Introduction

The COVID-19 pandemic has had significant impact in metropolitan areas where most infectious cases have occurred (Batty et al., 2022). In addition to the higher infection risk in metropolitan areas, restrictions of daily activities, such as social distancing, and business closures during the pandemic decreased the attractiveness of the areas (Florida et al., 2023). One of the long-term concerns regarding the changes in metropolitan areas was the ‘urban exodus’, reflecting the wide adoption of remote work during the pandemic, which decreased the physical commuting to working offices in the centres of metropolitan areas (Nathan & Overman, 2020). In addition to counter-urbanisation, that is the exodus, the changes in working arrangements could result in suburbanisation and urban sprawl (Denham, 2021; Li & Wei, 2023). Post-pandemic cities had many possibilities, including unpredictable components in some early predictions (Batty, 2020; Batty et al., 2022; Couclelis, 2020; Florida et al., 2023). An extreme example is that cities are expected to split into myriads of small towns and scattered settlements; otherwise, post-pandemic cities will not be much different from the pre-pandemic ones (Couclelis, 2020). By contrast, Florida et al. (2023) insisted that the broad macro-geographical pattern of urbanisation is expected to be unchanged, despite possible significant changes in the micro-geographical level such as intra-metropolitan and neighbourhood levels.

As a precursor of ‘urban exodus’, increased migration from metropolitan areas in the early phase of the pandemic were observed (Orford et al., 2023; Rowe, González-Leonardo & Champion, 2023). Despite available rich empirical evidence concerning the impact of the pandemic on internal migration in the first pandemic year, 2020, relatively little is known about the pandemic’s medium-term impact. According to forecasts on Australian regional net migration rates in 2021–2023 using time-series models (Borsellino et al., 2022), net gains in rural and provincial areas will slow down after 2022 but remain higher than the pre-pandemic level. Previous studies focusing on migration patterns in 2021 have mixed findings. In Spain, the macro patterns in internal migration returned to the pre-pandemic level while net out-migration in large cities, such as Madrid and Barcelona, continued in 2021 (González-Leonardo et al., 2022). In the United Kingdom, net migration patterns in 2021 were at an intermediate level, between those in pre- and early phases of the pandemic (Rowe, Calafiore et al., 2023), and they were getting back to the pre-pandemic stable and slow urbanisation trends (Wang et al., 2022). By contrast, the proportion of migration from metropolitan areas to the suburbs and rural areas in the United States increased in 2020 and remained the same in 2021 (Haslag & Weagley, 2022). In Serbia, net-migration rates in rural areas increased in 2020 and surpassed the pre-pandemic level in 2021 (Lukic et al., 2023). Few studies have extended their observations to the first half of 2022 in Argentina, Chile, and Mexico, reporting that negative and positive balances of short- and long distance (under and over 100 km, respectively) migration in the capital cities continued in 2020–2022 (González-Leonardo et al., 2024; Rowe, Cabrera-Arnau et al., 2023). Therefore, it is necessary to provide other cases to advance the understanding of the COVID-19 pandemic’s impact on internal migration.

To comprehend the pandemic’s medium-term impact on internal migration, this study aimed to investigate the annual series of migration patterns from 2019 to 2023 in Japan. The annual migration data during this period can be used to investigate whether the pandemic’s impact on internal migration was temporary. Using annual inter-municipal migration data, this study captures the changes in internal migration to/from and within metropolitan areas in Japan from 2019 to 2023.

Japanese context

In 2020, internal migration patterns in Japan and other countries changed. The urbanisation trend, especially net in-migration into the Tokyo metropolitan area from the rest of Japan since the late 1990s, became weak (Fielding & Ishikawa, 2021). The pandemic had a strong impact on migration to/from and within the Tokyo metropolitan area and a minimal one on migration among other areas (Kotsubo & Nakaya, 2023). Some studies point out that the ‘urban exodus’ comprised net out-migration from the centre of the Tokyo metropolitan area (Fukuda, 2024) and total population decrease in the centre of the Osaka metropolitan area (Kato & Takizawa, 2022). Some of the major destinations of migration from the centre of the Tokyo metropolitan area were identified as its suburbs (Koike, 2022; Kotsubo & Nakaya, 2023) as well as rural and provincial areas (Fielding & Ishikawa, 2021; Kotsubo & Nakaya, 2022).

