Participants
Overall, 721 young adults, who participated for a minimum of three days in the five-day survey period, were included in the analyses (54.4% female, Mage = 26.05, SD = 3.53). Overall, 284 respondents participated on the first day (60.6% were missing), 402 on the second day (44.2% were missing), 659 on the third day (8.6% were missing), 648 on the fourth day (10.1% were missing), and 640 on the fifth day (11.2% were missing). On an average, 26.8% respondents were reported missing per day. The number of participants varied from day to day; however, all those who completed the survey any given day answered all the items. Therefore, there were no missing values in the data for each day that the respondents were present.
Among them, 38.8%, 19.2%, and 13.9% lived in the Kanto, Kansai, and Chubu regions (highly urban) of Japan, respectively, and the others lived in less urbanized regions. 16.9% were university students (50.0% female, Mage = 20.68, SD = 2.17), 39.8% were office workers (46.0% female, Mage = 27.42, SD = 2.36), 9.0% were homemakers (100% female, Mage = 27.95, SD = 1.84), and 18.3% were civil servants (63.6% female, Mage = 26.64, SD = 2.94), with 15.9% belonging to other categories (e.g., unemployed; 43.5% female, Mage = 26.56, SD = 3.09).
A total of 150 young adults (20.8% of the survey participants) responded to all of the surveys. A t-test was conducted to examine the presence of an age difference between participants who responded to all the complete survey versus participants who did not. However, no significant differences in age were identified [t(719) = 1.608, p = 0.108, d = 0.15)]. Furthermore, chi-square tests were conducted to examine whether there was a difference in the percentage of sex, occupation, and residential areas between those who responded to all the surveys and those who did not. The distribution did not differ significantly according to sex [χ2(1, N = 721) = 0.203, p = 0.652, Cramer’s V = 0.017, p = 0.652], social position [χ2(4, N = 721) = 4.815, p = 0.307, Cramer’s V = 0.082, p = 0.307] and residential area [χ2(1, N = 721) = 3.795, p = 0.051, Cramer’s V = 0.073, p = 0.051]. To examine the missing data pattern, Little’s (1988). Missing Completely at Random test was performed. The result was χ2(500) = 498.051, p = 0.516, indicating that the missing data pattern was likely to be at random. The full information maximum likelihood estimation was employed for the missing values. To observe deviations from the normal distribution for the study variables, the maximum likelihood robust estimation method was applied using Mplus 8.6 (Muthén and Muthén 1998–2021).
Procedure
Data were collected by an online survey company, Cross Marketing, Inc. (https://www.cross-m.co.jp/en/). People of various ages, professions, and regions are registered with this company, and their data can be gathered on request. The company regularly conducts quality control checks, including periodically checking for irregularities in registration, and only those who pass these checks are registered. In this study, the authors collected the data of registrants aged between 18 and 30 years. In addition, they requested that the distribution of respondents based on sex be equal. In March 2019, there were 144,264 registrants (67.8% female) who were contacted with the request. Finally, 721 Japanese young adults were included in this study. The authors sent the survey items to the company, the company created a survey form based on the items, and the authors verified the created form.
A web address for the survey was emailed to the registrants who had provided informed consent and agreed to participate in the study. The survey was conducted over a five-day period in a single week, with participants responding to the survey every day during the week of March 13, 2019 to March 20, 2019.
The participants were instructed to complete the survey using their smartphones between 6 pm and 12 am. They were paid an honorarium of approximately 50 yen (0.5 USD) per survey; 300 JPY (3 USD) was awarded to those who answered all the survey questions in the allotted time period.
Measures
Daily identity process
A single item was developed to measure each identity processes. First, five items, one for each dimension, were created in Japanese that corresponded to those of the Dimensions of Identity Development Scale (DIDS; Luyckx et al., 2008) and the single-item version of the Utrecht-Management Identity Commitment Scale (U-MICS; Klimstra et al., 2010). The items were created based on the following criteria: (1) the word “today” was to be added to the beginning of each item and (2) the wordings should be easy for Japanese young adults to understand. The following five items in English were translated by a Japanese-English bilingual individual: “Today, I had a clear view on my future” (commitment making), “Today, I felt confident about my future plans” (identification with commitment), “Today, I considered other possible lifestyles that may suit me better” (exploration in breadth), “Today, I worked out for myself if my life’s purpose really suits me” (exploration in depth), and “Today, I worried about my future plans” (ruminative exploration). These items were rated on a five-point Likert scale ranging from one (completely untrue) to five (completely true).
