commission (ITC) standards, the instrument was validated in multi-country adult cohorts across the United States, Japan, Brazil, and Taiwan. Psychometric evaluation demonstrates robust internal consistency (Cronbach’s α ranging from .69 to .87 across dimensions) and outstanding structural validity through structural equation modeling (χ² = 77.195, p < .001; CFI = .981; NFI = .977; TLI = .948; RMSEA = .064, 90% CI [.051, .078]). Multigroup invariance testing confirmed partial metric invariance across distinct sociopolitical environments, establishing the inventory as a cross-culturally robust diagnostic instrument for health psychology, crisis communication, and epidemiological behavioral modeling.

Keywords

Theory of Planned Behavior, COVID-19, Stay-at-Home Compliance, Risk Perception, Injunctive Norms, Moral Norms, Health Communication, Perceived Behavioral Control, Non-Pharmaceutical Interventions, Structural Equation Modeling, Cross-Cultural Measurement Invariance, Preventive Health Behavior

Authors

The Determinants of Staying Home During the COVID-19 Pandemic–Model Inventory was formulated, tested, and published by an international, multidisciplinary consortium of researchers specializing in behavioral epidemiology, disaster resilience, public health policy, and psychometrics:

• Lucas Heiki Matsunaga — Department of International Environment and Resources Policies, Graduate School of Environmental Studies, Tohoku University, Sendai, Japan. (Correspondence: [email protected]).
• Toshiaki Aoki — Department of International Environment and Resources Policies, Graduate School of Environmental Studies, Tohoku University, Sendai, Japan.
• Cristiane Faiad — Department of Clinical Psychology, Institute of Psychology, University of Brasília (UnB), Brasília, Brazil.
• Daniel P. Aldrich — Security and Resilience Studies Program, Department of Political Science and School of Public Policy and Urban Affairs, Northeastern University, Boston, Massachusetts, United States.
• Po-Hsing Tseng — Department of Shipping and Transportation Management, National Taiwan Ocean University, Keelung, Taiwan.
• Jun Aida — Department of Health Infrastructure and Media Science, Graduate School of Medical and Dental Sciences, Tokyo Medical and Dental University (TMDU), Tokyo, Japan.

Purpose

The primary purpose of the Determinants of Staying Home During the COVID-19 Pandemic–Model Inventory is to provide an empirically validated, cross-culturally viable diagnostic tool capable of quantifying the socio-cognitive, normative, emotional, and communicational drivers of adherence to stay-at-home directives. During large-scale public health emergencies—particularly respiratory pandemics where vaccines and targeted therapeutics are initially absent—non-pharmaceutical interventions (NPIs) like shelter-in-place mandates, physical distancing, and mobility restrictions serve as primary defenses against exponential viral transmission. Nonetheless, voluntary and mandated adherence varies widely across demographics, jurisdictions, and cultural ecosystems.

Prior instruments assessing pandemic compliance often suffered from three main structural shortcomings: an over-reliance on basic sociodemographic correlations without theoretical modeling, a conflation of cognitive risk assessments with affective threat responses, and an absence of cross-cultural measurement invariance testing. This inventory was engineered to overcome these challenges by bridging Ajzen’s Theory of Planned Behavior with Schwartz’s Norm Activation Model and contemporary risk communication theory. It systematically dissects how rational evaluations, moral imperatives, perceived self-efficacy, and multi-channel information exposure converge to shape behavioral intentions and real-world mobility suppression.

In clinical, epidemiological, and governmental settings, the instrument provides health authorities and public policy planners with actionable behavioral intelligence. By determining whether low population adherence stems from deficits in perceived behavioral control (e.g., structural impediments to remote work), lack of injunctive social pressure, low moral obligation, or media fatigue, intervention strategies can be tailored precisely. Rather than deploying generic health warnings, communication architects can calibrate messages—amplifying moral solidarity, reinforcing personal agency, or mitigating specific infection fears depending on empirical construct scores across target populations.