The changes in migration patterns during the pandemic have influenced people’s daily lives. Following surveys conducted by the central government ministries and agencies of Japan, the pandemic has altered people’s daily lives significantly, but the duration of the alterations was dependent on the types of activities. One of the significant changes in Japan is wide spread remote work, as the share of remote workers increased from 10.3% as of December 2019 to 27.7% as of May 2020 (Cabinet Office, 2023). In the survey results, the share temporally decreased to 21.5% as of December 2020, then increased again to 30.8% as of April and May 2021; it was the highest at 32.2% as of September and October 2021. Finally, the share slightly decreased to 30.6% as of June 2022 and 30.0% as of March 2023. Another survey showed the same trend in share of employees who work remotely: 14.8, 23.0, 27.0, 26.1 and 24.8% in 2019, 2020, 2021, 2022 and 2023, respectively (Ministry of Land, Infrastructure, Transport and Tourism, 2024). These trends suggest that the working arrangements significantly changed in the early stage of the pandemic. Additionally, shares of remote workers differed regionally, and these differences continued during the pandemic. For instance, the shares of remote workers of the Tokyo, Osaka, and Nagoya metropolitan areas in 2021 were 42.1, 27.3, and 17.7%, respectively (Ministry of Land, Infrastructure, Transport and Tourism, 2024).

In contrast to the share of remote workers, the patterns of people’s daily activities reverted to the pre-pandemic ones in 2022. The shares of people going out for unessential purposes significantly decreased from 61.8% as of the pre-pandemic period to 13.1% as of April 2020 for leisure activities and from 81.7% to 33.7% for eating out alone (Ministry of Land, Infrastructure, Transport and Tourism, 2023). While these shares decreased during the pandemic, they recovered to the pre-pandemic level of 60.3% as of December 2022 for leisure activities and 83.3% for eating out alone. These findings suggest that the daily activities were significantly affected by the pandemic; however, its impact was temporary. In addition, in May 2023, the Japanese government (Ministry of Health, Labour and Welfare, 2023) and the World Health Organization (World Health Organization, 2023) declared that they would no longer request the public to take unified basic infection control measures.

Some studies, conducting surveys about residential preferences before and after the pandemic in Japan, reported increased preferences to live in the suburbs of Tokyo Metropolis (Sharifi & Lee, 2024) and Toyota City (Tsuboi et al., 2021). Additionally, the survey, which focused on the people who moved from the centres of Tokyo and Osaka metropolitan areas to other areas during the pandemic, indicated that the priority of working, among other reasons related to community, environment, and housing, reduced for all respondents, but remained the highest in single-person households (Komaki et al., 2023).

In the Japanese context before, during, and after the pandemic, the recovery of daily activities would contribute to the attractiveness of metropolitan areas, which had decreased during the pandemic. Considering that working at offices and face-to-face communication are preferable in the Japanese culture (Ono, 2022), the impact of the COVID-19 pandemic on the macro internal migration patterns in 2020 may have reduced even when remote work became a common option in the 2023 post-pandemic period. Therefore, the impact of the pandemic in metropolitan areas may be hypothesized to have also reduced in 2023.

Methods

Data

The source for the annual inter-municipal migration data is the Annual Report on the Internal Migration in Japan derived from the Basic Resident Registers, which is published by the Statistics Bureau in the Ministry of Internal Affairs and Communications. The number of migrants is collected from 1 January to 31 December in each year. While numbers of in- and out-migrants at the municipal scale (1,896 municipalities in 2023) are available, the small number of migrants in origin–destination matrix are suppressed. Table 1 shows the proportion of suppressed migrants for the period 2019–2023.

Table 1 Proportion of suppressed inter-municipal migrants
Full size table

Proportion of suppressed migrants in the origin–destination matrix slightly increased from 23.1% in 2019 to 24.1% in 2023. To compensate for the suppressed inter-municipal migration flows, the total out-migrants from 47 prefectures and 21 major cities (Tokyo special wards and 20 of ordinance-designated cities) to each municipality are available.

To use the origin–destination matrix at the municipal scale, this study estimated the suppressions in the matrix for each year from 2019 to 2023 following Kotsubo and Nakaya (2023). First, the origin–destination matrix was divided into 68 regions as rows corresponding to the 21 major cities and 47 prefectures, excluding these cities, where the total numbers of migrants from each region to each municipality were available. Second, suppressed parts in the matrix were identified using the published migration flows in the matrix, total out-migrants by municipalities, and total in-migrants to each municipality by 68 regions. Third, the suppressions were estimated for each region using the iterative proportional fitting. In the estimation, the marginal row totals were defined as the differences between total out-migrants by municipalities and published number of out-migrants in the matrix. The marginal column totals were defined as the differences between total in-migrants to each municipality from 68 regions and published number of in-migrants in the matrix. The seeds in the matrix and threshold value to end the iteration were set as same as Kotsubo and Nakaya (2023). Finally, we obtain the origin–destination matrix of 1,896 municipalities, including the estimated migrants of suppression parts in each year. The obtained matrices cover inter-municipal migrants of Japanese people and other nationalities in the period 2019–2023.

