Female engagement in STEM

Until the early 2000s, there was a culture in which women’s participation in the STEM sector was limited and socially reluctant. However, as perceptions of gender equality spread, women’s entry into the STEM fields has been increasing as the gap of their status and education opportunities to those of men is decreasing [9].

Conversely, gender segregation depending on the field of science discipline has become an issue. Recent research confirms the existence of varying levels of gender segregation in distinct fields of study ultimately disadvantaging women obtaining authority positions [10, 11]. Insights into the gender hierarchy within professions of academic science can be gained by understanding how scientists perceive different degrees of gender segregation among various fields of science. Prior research confirms the prevalence of the perceived gender role stereotypes in research disciplines in traditionally male-dominant businesses [12].

Mentors and role models discreetly dissuade a woman from sticking with a particular discipline if scientists themselves believe that one is more difficult for women to succeed in than another, which influences women’s thinking as they enter a particular science discipline [13]. Thus, many women may be drawn to the feminizing field of cosmetic science. Therefore, relatively active female participation can be expected in such research domains.

Women’s contribution in patenting activity

The trend of growing participation of females has been observed not only in the education or professions in STEM but also in the filing activities of intellectual property rights. In patenting activities, women’s contributions have been increasing in most countries in both the United States Patent and Trademark Office (USPTO) and the Patent Cooperation Treaty (PCT) applications [14, 15]. Specifically, Frietsch et al. [16] suggested that women’s contributions have increased in all countries between 1991 and 2005. In particular, women’s contribution in Spain was highest between 2003 and 2005. During 2011–2015, China, South Korea and Singapore accounted for the highest proportion of women’s participation in patent inventions, while Germany, Japan and the United Kingdom accounted for the poorest countries [15]. However, still, the patents applied at the USPTO and PCT during late 1990s to early 2010s had shown that the patent application acceptance of female was lower than that of male applicant [8, 14, 15, 17, 18]. In addition, the grant rate of female patent was much lower [8, 17] and approved number of claims was also smaller than that of male patents [8].

Recently, women are increasingly likely to patent in a large, gender-mixed inventor teams [17], highlighting the growing importance of exploration of innovative mixed gender collaboration. In particular, the trend of biotechnology has been shifting from invention by one gender to co-invention [19]. In addition, among the type of the assignee, such as university, company, and individual, patents from university had the highest female participation [14]. The proportions of female inventorships in university are relatively superior across all macro technological areas, with the exception of areas such as games, furniture, and other consumer goods [14]. Hence, investigating the changes in women’s participation in technology development may provide an understanding of the patenting trends of women inventors and gender disparities of patents.

Differences in patenting fields by gender

There exist some differences on patenting activities by gender. Among various technological fields, women inventors turn out to be largely focused on life sciences fields, such as biotechnology, pharmaceuticals, and agriculture [20]. On the contrary, several fields, such as engineering, physical sciences, mathematics, and transport, have low share of female inventions [21]. Specifically, Frietsch et al. [16] addressed that the women’s contribution had largest increase from 1991 to 2005 in chemistry field among the five large technical areas (Chemistry, Electrical, Instruments, Mechanical, Other). Similarly, among the patents filed in 2010, the proportion of patents including female inventors was the highest at 50% in technology class “Chemistry: Natural Resins or Derivatives” [4]. Subsequently, “Organic Compounds Drug”, and “Bio-Affecting and Body Treating Compositions” accounted for a high percentage. Meanwhile, in 2015, biotechnology (BT) (57.6%), pharmaceuticals (55.5%) and organic chemistry (50.7%) dominate other fields of technology on which women inventors participated, whereas the machinery sector is recorded to have lowest share among the fields [15]. This suggests that there exist the differences in the participation of female inventors by technological sectors.

As such, the female-majority teams are more likely to invent technologies for women, especially in the health and chemistry areas [4, 5]. Considering the much greater number of male inventors in overall patents, the technologies related to women’s health can be represented as more balanced fields in terms of the number of patents than other fields, overcoming the difference in the total number of inventions between women and men. Especially, the cosmetics industry was traditionally focused on women, which can be expected to be one of the gender-balanced technological field. These gender-balanced fields may allow fair comparison of technological trends or innovation performance by gender.