Psychological Construct

The instrument operationalizes stay-at-home compliance through a multi-tiered, nine-factor architecture structured within an integrated social-cognitive framework. Each subscale isolates a distinct psychological or behavioral dimension:

1. Attitude Toward the Behavior

Attitude reflects the individual’s favorable or unfavorable cognitive appraisal and affective evaluation of remaining at home during the pandemic. In accordance with Ajzen’s expectancy-value formulation, this construct measures beliefs regarding whether staying home produces positive utility and desirable societal or personal outcomes (e.g., self-preservation, slowing transmission) versus excessive personal deprivation.

2. Injunctive Norms

Injunctive norms quantify the perceived social pressure from significant others—such as family, peers, employers, and communal authorities—to comply with stay-at-home behaviors. This dimension gauges social approval or disapproval, assessing whether individuals perceive that those who are important to them expect them to shelter in place and whether social sanctions or endorsements operate within their immediate relational ecology.

3. Perceived Behavioral Control (PBC)

PBC measures an individual’s subjective appraisal of their ability, resources, and opportunity to execute the behavior successfully. Drawing upon Bandura’s self-efficacy theory, it captures perceived internal power and external feasibility: how facile or difficult it is to remain sequestered at home given work demands, domestic conditions, living space constraints, and access to essential goods.

4. Behavioral Intention

Intention functions as the proximal cognitive determinant and direct antecedent of manifest action in reasoned action frameworks. This construct captures the conscious motivational commitment, operational plan, and personal effort an individual dedicates toward staying inside and minimizing external community contact.

5. Moral Norms

Departing from purely utilitarian models, moral norms assess internalized personal ethical values and obligations rooted in Schwartz’s altruism frameworks. This dimension taps into anticipated self-evaluative moral emotions—such as guilt, conscience, and ethical responsibility toward shielding vulnerable populations—evaluating whether violating mobility restrictions clashes with internal ethical identity regardless of social surveillance.

6. Risk Perception: Cognitive Risk Evaluation

Risk perception is split into rational and affective channels. Cognitive risk evaluation measures rational probabilistic and severity estimates regarding viral transmission: the perceived likelihood that one will become infected in the near future and the anticipated physiological or personal severity of such an infection.

7. Risk Perception: Fear

Operating as the visceral counterpart to cognitive evaluation, fear captures raw emotional distress and worry surrounding viral exposure. Grounded in the Extended Parallel Process Model (Witte, 1992), this single-item affective indicator reflects the psychological tension generated by the contemplation of falling ill.

8. Risk Perception: Frequency of COVID-19 Information Exposure

Information flow shapes threat appraisals. This subscale measures the density of pandemic-related message encounters across three distinct media vectors: traditional broadcast/print media (television, radio, newspapers), algorithmic digital platforms (social media channels), and immediate interpersonal discourse (face-to-face peer communication).

9. Behavior (Compliance Metric)

The behavioral endpoint is measured as an ecologically anchored frequency index capturing the quantitative number of days per week an individual remains strictly at home, explicitly excluding excursions for purchasing indispensable daily survival necessities.

Theoretical Framework

The architecture of the Determinants of Staying Home During the COVID-19 Pandemic–Model Inventory synthesizes foundational theories from cognitive psychology, sociology, and risk communication into an integrative predictive model. Its theoretical roots draw upon three major paradigms:

The Theory of Planned Behavior (Ajzen, 1991)

At its core, the instrument relies on Icek Ajzen’s Theory of Planned Behavior. The TPB posits that human social behavior is reasoned and steered by three kinds of cognitive considerations: behavioral beliefs regarding consequences (producing attitudes), normative beliefs regarding social expectations (producing subjective/injunctive norms), and control beliefs regarding resources and impediments (producing perceived behavioral control). In combination, attitude, subjective norm, and PBC lead to the formation of a behavioral intention, which acts as the direct precursor to overt behavior, provided actual control over the behavior exists.

