Settlement Intention of Foreign Workers in Japan: Bayesian Multinomial Logistic Regression Analysis

Abstract

This study examines the intentions of foreign workers living in Okayama, Japan, to stay long-term in Japan. Utilizing a Bayesian multinomial logistic regression model, this research provides a novel analytical approach that captures parameter uncertainty and accommodates the categorical nature of migrants’ settlement intentions using primary data collected via a questionnaire survey from January to March 2024. The findings reveal that residence status, previous experience of living in Japan, and graduation from a Japanese education institution significantly influence long-term settlement intentions. In addition, respondents aged 26–35 intend to stay longer than those of other ages, and those from less developed countries, such as Myanmar and Vietnam, intend to stay longer than those from China. Conversely, highly educated migrants express lower settlement intentions, suggesting a potential loss of skilled foreign labor in Japan. Notably, migrants in the Technical Intern Training Program are more likely to stay longer than those with other residence statuses, such as Highly Skilled Professional. In contrast, workers with higher education levels tend to have less intention to stay long-term, indicating a high probability of Japan losing educated foreign labor in the future. These findings contribute to understanding the dynamics of migrant workers in Japan, which is crucial for creating policies for foreign workers that can attract and support long-term settlement. These findings have important implications for policy, particularly in enhancing community integration, reducing workplace discrimination, and designing residence pathways that support long-term retention.

1. Introduction

Japan is facing a prominent labor shortage because of its aging population and declining birth rate, resulting in persistent labor shortages in various sectors, especially the labor-intensive sectors of healthcare, construction, and services. As Japan’s workforce continues to shrink, foreign labor has become essential to fill gaps across various sectors of the economy, with foreign workers playing an increasingly crucial role.

Japan’s approach to foreign immigration has evolved significantly over the past century, shaped by economic needs and demographic challenges. Historically, Japan maintained restrictive immigration policies, with limited foreign labor until the 1980s. The economic boom of this period resulted in labor shortages, leading to the introduction of the “Nikkeijin” policy in 1990 and the Technical Intern Training Program (TITP) in 1993, marking the beginning of a gradual shift toward more open migration policies (Kondo, 2002; Sharpe, 2010). In addition, the Specified Skilled Worker (SSW) program was introduced in 2019, which allows a broader range of foreign workers to enter Japan across 16 designated sectors facing severe labor shortages. These sectors include healthcare, construction, agriculture, and service industries, where the demand for workers significantly outpaces the domestic supply. In January 2024, the number of immigrants in Japan reached over 3.32 million, accounting for about 2.66% of Japan’s population (Immigration Services Agency, 2024). The contribution of migrant workers extends beyond merely filling vacant positions; their diverse skills, perspectives, and cultural knowledge can enhance innovation and productivity in the Japanese economy.

Despite these developments, relatively little is known about the long-term settlement intentions of foreign workers in Japan, especially those with different residency statuses, such as TITP and SSW. Furthermore, limited empirical research focuses on how nationality, human capital characteristics, and past experiences within Japan affect the decision to stay. This creates a significant gap in the literature at the intersection of migration policy effectiveness and migrant integration outcomes.

In this study, we therefore explored the factors influencing the long-term settlement of foreign workers in Japan. The following three main research questions are addressed: What are the main factors affecting the intentions of foreign workers of different nationalities to stay long-term in Japan? Have Japanese immigration policies attracted immigrants and been effective in addressing labor shortage problems? Do foreign workers with high human capital intend to stay long-term in Japan? To examine these questions, we employed descriptive and Bayesian multinomial logistic regression analyses. To analyze settlement intentions, this study employed Bayesian multinomial logistic regression, which offers several advantages over traditional frequentist methods. Bayesian approaches provide more flexibility in estimating complex models, allow the incorporation of prior knowledge, and yield full posterior distributions for parameter uncertainty (Washington et al., 2009; van Erp & van Gelder, 2013), making them particularly useful for studies with relatively small or structured samples, such as migrant subgroups.

We found that migrant workers who have already stayed for a long time in Japan (>5 years) and those who graduated from a Japanese education institution have a significantly stronger intention to settle long-term. This study also highlights the effectiveness of Japan’s TITP and SSW program in attracting and retaining foreign workers. Conversely, foreign workers with high education levels (university degree or above) have less intention to stay long-term in Japan than foreign workers who have a low education level and are working in the TITP. In summary, although Japan’s work environment attracts foreign workers with a low education level, it still faces difficulties in encouraging the long-term settlement of highly skilled foreign workers.

This study makes significant key contributions to the literature. First, it is among the few empirical investigations of foreign workers’ settlement intentions in a Japanese regional context—specifically in Okayama Prefecture. Second, it examines the influences of residency status (such as TITP, SSW, etc.) and prior local experience on long-term migration intentions, offering practical insights for labor and immigration policymakers. Third, the use of Bayesian multinomial logistic regression enables nuanced modeling of complex decision-making processes related to migrant settlement.

The remainder of the paper is organized as follows: Section 2 provides an extensive review of the literature, including theoretical and empirical analyses related to the migration decisions and settlement intentions of foreign migrant workers. Section 3 describes the data collection method and methodology. In Section 4, the results are given and analyzed by descriptive and econometric approaches. Section 5 outlines the main conclusions of this study and gives policy recommendations.

2. Literature Review

In the international aspect, the main type of migration has been trans-migration, mainly from developing to developed countries (Barbiano di Belgiojoso, 2016; Paparusso & Ambrosetti, 2017). However, excessive migration creates a burden for major receiving countries, which has become a significant point of controversy in academic fields (Paparusso & Ambrosetti, 2017). Foreign workers play a crucial role in addressing the labor shortage of developed countries. Many studies argue that migration has reciprocal effects on the trajectories of both origin and destination countries (Joly, 2000; Tsuda, 2012). However, this impact may vary among countries facing a labor shortage, such as Japan.