The source for annual municipal population data is the Basic Resident Register. These municipal population were as of 1 January each year. The municipalities areas are as in the Report of Statistical reports on the land area by prefectures and municipalities in Japan, published by Geospatial Information Authority of Japan. These annual data were as of 1 October each year.

Investigation of migration patterns

To investigate overall patterns in internal migration in Japan, this study first examined the relationship between net migration rates and (the natural logarithm of) population density following the previous studies (Fielding & Ishikawa, 2021; Stawarz et al., 2022; Wang et al., 2022). Population density is used as a proxy for urbanicity (Rees et al., 2017). The positive and negative slopes of the regression line indicate urbanisation and counter-urbanisation, respectively. This study estimated the slope coefficients at the municipal scale each year for the period 2019–2023.

To elucidate the changes in internal migration in metropolitan areas, this study then investigates flows to/from and within the Tokyo, Nagoya, and Osaka metropolitan areas in Japan. Japanese migration studies have increasingly paid attention to these three major metropolitan areas for a long time (Fielding, 2018; Ishikawa, 2020). The ‘urban exodus’ of the Tokyo and Osaka metropolitan areas were pointed out based on the population loss in their centres (Fukuda, 2024; Kato & Takizawa, 2022). However, the population and net migration data they used does not explain people’s to and from movements. To overcome this limitation, this study aggregated inter-municipal migration flows based on the centre-suburbs in the metropolitan areas and the core-periphery dichotomy, such as the three major metropolitan areas and others (Inoue et al., 2022), in Japan.

In this study, metropolitan areas at the municipal scale are established based on the Urban Employment Area (UEA) proposed by Kanemoto and Tokuoka (2002). The UEA defines metropolitan areas as functional regions based on the commuting flows: the centres are densely populated areas and the municipalities where more than 10% of workers commute to the centres are defined as their suburbs. The definition allows the metropolitan area to have multiple centres. As shown in Fig. 1, the three major metropolitan areas with their centres and suburbs were defined using the UEA based on the 2015 census to provide comparable results to the previous studies (Kotsubo & Nakaya, 2023). Table 2 shows population in the three major metropolitan areas in 2020.

Fig. 1
figure 1

Three major metropolitan areas in Japan

Full size image
Table 2 Population in the three major metropolitan areas
Full size table

The population of the three major metropolitan areas accounted for 43.3% of the national population (55,033,523 of 1,127,138,033 people) in 2020.

Focusing on each of the three major metropolitan areas, this study aggregated annual inter-municipal flows into six flows: intra-migration flow from the centre to suburbs, its counter, out-migration flows from the centre or suburbs to the outsides of metropolitan area, and in-migration flows from the outsides to the centre or suburbs. To measure changes in sizes of migration flows, we first calculated the percent change in the number of migrants relative to that in 2019, as the pre-pandemic baseline:

$${Change}_{t}=\frac{{N}_{t}-{N}_{2019}}{{N}_{2019}}\times 100\left(\%\right)$$
(1)

where \({N}_{t}\) is the number of migrants in year \(t\). The changes were calculated for each migration flow of six types. To investigate the demographic impact of migration, we then calculated migration effectiveness index (MEI) as:

$${MEI}_{ij}=\frac{{M}_{ji}-{M}_{ij}}{{M}_{ji}+{M}_{ij}}\times 100\left(\%\right)$$
(2)

where \({M}_{ji}\) is a migration flow from area \(j\) to area \(i\). For a systematic analysis of migration patterns, the MEI is superior to the conventional net migration rate, which is a function of migration and population, because the index comprises net and gross migration and is not influenced by the cumulative population history (Plane, 1984). The MEI takes values from -100 to 100. The positive value indicates net gain of area \(i\) from the migration flows and the negative value indicates net loss. The absolute values of MEI show directional patterns that 100 and 0 indicate unidirectional and balanced flows between \(i\) and \(j\), respectively.

Results

Overall migration patterns

Figure 2 shows the relationships between municipal net migration rates and population density for the period 2019–2023.