Recently, cosmetics is no longer a preserve of women, which is shown by the increasing spendings on cosmetics of male [22]. The gender differences lead to different motivations of using cosmetics and consumption behaviors. However, existing studies on cosmetics lack the exploration of innovation characteristics and performance by gender diversity of inventor groups. Hence, research in cosmetics can provide insightful implications on the gender differences and the synergistic effects of gender diversity in patenting activities.

Female participation in cosmetics patent application / granted

Fig 3 shows the annual change in patent application rate by gender of cosmetics patent inventor. Until the early 1990s, most of the patents were applied only by male inventors. However, since then, the proportion of mixed patents has risen sharply, and in 2009 it overwhelmed patents with only male inventors. The proportion of patents with female inventors have also risen steadily and have remained at 10% since 2012.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Annual change in cosmetics patent application rate by the gender of patent inventors.

https://doi.org/10.1371/journal.pone.0305238.g003

Fig 4 shows the proportion of patent application by gender of inventors during 1970–2016. The patents invented by a group of mixed gender had the highest proportion of patent applications compared to those invented by single gender groups. Within a single gender group of inventors, women tend to invent alone rather than collaborators compared to men. Patents invented by single female inventor was 10% of the total patents with participation of female inventors (which exceeds that of multiple female inventors), whereas the proportion of patent applications with a single male inventor is lower than that of multiple male inventors. When calculating the average number of inventors of at least two inventors, a patent with male only has a value of 2.93, a patent with female only of 2.38, and a mixed patent of 3.89. These indicate that the patents invented by female-only groups are less likely to be associated with collaborative activities and tend to be usually developed by sole inventors. On the other hand, in mixed patents where women and men collaborate, cooperation is occurred more as a larger team.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Proportion of patent application by the gender of patent inventors.

https://doi.org/10.1371/journal.pone.0305238.g004

Table 1 shows the difference in patent grant according to the gender of the patent inventor. Grant lag refers to the period of time between the filing date of patent application and the date of its grant, which is shown to have inverse relation with patent value [25]. The table shows the highest rate of grant assigned to patents developed by male. On the other hand, the rate of grant of patents by female was the smallest. The chi-squared test indicates there are significant differences on the distribution of a categorical variable, that is, if patent is granted or not, between the inventor groups. Overall, the grant rate has significant differences by group, and decreases with the participation of women in patent inventions. We also observed differences in grant lag by gender for granted patents. The grant lag of a patent is influenced by the nature of the patent and generally averages 28 months [26]. In contrast, the average grant lag of cosmetic technology patents is observed at 35 months. The average grant lag of male, female, mixed was 32.62, 35.05, and 37.21, respectively. The analysis of variance (ANOVA) test showed at least one pair of groups has statistically significant difference. Then, we conduct post-hoc test, specifically the Games Howell test, considering that the size of sample is different, and the homogeneity of variance is not satisfied, which is proven by Levene test [27]. The post-hoc test showed every pair of groups had statistically significant differences. The inventions with the female-invented patents had significantly longer grant lag than male inventions. Surprisingly, it took the longest time for mixed patents to be granted. Overall, women’s inventions not only had a low grant rate, but also took a long time to be granted than men’s inventions.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Comparison of grant information by patent gender.