Norm Activation and Moral Altruism (Schwartz, 1977)

While standard TPB assumes subjective norms capture social influence, empirical health research shows that social expectations do not fully explain prosocial public health behaviors, which require personal sacrifice for collective safety. Matsunaga et al. integrated Shalom Schwartz’s Norm Activation Model, alongside modern health adaptations (Chan & Bishop, 2013; Tonglet et al., 2004). This framework posits that personal moral norms—internalized self-expectations anchored in core values—become activated when an individual recognizes that their actions carry consequences for others and feels personal responsibility for those outcomes. In pandemic mitigation, staying home is not merely a utility-maximizing calculation; it is an altruistic moral act where non-adherence activates personal guilt and moral distress.

Multidimensional Risk and Health Communication Theory

Classic models often treat risk perception as a uniform variable. However, Matsunaga and colleagues drew upon the work of Dryhurst et al. (2020) and the Extended Parallel Process Model (EPPM; Witte, 1992) to decouple cognitive risk appraisals (probabilistic vulnerability and severity) from visceral affective fear and media consumption patterns. By accounting for the media channels through which pandemic information is consumed—spanning legacy mass media, interactive social networks, and word-of-mouth exchange—the model acknowledges that information environments calibrate fear and risk evaluations, which subsequently shape attitudes and normative stances.

Sociopolitical and Cultural Conditioning

Crucially, the theoretical framework incorporates the modulating influence of local political culture and institutional governance. Public health measures operate within distinct civic traditions. By modeling behavior across four distinct cultural-political matrices—the United States (individualistic, decentralized, polarized), Japan (collectivistic, normative peer pressure, non-coercive legal frameworks), Brazil (high political tension, socioeconomic vulnerability), and Taiwan (early centralized response, collective civic solidarity)—the theoretical structure assesses how institutional trust and political culture recalibrate the structural weights linking norms, attitudes, and behaviors.

Validity

Validation of the Determinants of Staying Home During the COVID-19 Pandemic–Model Inventory was conducted through cross-national empirical testing across four metropolitan populations: the United States, Japan, Brazil, and Taiwan. The resulting psychometric evidence supports multiple dimensions of validity:

Construct and Structural Validity

Construct validity was evaluated using covariance-based structural equation modeling (SEM). The overall theoretical measurement model exhibited outstanding goodness-of-fit to the empirical cross-national data:

• Chi-Square Statistic (χ²): 77.195 (p < .001)
• Comparative Fit Index (CFI): .981 (exceeding the traditional .95 benchmark for superior fit)
• Normed Fit Index (NFI): .977 (indicating excellent relative fit compared to the null model)
• Tucker-Lewis Index (TLI): .948 (confirming robust penalization for model complexity)
• Root Mean Square Error of Approximation (RMSEA): .064 with a narrow 90% confidence interval of [.051, .078], well within the established threshold for acceptable to good approximation error.

Convergent and Discriminant Validity

The indicators for latent constructs (attitude, injunctive norm, perceived behavioral control, moral norm, cognitive risk, and information exposure) showed statistically significant factor loadings on their designated latent factors. The convergence of items measuring the moral norm dimension and TPB constructs confirms that moral obligation captures variance distinct from external social expectations (injunctive norms). Furthermore, the empirical separation of cognitive risk evaluation and affective fear supported discriminant validity within the risk domain, confirming that probabilistic threat assessments operate independently of emotional distress.

Predictive and Criterion Validity

Path coefficients within the structural equation models demonstrated strong predictive validity. Behavioral intention exerted a statistically significant positive effect on the concrete behavioral metric (weekly days staying at home). In parallel, perceived behavioral control, moral norms, and injunctive norms emerged as robust direct and indirect predictors of compliance, accounting for significant variance in sheltering adherence across all tested societies.