Fig. 2
figure 2

Relationship between net migration rates and population density at the municipal scale

Full size image

The slope of regression line in 2019, as the pre-pandemic period, was 0.199 (standard error: SE = 1.02 \(\times {10}^{-2}\), p < 0.001). The slope decreased to 0.176 (SE = 7.49 \(\times {10}^{-3}\), p < 0.001) in 2020. The slope then decreased to 0.145 (SE = 7.08 \(\times {10}^{-3}\), p < 0.001) in 2021 and slightly increased to 0.160 (SE = 8.87 \(\times {10}^{-3}\), p < 0.001) in 2022. In 2023, as the post-pandemic period, the slope decreased to 0.182 (SE = 9.29 \(\times {10}^{-3}\), p < 0.001). Overall, the estimated slopes indicated positive relationships for all years, but the values of the slopes fluctuated in the study period. The slope decreased in the pandemic years, 2020–2022, and then reverted to the level of first year of pandemic in 2023. The minimum and maximum values were observed in 2021 and 2019.

Migration patterns in the three major metropolitan areas

Figure 3 shows changes in sizes of migration flows from/to and within the three major metropolitan areas relative to those in 2019 (Fig. 5 in Appendix shows sizes of migration flows).

Fig. 3
figure 3

Changes in sizes of migration flows. a the Tokyo metropolitan area, b the Nagoya metropolitan area, c the Osaka metropolitan area

Full size image

In the Tokyo metropolitan area, migration from the centre to the outside increased in the period 2020–2022, compared to the level in 2019 (Fig. 3a). By contrast, migration from the outside to the centre and suburbs decreased during the pandemic. Migration from the centre to suburbs increased and its counter flow decreased during the pandemic. The out-migration from the suburbs was nearly the same level in 2020 and 2021 as in 2019, but it sharply increased in 2022. In 2023, the migration patterns seemingly recovered, with almost zero changes in migration sizes.

In the Nagoya metropolitan area, changes in migration sizes among the centre, suburbs, and outside had negative values in 2020 and 2021, indicating the decreases in size (Fig. 3b). In 2022, migration from the centre and suburbs to the outside increased whereas other flows remained in the decreased levels. The patterns were the same in 2023.

In the Osaka metropolitan area, migration from the outside to the centre, that from the centre to suburbs, and its counter remained in the same level as in 2019, whereas the other levels decreased in 2020. Migration from the suburbs to the outside and its counter remained in the decreased level after 2020. Changes in the migration size from the outside to the centre was positive but close to 0% in 2021; then, they increased in 2022 to almost 10%, the highest in the period, in 2023. Migration from the centre to suburbs increased in 2021 and 2022 and remained at a slightly high level in 2023, compared to the levels in 2019 and 2020.

The above changes in the migration size indicates that the COVID-19 pandemic seemingly influenced migration patterns in the cases of the three major metropolitan areas in Japan. The changes in migration size among the three major metropolitan areas during and after the pandemic had some similarities and differences. A similar point across the three areas was the decrease in migration from the outside to the centres during the pandemic. In the Nagoya and Osaka metropolitan areas, the year with largest absolute values of changes in migration size varied depending on the migration directions. In the Tokyo metropolitan area, the absolute values of changes, excluding that in migration from the suburbs to the outside, was the highest in 2021. While the recovery of the changes in the size of migration from the centres to the outside was observed for the three areas in 2023, the almost values of change remained at the increased or decreased levels during the pandemic in the Nagoya and Osaka metropolitan areas.

Figure 4 shows changes in migration effectiveness index of migration flows within and from/to the three metropolitan areas.

Fig. 4
figure 4

Migration effectiveness index. a the Tokyo metropolitan area, b Nagoya metropolitan area, c Osaka metropolitan area

Full size image

In the Tokyo metropolitan area, three migration effectiveness indices had similar characteristics of temporal changes (Fig. 4a). The indices decreased during the pandemic period and were at their lowest values in 2021. In 2023, the indices reverted close to the level in 2019. The indices of migration from the centre and suburbs to the outside took positive values in the period indicating net in-migration in the Tokyo metropolitan area. The negative values of index of migration from centre to suburbs showed the suburbanisation trends in the period.

In the Nagoya metropolitan area, the indices slightly decreased during the pandemic period; however, these changes were relatively small compared to those in the Tokyo metropolitan area (Fig. 4b). The indices of migration from the centre were nearly 0% and that of the migration from the suburbs to the outside took values between -4.11% in 2020 to -1.92% in 2019.