https://doi.org/10.1371/journal.pone.0305238.t001

Technological characteristics by gender

We observed how bibliographic information, which stands out the technical characteristics of patents, differs by gender. First, claims, backward citation, and family size were chosen as comparison factors. The claims in a patent have separate legal effects and expresses multiple technical features within an invention, which indicate the inventiveness and the scope of protection of patents [28, 29]. Tong and Frame [30] insisted that the number of patent claims is a better indicator of technological performance than just patent counts. As the patent fee generally relies on the number of claims, the claim counts reflect the financial value of the invention [31–33]. Backward citation notes the earlier patents that a given patent cited [28, 34–36]. Inventions containing fewer backward citations indicate higher level of novelty in their features or functionalities (i.e., novelty in application), suggesting previous inventions have less bearing on the invention [37]. A patent family is “the same invention disclosed by a common inventor and patented in more than one country.” The family size represents the number of jurisdictions that the patent has been sought for, which indicates the solid legal protection of patent rights recognized by several countries [25, 29, 30]. Hence, the patent family was used as an indicator to evaluate the market value of patents in various studies [34, 38].

Table 2 shows the average of three factors by patent inventor gender, and test results for significance for their differences. Surprisingly, female participation showed a statistically significant positive correlation with the number of claims. However, the difference between patents with female and mixed inventors was not statistically significant. Top graph of Fig 5 shows differences in the number of claims by gender between 1970 and 2016. The trend in claim numbers is similar regardless of gender. Between 1995 and 2003, the number of claims for female was the highest, but it has been the least since then. In most years, patents with female participation have many claims rather than patents whose inventors are only males. For the remaining two factors, patents of male dominated female inventors. In addition, two factors are higher when male and female inventors interact compared to when the inventor of the patent is biased by one gender. The graph in the middle of Fig 5 shows the change in the citation counts over time by gender. The backward citations showed the same pattern regardless of gender, and in most of the time, female has less backward citations than other groups. The less backward citations indicate the power to voluntarily create novel inventions in a positive sense [40]. Lastly, the bottom graph shows the differences in the family size by gender. In most of the time, the patents of an inventor group with mixed gender had greatest family size than those of the single gender groups. Moreover, the patents invented by male-only group had larger number of family patents compared to those of female-only group.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5.

The annual trend of patent claims (top), backward citations (middle) and family size (bottom) by gender.

https://doi.org/10.1371/journal.pone.0305238.g005

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Comparison of bibliometric information by the gender of patent inventors.

https://doi.org/10.1371/journal.pone.0305238.t002

Second, we observed how the technical areas covered by patents differed by gender. We used six-digit IPC codes that can specify and differentiate cosmetic technology. Table 3 shows top ten IPC codes for male, female, and mixed inventors. The ratio of the occurrence of IPC codes in total patents in cosmetics by gender indicates the degree of the technical area accounting for cosmetics technology. Regardless of the gender of the patent inventors, techniques for skin and hair care accounted for a high proportion and were connected to medical technology. For patent inventions with women participation, interest in makeup-related technologies increases greatly, but interest in dental care-related technologies decreases significantly.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Top 10 six-digit IPC codes by gender.

https://doi.org/10.1371/journal.pone.0305238.t003

Third, we additionally compared the novelty of the technologies that cover by the gender of the inventors in terms of scope. The novelty in scope indicates the extent to which the invention expands the boundaries of technical scope and can be quantified by z-score which evaluates the vagueness of IPC combination [28, 39–41]. The Z-score quantifies the degree of the novelty as shown: where α and β are different IPC six-digit codes; o_αβ is the number of patents that have both α and β; μ_αβ and σ_αβ are the expected co-occurrences of α and β and its standard deviation, generated from a null model of the data that randomizes IPC arrangement while preserving IPC usage and number of patents within the data. In general, the high Z-score value represents a popular IPC combination, while low Z-score value is the novel combination of IPC. Since there can be more than one IPC combination in a single patent, the min and median values of the Z-score are used [42]. While the median value measures the degree to which the main body of a patent conforms to technological conventionality, min value measures the extent to which the invention contains a novel IPC combination.