Numerous previous studies on settlement intention have emphasized the importance of economic, institutional, and cultural factors in migrants’ decisions on whether to stay long-term or permanently in destination countries, return to their origin countries, or remigrate to third countries (Ette et al., 2021). Barbiano di Belgiojoso (2016) explored the settlement intention of migrants in two ways by applying logistic regression. Their study found that attachment to the host country significantly influences the decision to stay long-term and also observed indecision resulting from negative experiences as migrants. Moreover, attachment to the host country was found to be affected by political, economic, and socio-cultural factors, with legal residential status, professional stability, the presence of family members, a feeling of belonging to the host country, the ability to save money in the host country, and life cycle strongly connected with the decision of long-term settlement.

Liu et al. (2018) studied rural–urban migration in China through the statistical and spatial analysis of GIS data and the hierarchical linear regression model. They found that education, age, income, employment status, and occupation type had significant effects on intention to settle in urban areas. Moreover, Bekaert et al. (2021) found that social networks significantly influence immigrants’ aspirations, with international networks encouraging outmigration and local networks encouraging settlement. In addition, economic integration in the destination country, aside from satisfaction with living standards, was not found to notably impact outmigration aspirations. Instead, factors such as community acceptance and satisfaction with local amenities (education, environment, and healthcare) reduced the desire to outmigrate. Sapeha (2017) explored the intention of immigrants to settle or leave their initial destination using multinomial logistic regression. They found that highly educated and skilled migrants tended to have a greater intention to move because of doubts about their current country of residence, although satisfactory employment had a positive impact on their intention to stay. Psychological integration through place attachment, linked with social integration, can encourage migrants to stay, with structural and cultural factors having stronger influences than economic factors. For example, many Ukrainian migrants in Warsaw have formed strong ties with local people, enhancing their place attachment and settlement intentions. Emotional social capital, rather than institutional factors, indirectly affects these intentions, especially for repeat migrants, regardless of structural integration or the time spent in Poland (Toruńczyk-Ruiz & Brunarska, 2020).

Su et al. (2018) examined rural–urban migration using a nested logit model and found that intra-provincial migrants are more likely than inter-provincial migrants to permanently settle in cities within their home provinces because higher wage differentials, a larger population size, higher GDP per capita, and faster employment growth attract migrants from both within and outside provinces. Hu (2023) applied latent class logit models to classify internal migrants in China, with the results indicating that access to social welfare is the most influential factor in destination city selection for all migrant types, especially those with lower mobility. For migrants with higher mobility, social welfare remains important, although expected wages and urban amenities are also significant factors in their location and settlement choices. Dalla Valle et al. (2020) proposed an efficient Bayesian approach for modeling binary response data based on generalized logistic regression and analyzed immigrants moving to and settling in the European Union, which has become a key political issue there. They introduced an efficient Bayesian approach for modeling binary response data using generalized logistic regression, addressing the complexities of immigration. Their analysis underscores the significant economic, social, and cultural impacts of rising immigration on the European Union and highlights the need for policymakers to correctly evaluate people’s attitudes toward immigration when designing integration policies.

Ette et al. (2021) analyzed the settlement and return of German migrants across their life course by multinomial logistic regression. The results suggested that highly qualified migrants tend to return sooner, with migration intentions influenced by individual life-course status rather than economic success or failure. A life-course perspective better explains return and settlement intentions, but there was little consideration of life events and destination country effects in these studies. Barbiano di Belgiojoso et al. (2023) applied three logistic regression models to study changes in the intentions of migrants regarding settlement in Italy. They found that financial stability, family situation, ties with the country of origin, and destination were critical factors determining the intention to settle. In particular, living with a partner from the destination country and having a positive self-assessed family financial condition were factors that tended to change settlement intention from a temporary to permanent stay.

Following conventional migration theories, Yang (2020) explored how firms differentiate between foreign and native workers, finding that such labor segmentation significantly affects outmigration intention. Specifically, for Asian-born highly skilled foreign workers in Japan, the country’s lifetime employment system for all workers was considered attractive. Chitose (2022) performed multinominal logistic regression on surveys conducted in Shizuoka Prefecture and Iwata City to analyze return intentions among Brazilian migrants. Their findings support the new economics of labor migration theory, indicating that higher income and remittances positively affect their return intentions. Other factors, including age, gender, status of residence, length of stay, fluency in Japanese, and having Japanese friends, did not have significant effects on long-term settlement.

Conversely, factors negatively linked to return intentions included home ownership and co-residence with children under 15 years old in Japan. This study, along with previous research (Constant & Massey, 2002; De Haas & Fokkema, 2011; De Haas et al., 2015), underscores the heterogeneity of immigrants and their settlement intentions, calling for future research to better understand the conditions under which immigrants decide to settle and what causes their motivations to shift.

While socio-economic variables such as income or type of employment are often considered important in studies of migration intention, these were not included in our model due to data limitations. Our primary focus was on demographic, institutional, and experiential factors (e.g., nationality, education level, residency status, and length of stay) that have been widely recognized in the migration literature as direct determinants of long-term settlement intention (Windzio et al., 2011; Barbiano di Belgiojoso et al., 2023). This study used proxy indicators such as residence status (e.g., HSP, SSW, and TITP) and education level, which are strongly associated with both the employment type and earning potential, allowing us to capture similar explanatory power while maintaining model reliability.

In summary, in the literature, there are numerous similarities and differences among the factors associated with the settlement intention of migrants in the destination country. However, most existing studies have focused on Western countries or large urban centers in East Asia, with limited attention to Japan’s unique immigration structure or the regional disparities within the country. Specifically, there is a lack of empirical research analyzing how Japan’s relatively new TITP and SSW program influence the settlement intentions of foreign workers.

However, there has been limited research on migrants in specific regions or with specific nationalities in Japan. There has also been no previous research on how the status of residence (for example, workers in the SSW program and TITP) and previous experience of living in Japan impact intentions of long-term settlement. To address this research gap, we explored the significant factors that influence the long-term settlement of foreign workers in Okayama, Japan. The results of this study highlight the key factors for policymakers to consider when formulating effective policies to retain foreign labor in sectors facing labor shortages. Furthermore, this study explores the significant factors that influence the long-term settlement of foreign workers in Okayama, Japan. This regional focus allows us to provide grounded insights into the everyday experiences of migrants beyond major urban centers. Furthermore, by explicitly engaging with theoretical frameworks, this study contributes to the development of a more context-sensitive understanding of migrant settlement in non-Western, rapidly aging societies.