In the Osaka metropolitan area, the index of migration from the centre to the outside increased in 2020 and then remained at the same level as in 2019 (Fig. 4c). The index of migration from the suburbs to the outside took negative values and hardly changed in the period. By contrast, the index of the migration from the centre to suburbs changed from positive to negative in 2021 and then took the positive value in 2023.

The changes indicate that the COVID-19 pandemic had the largest impact in the Tokyo metropolitan area. The indices of migration within and from/to the Tokyo metropolitan areas had the same trends. In the Osaka metropolitan area, the indices of migration from the centre changed in 2020 or 2021; however, the index of migration from the suburbs had little change. In the Nagoya metropolitan area, the changes in indices were relatively small compared to those in the other areas. Similar changes occurred among the three areas where the pandemic decreased the indices of the migration from the centres to the suburbs as suburbanisation trends. Additionally, in 2023, the indices of migration within the metropolitan areas reverted to the same level in 2019.

Discussion

The impact of the COVID-19 pandemic on internal migration in Japan

The regression slope between municipal net-migration rates and population density showed the urbanisation patterns in pre-, peri-, and post pandemic periods in Japan. The changes of slope indicated that the COVID-19 pandemic influenced macro internal migration patterns as a kind of weakened urbanisation trend. However, the impact might be relatively small compared to that in the United Kingdom, where the slope changed from positive in 2019 to negative in 2020 and then reversed to positive in 2021 (Wang et al., 2022). The changes of slope suggested that the impact was the largest in 2021, reducing in Japan in 2023.

Changes in the number of in- and out-migrants in the centre of the Tokyo metropolitan area were consistent with the change of regression slope. The impact was the largest in 2021, and the levels recovered to the pre-pandemic ones. Owing to the high population share (Table 1) and large number of migration to/from and within the Tokyo metropolitan area (Fig. 5), the patterns in the area contributed to the macro migration trends in Japan, as shown in the regression slope. The decrease in in-migration could reflect the increased number of remote workers in Japan, which peaked in 2021 and then slightly decreased (Ministry of Land, Infrastructure, Transport and Tourism, 2024). The changes in the Nagoya metropolitan area indicated the decrease in the number of migrants within and to/from the area during the pandemic. Although this is not normalised by annual population as migration intensity, the decrease could be the same changes at national scales in developed countries, considered as the impact of the pandemic (Rowe, González-Leonardo & Champion, 2023). The changes in the Osaka metropolitan area showed that, in 2020, the pandemic had minimum impact on migration within the area, but a large impact on migration from/to the area. Assuming that remote work is among the drivers of the movement from the centres to suburbs (Sharifi & Lee, 2024; Tsuboi et al., 2021), the differences among the three major metropolitan areas might be caused by the regional differences in the numbers of remote workers—whether people who live and work in the areas can choose remote work arrangements.

The changes in number of migrants in the three metropolitan areas indicated that the pandemic induced the decrease in in-migration rather than increase in out-migration. Besides, the migration from the centres to suburbs increased during the pandemic in the Tokyo and Osaka metropolitan areas. These changes suggest that the pandemic stimulated the intra-metropolitan migration as suburbanisation rather than net out-migration as the ‘urban exodus’, as pointed out in previous studies (Fukuda, 2024; Kato & Takizawa, 2022). The migration effectiveness index also showed that the pandemic affected intra-metropolitan migration in the three areas. The decreased indices during the pandemic can be interpreted as the pandemic induced suburbanisation. Besides, the positive indices of migration from the centres and suburbs to the outside indicate that the Tokyo metropolitan area continued to receive net in-migration following the previous trends (Fielding, 2018; Ishikawa, 2020). The changes in indices in the Nagoya and Osaka metropolitan areas indicate that the net in- or out-migration patterns did not change during the pandemic period. Therefore, the overall results revealed that the pandemic increased intra-metropolitan migration (suburbanisation). These could be consistent with the hypothesis of previous studies that a macro-geographical pattern is expected to persist, and that significant changes in micro-geographical level might occur (Florida et al., 2023).

