Comparison of consumers’ beauty awareness in the pandemic
Based on Tf-idf value
Since the announcement of the first confirmed case of coronavirus on December 31st, 2019, social data including the keyword “beauty” have been collected. Keywords were distilled and centrality score and Tf-idf, which can complement simple frequency, were measured. The top 50 keywords were used in the analysis and the top 30 keywords according to the Tf-idf are reported in Table 2.
Table 2 Consumers’ cosmetic awareness before and after COVID-19 outbreak Before the COVID-19 pandemic, keywords such as “makeup” (Tf-idf = 369.68), “product” (315.68), “brand” (276.77), “busy” (215.22), “routine” (200.66), “industry” (142.08), “natural” (140.15), “fashion” (138.77), “trick” (133.73), and “skincare” (132.91) appeared in the upper ranks of consumers’ awareness. Although keywords such as “product” (410.57), “makeup” (389.54), “brand” (328.95), “routine” (221.11), and “skincare” (191.71) appeared in the upper ranks, keywords such as “coronavirus” (191.71) and “pandemic” (85.51) appeared in the infectious disease keywords. These results indicate that a pandemic impacts general consumers’ awareness of cosmetics. Makeup remains the most significant product that composes consumers’ cosmetic awareness, irrespective of the pandemic situation. In the case of skincare, there has been an increase in rank since the pandemic.
Analysis of network structure and centrality
We extracted the top 50 keywords based on the co-occurrence frequency and then clustered them using the Wakita-Tsurumi algorithm. Consumer awareness of cosmetic behavior before COVID-19 outbreak is illustrated in Fig. 2. Figure 3 represents the beauty behavior after the COVID-19 outbreak. In Fig. 3, new keywords in relation to COVID-19 have emerged after the outbreak. The comparison of the network size is as follows. The maximum geodesic distance (diameter) of the networks for 2019 and 2020 is 3, and the overall networks were similar in size because both were equally limited to the top 50 keywords. In 2019, the total number of edges of the network was 590, the average geodesic distance was 1.76, and the graph density was 0.24. The total number of edges for 2020 was 706, the average geodesic distance was 1.72, and the graph density was 0.28. Compared to 2019, the significant increase in the number of edges in 2020 means that the correlation between keywords has increased within the context. As a result, we can see that the average geodesic distance of the graph has decreased, and the graph density has increased.
For centrality measurement, in the case of makeup, both degree and betweenness centrality declined in consumers’ cosmetic awareness after, rather than before, the pandemic. On the other hand, skincare showed an increase in all centrality items (Cd, Cb, and Ce) except closeness centrality after the outbreak. On this basis, while the influence of makeup products has decreased, that of skincare products has increased. Moreover, in the results of Tf-idf and centrality measurement, keywords of color cosmetics, such as “lipstick,” “foundation,” and “eyeshadow,” which appeared as representative keywords, did not appear in the upper ranks after the outbreak.
Before COVID-19, skincare was classified in the compact groups such as “salon,” “versatile,” and “home,” but after the outbreak, skincare joined larger groups comprising “global,” “digital,” “clean,” and “technology,” which account for 28% of the whole group. Furthermore, before the pandemic, makeup was classified into groups containing the words “celebrity,” “fashion,” “Fenty,” and “tutorial.” After the pandemic, it was reclassified into groups containing “pandemic,” “change,” “face,” “treatment,” “skin,” and so on. During the pandemic, skincare was considered as cleansing to prevent coronavirus infection, and makeup was considered in terms of skin problems due to the pandemic situation.
According to the network analysis, the appearance of keywords related to coronavirus determined that pandemics affect consumers’ cosmetic awareness. After the outbreak, global consumers recognized skincare as extremely important. Makeup was also regarded as being important; however, its importance importance showed a declining trend.
Consumers’ awareness of skincare and makeup
The results of preliminary investigation revealed that skincare and makeup appear as the main cosmetic behaviors and products. We therefore conducted a comparison of consumers’ awareness based on the keywords before and after pandemic (Table 3). After the outbreak, “coronavirus” appeared as an upper ranked Tf-idf keyword in every cosmetic product. This highlights the fact that the COVID-19 pandemic impacts consumers’ cosmetic awareness.