3. Data and Methodology

3.1. Data

Data were collected via a questionnaire survey administered through Google Forms from 20 January to 29 March 2024. Okayama City was purposefully selected as the survey area. Survey questionnaires were distributed via the homepage and Facebook page of Hashimoto Foundation, organizations supporting foreign workers, and companies recruiting foreign workers in Okayama Prefecture. A total of 501 responses were initially received from the survey.

However, it was discovered during the data-cleaning procedure that some respondents finished the survey quickly. In response, answers to surveys that took a total of under 30 s to complete were not included in the analysis. Based on the pilot survey, the complete questionnaire required approximately 15–20 min to answer thoroughly. Therefore, responses submitted in under 30 s were deemed unreliable and likely invalid, as such a short duration was insufficient to provide considered and accurate answers. Missing and incomplete surveys were also eliminated. Following these removals, 374 valid responses remained, giving our final sample size for this study.

3.2. Methodology

The dependent variable of this study is the intention to stay long-term in Japan, which is categorized as follows: (1) intention to stay less than 5 years, (2) intention to stay 5–9 years, (3) intention to stay 10 years or more, and (4) undecided. Gender, age, education, country of origin, residence status, graduation from a Japanese education institution, and years of staying in Japan are taken as independent variables.

When the outcome variable has more than two categories, a multinomial logistic regression analysis is suitable for predicting its probability (Hosmer et al., 2013; Osanya et al., 2020). In traditional multinomial regression, the results are often interpreted based on the p-value. However, the p-value has faced substantial criticism for its limitations in conveying practical significance or the probability of hypotheses, and this approach also has problems with reproducibility (Halsey, 2019; Held & Ott, 2018). These critiques have led many researchers to recommend Bayesian methods, which provide a fuller picture of uncertainty and offer more powerful interpretation tools, such as the Bayes factor and credible intervals, thereby enhancing the precision of the results.

Thus, we applied Bayesian multinomial logistic regression to obtain posterior distributions of model parameters, which provide more robust interpretations of stay intentions across the categories. Bayesian approaches allow us to incorporate prior information and better handle the uncertainty inherent in small sample sizes or sparse categories (Gelman et al., 2008). Bayesian multinomial logistic regression is applied due to its flexibility, capacity to incorporate prior information, and superior handling of parameter uncertainty compared to traditional logistic regression (Washington et al., 2009; van Erp & van Gelder, 2013). Additional, for the prior of this research study, it would be useful to use the normal prior for estimation. This prior is still one type of informative prior because the researchers of this study believe that the structure of the dataset in this research is very complex in terms of the settlement intention of foreign workers in Japan. Therefore, the normal prior, which is one type of informative prior, is used to address this problem by defining the ranges of means and variances of all parameters estimated in the Bayesian multinomial regression model, which are narrower (Gelman et al., 2008; Johnson, 2014; Gosho et al., 2024).

For a multinomial outcome Y_i, the probability of individual i falling into category k (k ∈ {1, 2, 3, 4}) is given by

$$\mathrm{P}(Y_i=k|X_i)={\displaystyle \frac{\mathrm{e}\mathrm{x}\mathrm{p} (X_i{\beta }_k)}{1+{\sum }_{j=2}^k \mathrm{e}\mathrm{x}\mathrm{p} (X_i{\beta }_j)}}, \mathrm{for} k=2,3,4.$$

(1)

For the reference category k = 1 (less than 5 years), the equation is

$$\mathrm{P}(Y_i=1|X_i)={\displaystyle \frac{1}{1+{\sum }_{j=2}^k\mathrm{e}\mathrm{x}\mathrm{p} (X_i{\beta }_j)}}.$$

(2)

Here,

• P($Y_i$= k|$X_i$) is the probability of the ith individual choosing category k.
• $X_i$ is the vector of independent variables.
• ${\beta }_k$ is the vector of coefficients for category k estimated relative to the reference category.

In Bayesian multinomial logistic regression, the posterior distribution of the parameters ${\beta }_k$ is estimated.

The steps in Bayesian estimation are as follows:

• Prior distribution—a prior distribution is assigned to the parameters. Typically, a normal prior is used

$${\beta }_k\sim \mathrm{N}(0,{\sigma }^2).$$

This prior reflects beliefs about the parameters before observing the data.

2.

Likelihood function—the likelihood is the probability of observing the data given the parameters

$$L\left(\beta \right)={\Pi }_{i=1}^n{\Pi }_{k=1}^K\mathrm{P}{(Y_i=k|X_i)}^{\mathit{yik}}.$$

3.

Posterior distribution—the posterior distribution of the parameters is obtained using Bayes’ theorem

p(β∣Y, X) ∝ L(β)  p(β),

where p(β) is the prior and L(β) is the likelihood.

4.

Markov chain Monte Carlo (MCMC) sampling—since the posterior distribution is analytically intractable, MCMC methods (such as Gibbs sampling or Hamiltonian Monte Carlo) are used to sample from the posterior distribution and generate estimates for β_k.

4. Analysis of the Results

4.1. Descriptive Analysis

Descriptive analysis results, including the Pearson chi-square value and p-value, are presented in

Table 1. Intention to stay by gender.

Gender	Intended Years of Stay in Japan
	Less than 5	5–9	10 or More	Undecided	Total
Female	49 (29.4)	51 (30.5)	51 (30.5)	16 (9.6)	167 (44.7)
Male	72 (35.5)	86 (42.4)	38 (18.7)	7 (3.4)	203 (54.3)
Other	2 (50)	2 (50)	0	0	4 (1.0)
Total	123 (32.9)	139 (37.2)	89 (23.8)	23 (6.1)	374 (100)

Pearson chi² = 17.23, p = 0.008.

, Table 2, Table 3, Table 4, Table 5 and Table 6. Table 1 illustrates the intended years of stay in Japan by gender, where females, males, and other responses make up 44.7%, 54.3%, and 4% of the total sample, respectively. For female respondents, 29.4% intend to stay less than 5 years, 30.5% intend to stay 5–9 years, 30.5% intend to stay 10 years or more, and 9.6% are undecided. For male respondents, 35.5% intend to stay less than 5 years, 42.4% intend to stay 5–9 years, 18.7% intend to stay 10 years or more, and 3.4% are undecided. These results indicate that male respondents intend to stay for less time than female respondents and have a clearer idea about how long they intend to stay.