Reliability

The scale items demonstrate satisfactory to high internal consistency across all latent dimensions and cultural subsamples. Reliability analysis utilizing Cronbach’s coefficient alpha (α) confirmed structural stability:

• Overall Construct Alpha Range: Across all multi-item subscales, Cronbach’s α coefficients ranged between .69 and .87, satisfying standard psychometric thresholds for academic research and behavioral profiling (Nunnally & Bernstein, 1994).
• Moral Norm Subscale: Demonstrated high internal consistency (α ≥ .83), reflecting strong item covariance among items targeting personal ethical duty, anticipated guilt, and moral alignment.
• Attitude Subscale: Exhibited good internal consistency (α ≈ .81), indicating strong mutual reinforcement between items assessing perceived personal benefit and behavioral desirability.
• Injunctive Norm Subscale: Demonstrated acceptable internal consistency (α > .75), confirming reliable measurement of external social expectations across peer and institutional spheres.
• Perceived Behavioral Control Subscale: Maintained acceptable internal consistency (α > .70), demonstrating that items assessing subjective confidence and situational ease reliably captured behavioral control.
• Information Frequency Subscale: Displayed moderate-to-high internal consistency (α ≈ .69 to .74), adequately capturing broad media usage across traditional, digital, and interpersonal modalities.

No excessive item redundancy was observed, confirming that high internal consistency was driven by genuine conceptual overlap rather than repetitive phrasing.

Factor Analysis and Measurement Invariance

The underlying factor structure and cross-cultural generalizability of the inventory were investigated using confirmatory factor analysis (CFA) alongside multigroup structural equation modeling (MGSEM).

Structural Equation Model Specification

The operationalized measurement model specified distinct latent variables corresponding to the extended TPB and risk perception dimensions:

• Latent construct Attitude indicated by 2 items.
• Latent construct Injunctive Norm indicated by 2 items.
• Latent construct Perceived Behavioral Control indicated by 2 items.
• Latent construct Intention indicated by 2 items.
• Latent construct Moral Norm indicated by 4 items.
• Latent construct Cognitive Risk Evaluation indicated by 2 items.
• Latent construct Frequency of Information indicated by 3 items.
• Single-indicator observed constructs for Fear (1 item) and Behavior (1 item).

The structural paths linked risk perception constructs and information exposure to attitudes, norms, and control beliefs, which channeled into intention and subsequent manifest behavior.

Measurement Invariance Across Four Nations

To establish whether the inventory measures the same underlying social-cognitive constructs equivalently across diverse national populations, multigroup invariance testing was conducted treating the United States, Japan, Brazil, and Taiwan as independent groups:

• Unconstrained vs. Constrained Model Comparison: An unconstrained baseline model (where factor loadings, intercepts, and regression paths varied freely across countries) was compared against progressively constrained structural models.
• Establishment of Partial Invariance: While a totally constrained structural model (imposing strict equality across all structural regression weights) yielded minor decrements in fit due to genuine cultural differences in how norms shape intentions, a partially constrained model achieved exceptional psychometric tenability.
• Invariance Test Statistics: The partially constrained model displayed no statistically significant deterioration in fit compared to the fully unconstrained baseline (ΔCMIN = 16.546, df = 12, p = .167). Changes in global goodness-of-fit indices were well below established cutoff points for rejecting invariance: ΔCFI = .001, ΔTLI = .012, and ΔRMSEA = .004 (Cheung & Rensvold, 2002; Chen, 2007).

These findings demonstrate that the factor structure and item-level operationalizations remain equivalent across Western and East Asian contexts, permitting cross-national comparisons while highlighting meaningful cultural variations in the strength of individual path coefficients.

Instrument / Measurement Tool

The complete instrument parameters and structural characteristics are detailed below:

• Test Type: Original multi-dimensional social-cognitive self-report survey inventory.
• Target Population: General adult population (aged 18 years and older, including young adults and middle-aged/older demographics).
• Number of Items: 19 standardized items (18 social-cognitive indicators + 1 manifest behavioral index).
• Response Scales:
  • Items 1 through 18: 5-point Likert response scale ranging from 1 (“Strongly Disagree” / “Never”) to 5 (“Strongly Agree” / “Very Frequently”).
  • Item 19 (Behavior): 7-point ordinal response scale quantifying weekly adherence (0 to 7 days per week).
• Subscale Breakdown:
  • Attitude (2 items)
  • Injunctive Norms (2 items)
  • Perceived Behavioral Control (2 items)
  • Behavioral Intention (2 items)
  • Moral Norms (4 items)
  • Cognitive Risk Evaluation (2 items)
  • Frequency of COVID-19 Information Exposure (3 items)
  • Fear of Infection (1 item)
  • Manifest Staying Home Behavior (1 item)
• Available Linguistic Adaptations: English, Portuguese (Brazilian), Mandarin Chinese (Traditional), and Japanese. All translations adhered to International Test Commission (ITC, 2017) guidelines for cross-cultural adaptation via forward-translation, back-translation, and expert committee review.
• Scoring and Computational Procedures: Subscale composite scores can be derived by calculating the unweighted arithmetic mean of items loading onto each respective dimension. Alternatively, researchers using latent variable modeling can assign factor loadings via structural equation modeling routines (e.g., in AMOS, Mplus, or the R lavaan package). Higher composite scores indicate stronger endorsement of the respective psychological construct (e.g., higher moral obligation, elevated risk appraisal, or greater perceived behavioral control).

Permissions, Fee, and Test Year

The Determinants of Staying Home During the COVID-19 Pandemic–Model Inventory was developed in 2023 by Lucas Heiki Matsunaga and colleagues. It was initially published in the peer-reviewed Journal of Health Communication under the title “Why people stayed home during the COVID-19 pandemic: Implications for health communication across four countries.”

Fee and Licensing: The measurement model and survey inventory are free of financial charge for scholarly, educational, and public health research purposes. The authors and publishing entity encourage academic implementation and replication. Researchers planning to utilize, adapt, or integrate the inventory within clinical or epidemiological studies are asked to cite the seminal empirical publication (Matsunaga et al., 2023) and, where formal institutional permissions are required, coordinate with the corresponding author via academic correspondence.

References

• Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179–211. https://doi.org/10.1016/0749-5978(91)90020-T
• Chan, D. K., & Bishop, G. D. (2013). Moral obligation, self-efficacy, and the intention to perform health behaviors. Journal of Applied Social Psychology, 43(11), 2207–2219. https://doi.org/10.1111/jasp.12173
• Chen, F. F. (2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modeling: A Multidisciplinary Journal, 14(3), 464–504. https://doi.org/10.1080/10705510701301834
• Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling: A Multidisciplinary Journal, 9(2), 233–255. https://doi.org/10.1207/S15328007SEM0902_5
• Dryhurst, S., van der Bles, A. M., Spiegelhalter, D., & van der Linden, S. (2020). Risk perceptions and science engagement during the COVID-19 pandemic. Journal of Risk Research, 23(7–8), 987–999. https://doi.org/10.1080/13669877.2020.1758193
• International Test Commission. (2017). ITC guidelines for translating and adapting tests (2nd ed.). International Test Commission. https://www.intestcom.org
• Matsunaga, L. H., Aoki, T., Faiad, C., Aldrich, D. P., Tseng, P.-H., & Aida, J. (2023). Why people stayed home during the COVID-19 pandemic: Implications for health communication across four countries. Journal of Health Communication, 28(4), 218–230. https://doi.org/10.1080/10810730.2023.2193149
• Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory (3rd ed.). McGraw-Hill.
• Schwartz, S. H. (1977). Normative influences on altruism. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 10, pp. 221–279). Academic Press. https://doi.org/10.1016/S0065-2601(08)60358-5
• Tonglet, M., Phillips, P. S., & Read, A. D. (2004). Using the theory of planned behaviour to explain recycling behaviour: A case study from Nottingham. Resources, Conservation and Recycling, 41(3), 191–214. https://doi.org/10.1016/j.resconrec.2003.11.001
• Witte, K. (1992). Putting the fear back into fear appeals: The extended parallel process model. Communication Monographs, 59(4), 329–349. https://doi.org/10.1080/03637759209376276

Items of the Scale