Daily emotions
A single measure was developed to examine the participants’ positive and negative emotions. For their positive emotions, items measuring life satisfaction and happiness were developed with reference to the Satisfaction with Life Scale (Diener et al., 1985) and the Subjective Happiness Scale (Lyubomirsky & Lepper, 1999). To assess their negative emotions, items measuring depressive moods were created with reference to the Daily Mood Scale (Hoeksma et al., 2000). The items were created based on the following criteria: (1) the word “today” was to be added to the beginning of each item, and (2) the wording should be easy for Japanese young adults to understand. These items were initially written in Japanese and were subsequently translated to English, with them being examined by a Japanese bilingual person who also spoke English. The items developed through this process included “Today, I am feeling satisfied” (life satisfaction), “Today, I am feeling happy” (happiness), and “Today, I am feeling down” (depression). The instruction for each item was as follows: “To what extent did you experience the following thoughts today?” These items were then rated by the participants on a five-point Likert scale from one (completely untrue) to five (completely true). For all analyses, a mean score across days for daily emotions was computed.
Statistical Analysis
As a preliminary analysis, factor structure was examined by confirmatory factor analysis (CFA). For the factor structure, the analysis was adapted from the study that examined the factor structure of a single-item version of the scale using data from a diary method (Becht et al., 2016a). A latent variable as a single item with a daily mean value was assumed (Fig. S1); that is, for each identity factor, the five daily single items were used as indicators of the respective factor. In addition, a measurement invariance test was conducted for sex. Three different levels were examined: configural (same number of factors and pattern of factor loadings across sex), metric (constraining the indicator factor loading to be equal across sex), and scalar (constraining the indicator factor loading and item intercepts to be equal across sex). For optimal model fit, the comparative fit index (CFI) should exceed 0.95, with values higher than 0.90 considered acceptable, and the root mean square error of approximation (RMSEA) should be less than 0.05, with values less than 0.08 representing reasonable fit (Kline, 2015), and the standardized root mean square residual (SRMR) should be less than 0.08, representing reasonable fit (Byrne, 2012). To test whether the fit of the model was equivalent across sex, we used the Satorra–Bentler χ2 difference test (S-B χ2) (Cheung & Rensvold, 2002), and differences in CFI (ΔCFI), RMSEA (ΔRMSEA), and SRMR (ΔSRMR) between the models. If the differences in model fit indices exceeded the following criteria, the null hypothesis of invariance was rejected: ΔCFI ≥ −0.010, ΔRMSEA ≥ 0.015, and ΔSRMR ≥ 0.030 (Cheung & Rensvold, 2002; Kline, 2015). If criteria are met the null hypothesis of invariance was rejected.
With respect to the reliability of the developed scale, the intraclass correlation coefficients were calculated (ICCs; Shrout & Fleiss, 1979). The ICC represents consistency within the measure. If the ICC value was greater than or equal to 0.70, test-retest reliability was considered to be sufficient (e.g., Landis & Koch, 1977). For the relationships between identity process and daily emotions, the correlation coefficients were calculated.
To achieve the first research objective, the correlation coefficients between identity processes and daily emotions were calculated. A mean score across days for identity processes and daily emotions was used. For the second research objective, the relationships between identity profiles and positive and negative emotions were examined. In this analysis, a mean score across days for identity processes and daily emotions was used. First, to identify the identity profiles, latent profile analysis was used (LPA; Lanza et al., 2003). LPA groups individuals on the basis of empirically distinct patterns of scores on the variables (i.e., identity processes). The continuous scores for each of the identity dimensions within each profile represent the measurement parameters, whereas the structural parameters refer to the profile membership probabilities assigned to groups of individuals (Nylund et al., 2007). The Sample Size Adjusted Bayesian Information Criterion (SSABIC), the bootstrapped likelihood ratio test (BLRT; Nylund et al., 2007), and entropy were used to determine the number of profiles. The lower the SSABIC value, the better the model fits the data; a significant BLRT score indicates that a model with k profiles fits better than that with k-1 profiles. Entropy ranges from 0.00 to 1.00, with values greater than or equal to 0.75 indicating accurate classification (Reinecke, 2006). Furthermore, if the number of participants in a profile is too small, it is difficult to replicate that profile. Hence, this study included the criterion that all profiles needed to include at least 5% of the participants (Muthén & Muthén, 2000). In addition to these criteria, the number of profiles was determined based on theoretical interpretation.
Second, to assess the relationships between identity profiles and daily emotions, the R3STEP command in Mplus was employed for membership prediction, using multinomial logistic regression analyses (Asparouhov & Muthén, 2014). This analysis assumes that daily emotions predict identity profiles. However, since this study simultaneously measures identity dimensions and daily emotions, it is not appropriate, nor is it the purpose of this study to refer to the direction of associations. Therefore, this study’s results are discussed without reference to the direction of associations between identity profiles and daily emotions.