Values inside and outside parentheses are the frequencies and percentages of the row total, respectively. The “Total” row gives the frequency and percentage of the total sample for each duration. The “Total” column gives the frequency and percentage of the total sample for each gender.

Table 2 illustrates the intended years of stay in Japan by nationality. Chinese nationals comprise 47.9% of the total sample, Myanmar nationals 11.2%, Vietnamese nationals 31.3%, and other nationalities 9.6%. Among the Chinese nationals, 43% intend to stay less than 5 years, 44.7% intend to stay 5–9 years, 11.2% intend to stay 10 years or more, and 1.1% are undecided. Similarly, among the Myanmar nationals, 16.7% intend to stay less than 5 years, 30.9% intend to stay 5–9 years, 42.9% intend to stay 10 years or more, and 9.5% are undecided. Among the Vietnamese nationals, 20.5% intend to stay less than 5 years, 30.8% intend to stay 5–9 years, 36.7% intend to stay 10 years or more, and 12% are undecided. Among the people of other nationalities, 41.7% intend to stay less than 5 years, 27.8% intend to stay 5–9 years, 22.2% intend to stay 10 years or more, and 8.3% are undecided. The Pearson chi-square value (62.09) and p-value (0.000) indicate that the differences observed are statistically significant at less than the 5% significance level, meaning they are not due to chance.

Table 2. Intention to stay by nationality.

Nationality	Intended Years of Stay in Japan
	Less than 5	5–9	10 or More	Undecided	Total
China	77 (43.0)	80 (44.7)	20 (11.2)	2 (1.1)	179 (47.9)
Myanmar	7 (16.7)	13 (30.9)	18 (42.9)	4 (9.5)	42 (11.2)
Vietnam	24 (20.5)	36 (30.8)	43 (36.7)	14 (12.0)	117 (31.3)
Others	15 (41.7)	10 (27.8)	8 (22.2)	3 (8.3)	36 (9.6)
Total	123 (32.9)	139 (37.2)	89 (23.8)	23 (6.1)	374 (100)

Pearson chi² = 62.09, p = 0.000.

Table 2. Intention to stay by nationality.

Nationality	Intended Years of Stay in Japan
	Less than 5	5–9	10 or More	Undecided	Total
China	77 (43.0)	80 (44.7)	20 (11.2)	2 (1.1)	179 (47.9)
Myanmar	7 (16.7)	13 (30.9)	18 (42.9)	4 (9.5)	42 (11.2)
Vietnam	24 (20.5)	36 (30.8)	43 (36.7)	14 (12.0)	117 (31.3)
Others	15 (41.7)	10 (27.8)	8 (22.2)	3 (8.3)	36 (9.6)
Total	123 (32.9)	139 (37.2)	89 (23.8)	23 (6.1)	374 (100)

Pearson chi² = 62.09, p = 0.000.

Values inside and outside parentheses are the frequencies and percentages of the row total, respectively. The “Total” row gives the frequency and percentage of the total sample for each duration. The “Total” column gives the frequency and percentage of the total sample for each nationality.

Table 3 shows the expected years of stay in Japan by age group. The age range of the respondents is from 19 to 53 years old. The majority of respondents fall within the 26–35 year age group, making up around 60% of the total sample size. Among the respondents aged 19–25 years, which comprise 31.5% of the total sample, most (46.6%) intend to stay less than 5 years, 26.3% plan to stay for 5–9 years, 22.9% intend to stay for 10 years or more, and 4.2% are undecided. Among those in the 26–35 year age group, 25.9% intend to stay less than 5 years, 43.3% plan to stay for 5–9 years, 24.1% intend to stay for 10 years or more, and 6.7% are undecided.

Table 3. Intention to stay by age group.

Age Group	Intended Years of Stay in Japan
	Less than 5	5–9	10 or More	Undecided	Total
19–25	55 (46.6)	31 (26.3)	27 (22.9)	5 (4.2)	118 (31.5)
26–35	58 (25.9)	97 (43.3)	54 (24.1)	15 (6.7)	224 (59.9)
36–53	10 (31.2)	11 (34.4)	8 (25.0)	3 (9.4)	32 (8.6)
Total	123 (32.9)	139 (37.2)	89 (23.8)	23 (6.1)	374 (100)

Pearson chi² = 17.64, p = 0.007.

Table 3. Intention to stay by age group.

Age Group	Intended Years of Stay in Japan
	Less than 5	5–9	10 or More	Undecided	Total
19–25	55 (46.6)	31 (26.3)	27 (22.9)	5 (4.2)	118 (31.5)
26–35	58 (25.9)	97 (43.3)	54 (24.1)	15 (6.7)	224 (59.9)
36–53	10 (31.2)	11 (34.4)	8 (25.0)	3 (9.4)	32 (8.6)
Total	123 (32.9)	139 (37.2)	89 (23.8)	23 (6.1)	374 (100)

Pearson chi² = 17.64, p = 0.007.

Values inside and outside parentheses are the frequencies and percentages of the row total, respectively. The “Total” row gives the frequency and percentage of the total sample for each duration. The “Total” column gives the frequency and percentage of the total sample for each age group.

Table 4 gives the intended years of stay in Japan for the respondents with different levels of education. Among the respondents, 20.3% did not graduate from university and 79.7% have a university degree or above. For the former respondents, 28.9% intend to stay less than 5 years, 23.7% intend to stay 5–9 years, 39.5% intend to stay 10 years or more, and 7.9% are undecided. For the latter category, 33.9% intend to stay less than 5 years, 37.2% intend to stay 5–9 years, 19.8% intend to stay 10 years or more, and 5.7% are undecided.

Table 4. Intention to stay by education level.

Education	Intended Years of Stay in Japan
	Less than 5	5–9	10 or More	Undecided	Total
School	22 (28.9)	18 (23.7)	30 (39.5)	6 (7.9)	76 (20.3)
University or above	101 (33.9)	121 (40.6)	59 (19.8)	17 (5.7)	298 (79.7)
Total	123 (32.9)	139 (37.2)	89 (23.8)	23 (6.1)	374 (100)

Pearson chi² = 15.43, p = 0.001.