Table 3 Consumers’ cosmetic awareness before and after COVID-19 The results of comparing representative keywords of individual cosmetic products before and after the pandemic are shown below. First, in the case of skincare, “self” (Tf-idf = 55.17), “mask” (47.81), “prevention” (47.50), “maskne” (46.92), “acne” (38.39), and “lockdown” (36.75) appeared as the representative keywords. In terms of consumer’s cosmetic awareness, this indicates the appearance of skin problems (or acne) caused due to wearing masks to prevent infection and the phenomena of self-care at home due to lockdown. The main products are serum (56.61/31.43), sunscreen (41.37/45.55), cream (37.61/28.58), and cleansers (24.29/45.30). Additionally, keywords such as “mask” (47.81) and “Maskne” (46.92) appear as representative keywords in cosmetic awareness, include skincare products for “acne” (38.39).
Second, in terms of consumers’ awareness on makeup, keywords such as “mask” (43.98), “lockdown” (32.80), “home” (26.39), “COVID-19” (26.39), “distance” (26.39), “maskne” (25.05), and “washing” (25.05) appeared after the outbreak. As with skincare, skin problems caused by wearing masks and cleansing the face appeared as the main interest after the spread of the disease. In addition, words related to social distancing and lockdown phenomena related to makeup started to appear. The representative products are “lipstick” (48.85/60.95), “eyeliner” (38.86/ 53.80), and “foundation” (35.07/45.26).
The analysis of consumers’ cosmetic awareness of skincare and makeup products before and after the outbreak of the COVID-19 pandemic revealed keywords related to infectious disease (coronavirus, mask, lockdown). Therefore, it can be argued that consumers’ awareness regarding individual cosmetic products has been affected by the pandemic.
Spread of the pandemic and consumers’ interest in cosmetic products
Model evaluation
Since the confirmed cases of COVID-19 and search volume of individual cosmetic products’ keywords are measured with a single measurement object, significance and suitability, such as convergent validity, multi-collinearity, outer weights, and outer loading do not need to be judged. The PLS-SEM model is assessed by R2, f2, and Q2 values (Shin, 2018). In investigation of consumer behaviors, if the coefficient of determination for the endogenous variable is larger than R2 = 0.20, it is judged to have a very high estimated suitability (Hair et al., 2017). The values of R2 in the endogenous variable are skincare products (R2serum = 0.31, R2sunscreen = 0.21, R2acne = 0.32, R2cleanser = 0.15, and R2cream = 0.25), makeup products (R2eyeliner = 0.21, R2foundation = 0.36, and R2lipstick = 0.27), and mask (R2 = 0.20). The values of R2 for acne, serum, and foundation are larger than R2 = 0.30. The majority of the other endogenous variables were larger than R2 = 0.20, indicating a high level of suitability.
If the effect size (f2) in PLS-SEM is greater than f2 = 0.02, it indicates a small effect size for latent exogenous variables on latent endogenous variables. If it is greater than f2 = 0.15, it indicates a medium effect size. If it is greater than f2 = 0.35, it indicates a large effect size. The interest in cosmetic products due to the spread of disease showed foundation as (f2 = 0.53), which has an effect size larger than f2 = 0.35 and it was identified as having the largest contribution. Serum (f2 = 0.17), acne (f2 = 0.25), lipstick (f2 = 0.28), and mask (f2 = 0.25) have medium effect sizes. Sunscreen (f2 = 0.09), cleanser (f2 = 0.07), and eyeliner (f2 = 0.04) have small effect sizes. Masks, cream (f2 = 0.21), foundation (f2 = 0.24), and lipstick (f2 = 0.26) showed medium effect sizes, and the others showed small effect sizes (f2serum = 0.08, f2sunscreen = 0.05, f2acne = 0.03, f2cleansing = 0.03, and f2eyeliner = 0.11). Next, by identifying the estimated suitability (Q2) through the blindfolding process, all Q2 values of the latent endogenous variables were larger than 0, indicating the suitability of the entire constitutive model.