Medium-term impact of the pandemic

The changes in migration effectiveness index in the three major metropolitan areas demonstrate that the COVID-19 pandemic had the largest impact in the Tokyo metropolitan area in 2021, which expands the findings in Japan that the pandemic strongly affected migration from/to and within the Tokyo area (Kotsubo & Nakaya, 2023). The changes in indices may reflect the trend in the share of remote workers in Japan, similar to the case of changes in the number of migrants. Furthermore, the changes suggest reduced impacts of the pandemic on migration to/from and within the Tokyo metropolitan area, which is similar to implications of changes in regression slopes. Previous studies in other countries have shown the recovery to the pre-pandemic level (González-Leonardo et al., 2022; Wang et al., 2022) or intermediate level between the pre- and pandemic periods (González-Leonardo et al., 2024; Rowe, Cabrera-Arnau et al., 2023a; Rowe, Calafiore et al., 2023b), as well as the remaining changes (Haslag & Weagley, 2022; Lukic et al., 2023) in 2021. Although the share of remote worker in the Tokyo metropolitan area was relatively high in Japan, it was lower than that in other global cities. For instance, the shares of workers who adopt remote work once and more per week was 86% in central London (Swinney, 2023) and 37.1% (51.6% including workers who occasionally adopt remote work) in Tokyo special wards (Cabinet Office, 2023). Another difference was that Japan had a soft (non-binding) lockdown during the pandemic, meaning that the public transport system worked as normal, and people were allowed to commute (Okubo, 2022). Notably, the Japanese culture favours working at offices and face-to-face communication (Ono, 2022). In addition, COVID-19-caused mortality rate in Japan was low compared to those in European and American countries (Mathieu et al., 2024). These differences may be attributed to the timing of the largest impact and its recovery compared to the other countries.

The recovery of the migration effectiveness indices in the Tokyo metropolitan area and the reduced impact on macro patterns in 2023 suggested the need to re-think the regional policies in Japan. In 2020, the pandemic weakened the net in-migration into the Tokyo metropolitan area (Fielding & Ishikawa, 2021; Kotsubo & Nakaya, 2023). This change was one of the goals of national government policies, including regional revitalization under the national population decline (Dilley et al., 2022, 2024; Saito, 2021). With the wide adoption of remote work and changes in residential preferences induced by the pandemic, the Government of Japan aimed to build digital infrastructure in rural and provincial areas in the ‘Vision for a Digital Garden City Nation’, which is regarded as one of the mitigations of ‘mono-polar concentration’ in the Tokyo metropolitan area (Cabinet Office, 2022). However, the results indicate that the pandemic influenced intra-metropolitan migration and the migration patterns were recovering to the pre-pandemic ones, that is urbanisation and ‘mono-polar concentration’. These could re-emphasise the importance of traditional factors of the urbanisation such as regional disparities of income level and working opportunities between the Tokyo metropolitan area and others (Fielding, 2018; Ishikawa, 2020).

Conclusions

This study investigated the medium-term impact of the COVID-19 pandemic on internal migration in Japan using annual series of migration data for the period 2019–2023. The results showed that the pandemic had the largest impact in 2021 based on the macro migration patterns. These impacts were observed in the three major metropolitan areas, Tokyo, Nagoya and Osaka, with the largest impact observed in the Tokyo metropolitan area, which is consistent with Kotsubo and Nakaya (2023). These changes could reflect the wide adoption of remote work during the pandemic. By aggregate the inter-municipal migration flows from/to and within the metropolitan areas, the results revealed that the pandemic stimulated intra-metropolitan migration as suburbanisation rather than the net out-migration as ‘urban exodus.’ The macro migration patterns and the migration effectiveness indices in the Tokyo metropolitan area suggested that the impact of pandemic was confirmed in 2020–2022; in 2023, the migration patterns were recovering to the pre-pandemic ones.

One of the limitations of this study is that the reported migration patterns were aggregated by age and sex. In addition to age and sex differences (Lukic et al., 2023; Perales & Bernard, 2023), some studies showed the differences in migration patterns during the pandemic based on individual characteristics (Tønnessen, 2021; Vogiazides & Kawalerowicz, 2023). To understand internal migration patterns in Japan better, future studies should address these issues by estimating inter-municipal migration data by age and sex or using microdata, including the socio-economic characteristics of people. Another limitation is that this study focused on only the three major metropolitan areas. The framework of migration system theory (Fawcett, 1989), which posits that when one place experiences a change, the effect is manifested throughout the system, could be applicable to investigate the changes in migration patterns related to the impact of the pandemic, such as forced migration caused by the disasters. The analysis, which used all origin–destination combination based on the theory (Curtis et al., 2015; Fussell et al., 2014; Hauer et al., 2020), could reveal how municipalities contribute to the changes in migration patterns in the Tokyo metropolitan area and identify large changes in the rural and provincial areas.