Table 4. Intention to stay by education level.

Education	Intended Years of Stay in Japan
	Less than 5	5–9	10 or More	Undecided	Total
School	22 (28.9)	18 (23.7)	30 (39.5)	6 (7.9)	76 (20.3)
University or above	101 (33.9)	121 (40.6)	59 (19.8)	17 (5.7)	298 (79.7)
Total	123 (32.9)	139 (37.2)	89 (23.8)	23 (6.1)	374 (100)

Pearson chi² = 15.43, p = 0.001.

Values inside and outside parentheses are the frequencies and percentages of the row total, respectively. The “Total” row gives the frequency and percentage of the total sample for each duration. The “Total” column gives the frequency and percentage of the total sample for each level of education.

Table 5 gives the responses in terms of previous number of years of stay in Japan. Among the respondents who have stayed in Japan for less than 5 years, 50% intend to stay less than 5 years, 25.7% intend to stay 5–9 years, 18.6% intend to stay 10 years or more, and 5.7% are undecided. On the other hand, among the respondents who have stayed 5 years or more in Japan, 6.8%, 54.7%, 31.8%, and 6.7% intend to stay less than 5 years, 5–9 years, and 10 years or more, and are undecided, respectively.

Table 5. Intention to stay by previous years of stay.

Years of Stay in Japan	Intended Years of Stay in Japan
	Less than 5	5–9	10 or More	Undecided	Total
Less than 5	113 (50.0)	58 (25.7)	42 (18.6)	13 (5.7)	226 (60.4)
5 or more	10 (6.8)	81 (54.7)	47 (31.8)	10 (6.7)	148 (39.6)
Total	123 (32.9)	139 (37.2)	89 (23.8)	23 (6.1)	374 (100)

Pearson chi² = 77.85, p = 0.000.

Table 5. Intention to stay by previous years of stay.

Years of Stay in Japan	Intended Years of Stay in Japan
	Less than 5	5–9	10 or More	Undecided	Total
Less than 5	113 (50.0)	58 (25.7)	42 (18.6)	13 (5.7)	226 (60.4)
5 or more	10 (6.8)	81 (54.7)	47 (31.8)	10 (6.7)	148 (39.6)
Total	123 (32.9)	139 (37.2)	89 (23.8)	23 (6.1)	374 (100)

Pearson chi² = 77.85, p = 0.000.

Values inside and outside parentheses are the frequencies and percentages of the row total, respectively. The “Total” row gives the frequency and percentage of the total sample for each duration. The “Total” column gives the frequency and percentage of the total sample for the different years of stay in Japan.

Table 6 gives the results for intended years of stay in Japan according to residence status. Highly Skilled Professional visa holders comprise 26.2% of the total sample, h Engineer/Specialist in Humanities/International Services visa holders account for 22.5%, students make up 19.8%, people in the TITP and SSW program make up 11.2% and 10.7%, respectively, and other visa holders make up 9.6%. Highly Skilled Professional visa holders mainly intend to stay for 5–9 years and 10 years or more. On the other hand, most students plan to stay less than 5 years and are likely to return earlier than those with other residence statuses.

Table 6. Intention to stay by residence status.

Residence Status	Intended Year of Stay in Japan
	Less than 5	5–9	10 or More	Undecided	Total
HSP	28 (28.6)	51 (52.0)	18 (18.4)	1 (1.0)	98 (26.2)
ESHI	27 (32.1)	33 (39.3)	17 (20.3)	7 (8.3)	84 (22.5)
Student	42 (56.8)	15 (20.3)	12 (16.2)	5 (6.7)	74 (19.8)
TITP	9 (21.4)	21 (50.0)	9 (21.4)	3 (7.2)	42 (11.2)
SSW	9 (22.5)	11 (27.5)	18 (45.0)	2 (5.0)	40 (10.7)
Other	8 (2.1)	8 (2.1)	15 (4.0)	5 (1.4)	36 (9.6)
Total	123 (32.9)	139 (37.2)	89 (23.8)	23 (6.1)	374 (100)

Pearson chi² = 58.76, p = 0.000. HSP = Highly Skilled Professional, ESHI = Engineer/Specialist in Humanities/International Services, TITP = Technical Intern Training Program, SSW = Specified Skilled Worker.

Table 6. Intention to stay by residence status.

Residence Status	Intended Year of Stay in Japan
	Less than 5	5–9	10 or More	Undecided	Total
HSP	28 (28.6)	51 (52.0)	18 (18.4)	1 (1.0)	98 (26.2)
ESHI	27 (32.1)	33 (39.3)	17 (20.3)	7 (8.3)	84 (22.5)
Student	42 (56.8)	15 (20.3)	12 (16.2)	5 (6.7)	74 (19.8)
TITP	9 (21.4)	21 (50.0)	9 (21.4)	3 (7.2)	42 (11.2)
SSW	9 (22.5)	11 (27.5)	18 (45.0)	2 (5.0)	40 (10.7)
Other	8 (2.1)	8 (2.1)	15 (4.0)	5 (1.4)	36 (9.6)
Total	123 (32.9)	139 (37.2)	89 (23.8)	23 (6.1)	374 (100)

Pearson chi² = 58.76, p = 0.000. HSP = Highly Skilled Professional, ESHI = Engineer/Specialist in Humanities/International Services, TITP = Technical Intern Training Program, SSW = Specified Skilled Worker.

Values inside and outside parentheses are the frequencies and percentages of the row total, respectively. The “Total” row gives the frequency and percentage of the total sample for each duration. The “Total” column gives the frequency and percentage of the total sample for each residence status.

4.2. Econometric Analysis

The Bayesian multinomial logistic regression model was implemented using Stata’s default priors. The values in Table 7 are the mean coefficients, standard deviations, Monte Carlo standard errors (MCSEs), medians, and 95% equal-tailed credible intervals (95% CIs) for residents intending to stay another 5–9 years in Japan. Positive coefficients indicate an increased likelihood of staying within the specified group (group 2 vs. group 1). The age of the respondents was categorized into three groups: 19–25 years, 26–35 years, and 36 years or above. The 26–35 year age group has a positive mean coefficient of 0.821 and its confidence interval does not include zero (95% CI: 0.265, 1.371), indicating that respondents in this group are more likely to intend staying for 5–9 years than those in the other age groups.