Hypothesis testing
SOP and interest in cosmetic products
The impact of the spread of the COVID-19 pandemic on global consumers’ awareness regarding individual cosmetic products is as follows (Table 4; Fig. 4). The SOP had a positively influencing relationship with all skincare products: serum (β = 0.50, p < 0.001), sunscreen (β = 0.40, p < 0.001), cream (β = 0.33, p < 0.001), cleanser (β = 0.34, p < 0.001), and acne (β = 0.52, p < 0.001). Therefore, it was clear that as the number of confirmed cases increased, global consumers’ interest in skincare products also increased. This result statistically proves news reports and interviews about changes in cosmetic behaviors due to the pandemic (Cerullo, 2020; Wallace & CNN Business, 2020). Therefore, H1 (H1a, H1b, H1c, H1d, H1e) was supported.
Table 4 Direct path coefficients of SOP and interest in mask and individual cosmetic products The SOP had a negatively influencing relationship with lipstick (β = − 0.23, p < 0.001) and foundation (β = − 0.41, p < 0.001). This is statistical proof that the sales of makeup products have decreased (Cerullo, 2020; Halliday, 2020). This decrease is due to wearing masks or staying at home to prevent infection and spreading of the virus. Since people decide not to apply makeup to the area of the face covered by a mask does not need to have makeup applied, interest in makeup has naturally fallen. Meanwhile, eyeliner (β = 0.38, p < 0.001) showed a positively influencing relationship with regards to interest in the product. Opposite to the case of lipsticks and foundations, the eyes are still visible when wearing masks, so consumers tend to pay more attention to eye makeup, leading to the increased interest in eyeliners (see Table 4).
“Maskne” behavior manifests in that consumers tend to focus their efforts on the exposed area of their face. This is also similar to Muslim females’ cosmetic behavior, in which they pay more attention to eye makeup as they cover the rest of their face due to their religious ideology. In each case, the social setting forces people to hide their facial area, which leads to the limited application of makeup, that is, eye makeup. Hill et al. (2012) argued cosmetics such as lipsticks are a decorative product. In a pandemic situation, since a mask covers the nose, lips, and chin, eye makeup products are considered a decorative product. Therefore, H2c was accepted, and H2a and H2b were rejected.
Interest in face mask and cosmetic products
The SOP and consumers’ interest in masks were β = 0.11, p > 0.05, indicating that it does not have a significantly influencing relationship; H3 is therefore rejected (Table 5). However, the result of dividing the confirmed cases of COVID-19 by the timing of the spread was interesting. In the early phase from January to April 2020, it presented a high coefficient (β = 0.80, p < 0.001). In the next phase from May to July 2020, SOP also revealed a positive impact on masks (β = 0.57, p < 0.001), but it was lower than the previous phase. However, since the pandemic has been prolonged, the interest in masks has fallen and it seems to have levelled off. Consequently, the increase of confirmed cases may have influenced the relationship with the interest in masks. However, as the pandemic continues, it does not have a significantly influencing relationship; H3 was therefore accepted.
Table 5 Direct path coefficients of SOP and interest in mask according to three phases of COVID-19 pandemic Although the interest in face masks did not form significantly influencing relationships with serum, sunscreen, and cream, it showed a positively influencing relationship with acne (β = 0.15, p < 0.001) and cleansing (β = 0.11, p < 0.05). This is similar to the background of the appearance of the word “maskne,” which proves the interest in masks has a relationship with skin problems, such as acne, caused by wearing masks. On the other hand, cream (β = − 0.12, p < 0.05) had a significantly negative influencing relationship. Since “cream” can include moisturizing cream, BB cream, sun cream, and so on, it can be assumed that this affected the result.
Therefore, H4a, H4b, and H4c were rejected, and H4d and H4e were accepted. Meanwhile, interest in masks showed a negatively influencing relationship with lipstick (β = − 0.46, p < 0.001), foundation (β = − 0.36, p < 0.001), and eyeliner (β = − 0.25, p < 0.001). As consumers’ interest in masks increases, interest in makeup products for the entire face decreases. Therefore, H5 (H5a, H5b, H5c) was accepted.