Respondents from Myanmar and Vietnam are more likely to intend staying 5–9 years than Chinese nationals, having mean coefficients of 1.939 (95% CI: 1.162, 2.538) and 0.897 (95% CI: 0.388, 1.418), respectively. Additionally, respondents with HSP and TITP statuses, with mean coefficients of 0.519 (95% CI: 0.062, 1.004) and 1.564 (95% CI: 1.063, 2.084), respectively, are more likely to stay for 5–9 years than those with other residence statuses.

Similarly, the mean coefficient of 1.007 (95% CI: 0.317, 1.628) for respondents who graduated from Japanese education institutions indicates a positive and significant association with intending to stay 5–9 years compared with the reference category. Furthermore, the mean coefficient of 2.793 (95% CI: 1.984, 3.501) for respondents who have stayed in Japan more than 5 years indicates a strong and positive effect; namely, people who have stayed in Japan for more than 5 years are more likely to intend to stay 5–9 years than those who have stayed less than 5 years.

Table 7. Bayesian multinomial logistic regression results for the stay intention of 5–9 years (group 2 vs. group 1).

Group 2 vs. Group 1	Mean	Std. Dev.	MCSE	Median	95% CI (Equal-Tailed)
Male	0.229	0.207	0.037	0.216	(−0.163, 0.632)
Age 26–35 years	0.821	0.287	0.046	0.807	(0.265, 1.371)
Age 36 years or above	0.406	0.415	0.043	0.414	(−0.433, 1.206)
University degree or above	−0.366	0.279	0.036	−0.368	(−0.943, 0.172)
Nationality: Myanmar	1.939	0.330	0.068	1.969	(1.162, 2.538)
Nationality: Vietnam	0.897	0.265	0.023	0.896	(0.388, 1.418)
Nationality: Other	0.153	0.441	0.050	0.158	(−0.702, 0.996)
Residence status: HSP	0.519	0.239	0.021	0.517	(0.062, 1.004)
Residence status: ESHI	−0.281	0.278	0.038	−0.273	(−0.833, 0.266)
Residence status: TITP	1.564	0.252	0.035	1.565	(1.063, 2.084)
Residence status: SSW	0.648	0.357	0.030	0.649	(−0.044, 1.357)
Residence status: Others	0.242	0.486	0.101	0.217	(−0.649, 1.279)
Recent education in Japan	1.007	0.331	0.032	1.008	(0.317, 1.628)
Stay in Japan >5 years	2.793	0.401	0.087	2.819	(1.984, 3.501)
_cons	−2.132	0.301	0.039	−2.128	(−2.719, −1.556)

Table 7. Bayesian multinomial logistic regression results for the stay intention of 5–9 years (group 2 vs. group 1).

Group 2 vs. Group 1	Mean	Std. Dev.	MCSE	Median	95% CI (Equal-Tailed)
Male	0.229	0.207	0.037	0.216	(−0.163, 0.632)
Age 26–35 years	0.821	0.287	0.046	0.807	(0.265, 1.371)
Age 36 years or above	0.406	0.415	0.043	0.414	(−0.433, 1.206)
University degree or above	−0.366	0.279	0.036	−0.368	(−0.943, 0.172)
Nationality: Myanmar	1.939	0.330	0.068	1.969	(1.162, 2.538)
Nationality: Vietnam	0.897	0.265	0.023	0.896	(0.388, 1.418)
Nationality: Other	0.153	0.441	0.050	0.158	(−0.702, 0.996)
Residence status: HSP	0.519	0.239	0.021	0.517	(0.062, 1.004)
Residence status: ESHI	−0.281	0.278	0.038	−0.273	(−0.833, 0.266)
Residence status: TITP	1.564	0.252	0.035	1.565	(1.063, 2.084)
Residence status: SSW	0.648	0.357	0.030	0.649	(−0.044, 1.357)
Residence status: Others	0.242	0.486	0.101	0.217	(−0.649, 1.279)
Recent education in Japan	1.007	0.331	0.032	1.008	(0.317, 1.628)
Stay in Japan >5 years	2.793	0.401	0.087	2.819	(1.984, 3.501)
_cons	−2.132	0.301	0.039	−2.128	(−2.719, −1.556)

Table 8 illustrates the results of the Bayesian multinomial logistic regression comparing the intention to stay 10 years or more (group 3) with the base group (less than 5 years, group 1). The mean coefficient of 1.357 (95% CI: 0.711, 1.949) for the age group of 26–35 years indicates a positive and statistically significant association with intending to stay 10 years or more compared with the reference age group. The mean coefficient of −1.976 (95% CI: −2.852, −1.134) for people with a university degree or higher illustrates a strong negative association with intending to stay 10 years or more compared with the base group, indicating that more highly educated respondents are less likely to stay in Japan for 10 years or more and are more likely to seek careers in their home country or elsewhere than in Japan.

The highly significant mean coefficients of 4.689 (95% CI: 3.911, 5.540), 2.839 (95% CI: 2.363, 3.322), and 1.997 (95% CI: 1.014, 2.956) for respondents from Myanmar, Vietnam, and other countries, respectively, indicate their greater likelihood of intending to stay in Japan for 10 years or more than the Chinese respondents.

The mean coefficients of 1.772 (95% CI: 1.031, 2.477) and 3.300 (95% CI: 2.478, 4.104) for residents with a recent education acquired from Japanese institutions and people having already stayed in Japan for more than 5 years, respectively, indicate a strong positive association with intending to stay 10 years or more in Japan, suggesting that they are more likely to intend staying in Japan for at least 10 years than the reference category.

Table 8. Bayesian multinomial logistic regression results for the stay intention of 10 years or more (group 3 vs. group 1).

Group 3 vs. Group 1	Mean	Std. Dev.	MCSE	Median	95% CI (Equal-Tailed)
Male	−0.184	0.305	0.039	−0.181	(−0.781, 0.393)
Age 26–35 years	1.357	0.315	0.051	1.369	(0.711, 1.949)
Age 36 years or above	0.479	0.514	0.091	0.422	(−0.414, 1.595)
University degree or higher	−1.976	0.441	0.054	−1.990	(−2.852, −1.134)
Nationality: Myanmar	4.689	0.413	0.053	4.677	(3.911, 5.540)
Nationality: Vietnam	2.839	0.246	0.030	2.832	(2.363, 3.322)
Nationality: Other	1.997	0.501	0.047	1.994	(1.014, 2.956)
Residence status: HSP	0.608	0.393	0.037	0.610	(−0.137, 1.358)
Residence status: ESHI	−0.879	0.467	0.043	−0.882	(−1.800, 0.068)
Residence status: TITP	0.008	0.246	0.032	0.013	(−0.494, 0.496)
Residence status: SSW	0.406	0.464	0.054	0.409	(−0.489, 1.324)
Residence status: Others	0.516	0.328	0.053	0.510	(−0.099, 1.154)
Recent education in Japan	1.772	0.364	0.059	1.768	(1.031, 2.477)
Stay in Japan >5 years	3.300	0.409	0.073	3.318	(2.478, 4.104)
_cons	−3.099	0.278	0.029	−3.092	(−3.658, −2.552)

Table 8. Bayesian multinomial logistic regression results for the stay intention of 10 years or more (group 3 vs. group 1).

Group 3 vs. Group 1	Mean	Std. Dev.	MCSE	Median	95% CI (Equal-Tailed)
Male	−0.184	0.305	0.039	−0.181	(−0.781, 0.393)
Age 26–35 years	1.357	0.315	0.051	1.369	(0.711, 1.949)
Age 36 years or above	0.479	0.514	0.091	0.422	(−0.414, 1.595)
University degree or higher	−1.976	0.441	0.054	−1.990	(−2.852, −1.134)
Nationality: Myanmar	4.689	0.413	0.053	4.677	(3.911, 5.540)
Nationality: Vietnam	2.839	0.246	0.030	2.832	(2.363, 3.322)
Nationality: Other	1.997	0.501	0.047	1.994	(1.014, 2.956)
Residence status: HSP	0.608	0.393	0.037	0.610	(−0.137, 1.358)
Residence status: ESHI	−0.879	0.467	0.043	−0.882	(−1.800, 0.068)
Residence status: TITP	0.008	0.246	0.032	0.013	(−0.494, 0.496)
Residence status: SSW	0.406	0.464	0.054	0.409	(−0.489, 1.324)
Residence status: Others	0.516	0.328	0.053	0.510	(−0.099, 1.154)
Recent education in Japan	1.772	0.364	0.059	1.768	(1.031, 2.477)
Stay in Japan >5 years	3.300	0.409	0.073	3.318	(2.478, 4.104)
_cons	−3.099	0.278	0.029	−3.092	(−3.658, −2.552)

Table 9 illustrates the results of the Bayesian multinomial logistic regression comparing the undecided group (group 4) and the base group (less than 5 years, group 1). The negative and significant mean coefficient of −0.897 (95% CI: −1.731, −0.105) for the male respondents indicates that they are clearer in their intention to stay in Japan than the female respondents. Likewise, the positive and significant mean coefficients of 2.338 (95% CI: 1.448, 3.309) and 1.500 (95% CI: 0.061, 2.899) for the age groups of 26–35 years and 36 years and above, respectively, indicate that these respondents are more likely to fall into the undecided group than the baseline group.

The mean coefficients of 4.453 (95% CI: 3.246, 5.838), 3.559 (95% CI: 2.594, 4.504), and 2.542 (95% CI: 2.009, 3.065) for respondents from Myanmar, Vietnam, and other nations, respectively, also suggest their relatively high likelihood of being undecided about how long they intend to stay in Japan. In addition, the mean coefficients for residence statuses HSP and ESHI of −1.329 (95% CI: −2.686, −0.143) and −1.056 (95% CI: −2.078, −0.051), respectively, suggest a negative impact on being undecided about their stay intention; namely, these respondents have clearer intentions than the reference category.

The mean coefficients of 2.454 (95% CI: 1.239, 3.748) and 2.187 (95% CI: 0.858, 3.363) for respondents with a recent education acquired from a Japanese institution and those having already stayed in Japan for more than 5 years, respectively, suggest that they are more likely to fall into the undecided group.

Table 9. Bayesian multinomial logistic regression results for the undecided stay intention (group 4 vs. group 1).

Group 4 vs. Group 1	Mean	Std. Dev.	MCSE	Median	95% CI (Equal-Tailed)
Male	−0.897	0.416	0.032	−0.891	(−1.731, −0.105)
Age 26–35 years	2.338	0.468	0.055	2.331	(1.448, 3.309)
Age 36 years and above	1.500	0.731	0.079	1.514	(0.061, 2.899)
University degree or above	−1.322	0.671	0.085	−1.309	(−2.574, 0.024)
Nationality: Myanmar	4.453	0.675	0.087	4.408	(3.246, 5.838)
Nationality: Vietnam	3.559	0.490	0.047	3.569	(2.594, 4.504)
Nationality: Other	2.542	0.277	0.045	2.538	(2.009, 3.065)
Residence status: HSP	−1.329	0.657	0.094	−1.297	(−2.686, −0.143)
Residence status: ESHI	−1.056	0.506	0.042	−1.072	(−2.078, −0.051)
Residence status: TITP	−0.274	0.831	0.094	−0.262	(−1.933, 1.358)
Residence status: SSW	−0.718	0.747	0.081	−0.719	(−2.233, 0.741)
Residence status: Others	0.196	0.628	0.055	0.186	(−0.989, 1.463)
Recent education in Japan	2.454	0.679	0.069	2.430	(1.239, 3.748)
Stay in Japan >5 years	2.187	0.632	0.102	2.219	(0.858, 3.363)
_cons	−5.246	0.885	0.093	−5.201	(−7.023, −3.576)

Table 9. Bayesian multinomial logistic regression results for the undecided stay intention (group 4 vs. group 1).

Group 4 vs. Group 1	Mean	Std. Dev.	MCSE	Median	95% CI (Equal-Tailed)
Male	−0.897	0.416	0.032	−0.891	(−1.731, −0.105)
Age 26–35 years	2.338	0.468	0.055	2.331	(1.448, 3.309)
Age 36 years and above	1.500	0.731	0.079	1.514	(0.061, 2.899)
University degree or above	−1.322	0.671	0.085	−1.309	(−2.574, 0.024)
Nationality: Myanmar	4.453	0.675	0.087	4.408	(3.246, 5.838)
Nationality: Vietnam	3.559	0.490	0.047	3.569	(2.594, 4.504)
Nationality: Other	2.542	0.277	0.045	2.538	(2.009, 3.065)
Residence status: HSP	−1.329	0.657	0.094	−1.297	(−2.686, −0.143)
Residence status: ESHI	−1.056	0.506	0.042	−1.072	(−2.078, −0.051)
Residence status: TITP	−0.274	0.831	0.094	−0.262	(−1.933, 1.358)
Residence status: SSW	−0.718	0.747	0.081	−0.719	(−2.233, 0.741)
Residence status: Others	0.196	0.628	0.055	0.186	(−0.989, 1.463)
Recent education in Japan	2.454	0.679	0.069	2.430	(1.239, 3.748)
Stay in Japan >5 years	2.187	0.632	0.102	2.219	(0.858, 3.363)
_cons	−5.246	0.885	0.093	−5.201	(−7.023, −3.576)

5. Discussion and Limitation

This study provides new insights into the long-term settlement intentions of foreign workers in Japan, with findings that both align with and diverge from international patterns. In particular, we find that age (26–35 years), lower education levels, nationality (e.g., Myanmar and Vietnam), and certain residence statuses (TITP and SSW) are positively associated with the intention to stay long-term in Japan.

A comparative study on labor migrants in Germany (Ette et al., 2016) shows that while many migrants possess high human capital, permanent settlement intentions are driven more by socio-cultural and institutional factors—such as language skills, family context, and legal status—than by economic considerations. This aligns with our findings in Japan, where the residence status, local education, and duration of stay outweigh income or employment type in influencing long-term settlement. These parallels suggest that effective integration policies must address social and legal inclusion, not just labor market access.

Despite its contributions, this study has several limitations. The cross-sectional design limits causal inference, and the focus on Okayama Prefecture may affect the generalizability of findings. Moreover, socio-economic variables such as income and employment type were not included, though we used proxies such as education and residence status. Importantly, while the survey was conducted online, steps were taken to enhance data reliability, including in-person recruitment at community events, clear communication of the research objectives, and follow-up interviews with over 20 participants to validate and deepen the findings. However, this survey is conducted annually by the Hashimoto Foundation, and a second round is currently underway, incorporating additional socio-economic variables as well as cultural and institutional factors that influence long-term settlement intentions. These future data collections are expected to provide a more comprehensive understanding and allow for a longitudinal analysis in subsequent research. While this study offers critical insights into migrant settlement dynamics in Japan, its findings are limited in generalizability beyond the Okayama context.

Therefore, future research would consider longitudinal and multi-regional designs to capture temporal changes and diverse regional dynamics within Japan. Comparative studies between Japan and other migrant-receiving countries would also help to further clarify the institutional and cultural factors that shape migrants’ long-term settlement decisions.

6. Conclusions and Policy Recommendations

We analyzed the factors influencing long-term settlement intentions among foreign workers in Okayama Prefecture of Japan. Our findings highlight the significant roles of the age group, level of education, country of origin, type of residence status, and existing experience of living in Japan in shaping the decisions of foreign workers regarding long-term settlement in Japan. Specifically, residence statuses of TITP and SSW, experience of living in Japan for more than 5 years, and graduation from a Japanese education institution significantly increase the probability of intention to settle long-term. Among the age groups, respondents aged 26–35 were most likely to intend to stay in Japan long-term. Moreover, different nationalities have different intentions regarding long-term settlement, with respondents from low-income and lower-middle-income countries such as Myanmar and Vietnam more likely to intend settling in Japan long-term than those from China, an upper-middle-income and emerging economy.

The education level of foreign workers is another significant factor: respondents with higher education levels (university degree or above) have less intention to stay in Japan long-term than workers with school-level education. These results indicate that Japan is at risk of losing its educated foreign labor force in the future. Moreover, among the various residence statuses, respondents in the TITP and SSW program are more likely to intend settling in Japan long-term than those with other residence statuses such as Highly Skilled Professional, even though immigration rules do not permit workers in the TITP to extend their stay. The study highlights the effectiveness of Japan’s immigration policies, such as the TITP and SSW program, in attracting and retaining foreign labor, with challenges remaining in retaining highly skilled labor.

In line with Mosbah et al. (2020), we found that a long (>5 years) experience of staying in Japan is a crucial determinant of settlement intention. We consider that migrants who have already stayed in Japan for a long time are more likely to have integrated socially and become attached to the country, leading to a stronger intention to settle long-term, suggesting that policies fostering community engagement and emotional support may effectively enhance the settlement intention. However, segmentation and differentiation between foreigners and Japanese people may reduce their intention to settle in Japan.

Similar to the study of Cao (2019), our results indicate that economic factors, such as employment and income, have relatively little impact on the intention of long-term settlement. This may be due to the high level of economic development in Japan, which reduces the relative importance of economic considerations. Instead of those economic factors, access to social security such as pensions, insurance, and subsidies has a significant influence on long-term settlement.

In conclusion, we found that education, nationality, age, and residence status influence long-term settlement intentions. It is also confirmed that highly skilled workers are less likely to intend to stay long-term than low-skilled workers. Yang (2020) found that discrimination between native and foreign workers significantly affects the outmigration intention of highly skilled foreign workers from Japan. Therefore, by reducing discrimination against foreign workers in Japanese organizations and encouraging stronger community ties and emotional bonds with Japan, policymakers can enhance the long-term settlement intention of skilled migrant workers, thus helping address the labor shortage in Japan. Future research should further explore the dynamic interplay between social, economic, and psychological factors in shaping migrant settlement intentions.

Despite these contributions, our study has several limitations. The cross-sectional design limits the ability to infer causality, and our sample was limited to immigrants in Okayama Prefecture. Future studies should employ longitudinal designs and explore other contexts to validate and extend our findings.