Before calculating the z-score of cosmetic patents, we removed the IPC code: A61K 8 assigned to about 98% of patents. When only a single IPC is assigned to a patent, we consider it as missing value, not creating any novel recombination [40]. As shown in Table 4, the average Z_median for inventions of male, female, and mixed were 7.19, 6.91, and 10.11, respectively. The value of technological conventionality becomes greater when inventors of different genders interact in a patent. This shows that an inventor group of mixed gender generally includes highly conventional IPC combinations in their inventions compared to the groups of a single gender. On the other hand, the average Z_min for male, female, and mixed were 0, 0.98, and -0.96, respectively. The inventions of cosmetic technology with the interaction between men and women include more novel IPC combinations on average. These results indicate that the inventions from mixed gender group tend to have both novel IPC combinations and conventional IPC combinations, that resulting inventions can provide groundbreaking or practical solutions for specific needs. Overall, in perspectives of novelty, the average number of backward citations is less in female inventions as shown in Table 2, whereas the novelty in scope is greater in male or mixed gender group. This shows that the technical features of inventions from the female inventor groups are likely to be more innovative, having fewer existing prior arts, whereas their scope of technologies is comparatively conventional. Conversely, the inventions of the male group tend to be novel in technological scope, expanding the boundaries of the technological field, albeit with less emphasis on specific technical features that are genuinely new.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Conventionality and novelty of inventions by gender.

https://doi.org/10.1371/journal.pone.0305238.t004

Specifically, Table 5 presents the ratio of inventions by innovation type and gender in cosmetics technology. We define the inventions as four types based on the degree of conventionality and novelty: high conventionality and high novelty (H-H), high conventionality and low novelty (H-L), low conventionality and high novelty (L-H), and low conventionality and low novelty (L-L) inventions. Here, the level of high and low is a relatively compared to the average of the values of all patents in cosmetics technological field. H-H invention is more conventional patent than other patents in terms of technical scope, which means IPC combinations in the patent are frequently occurring in general, and also may include an element that is more novel, that is, rarely occurring IPC combination in the field. H-L invention is composed of conventional IPC combinations in general, but the no combination that is more novel than average. L-H invention has more than one IPC combinations that are more novel but the IPC combinations in general are less conventional. Most of the elements in L-L invention are neither too novel nor too conventional in their technological scope. The table shows that male inventors usually create the inventions that have high novelty in scope in general (L-H inventions) compared to those of other groups, which can result in radical innovation with high invention value. The invention value indicates the extent to which the patent receives forward citations that is highly correlated to the patent’s actually achieved value [41]. On the other hand, group of mixed gender often invent patents with more conventional IPC combinations than the groups of single gender, some with novel combinations included (H-H invention) and some with less novel combinations (H-L invention). These inventions may have relatively low inventive value regardless of novelty level since it would be difficult for novel knowledge to be the integrated into highly conventional knowledge [41]. Meanwhile, female inventors are more likely to participate in L-L invention that can bring gradual and modest improvements on existing solutions.

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Ratio of inventions by invention type and gender.

https://doi.org/10.1371/journal.pone.0305238.t005

Moreover, we used the text information of patents to compare the differences in the cosmetics technology characteristics by gender. We extracted an abstract that briefly summarizes the technology of the patent. We leveraged only one patent from each family patent to avoid using duplicated information. If different patents have the same family ID, the patents represent the same invention, so we retained patents with more claims. If the number of claims is also the same, we retained patents with more forward citations. We changed the uppercase to lowercase, performed stopwords, and removed words that have little contribution in constructing the topic as preprocessing of the text.

We conducted the structural topic modeling (STM) to observe differences in technology topics of cosmetics patents by gender. The gender of patent inventors was used as covariates. We used spectral initialization method, which conducts a spectral decomposition of the word co-occurrence matrix. This method is most often used among STM initialization methods because it shows a deterministic and globally consistent result under reasonable conditions. STM needs to pre-determine the number of topics in advance as a hyperparameter. We used four indicators to find the optimal number of topics. Semantic coherence [43] and exclusivity are indicators for evaluating the quality of the topic. Held-out likelihood and residuals are indicators for evaluating the generalization of the topic model. Through the four indicators, the optimal number of topics was 9 and 11 with the highest semantic coherence, as shown in Fig 6. We set 11 topics considering the other three indicators.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Four indicators for determining the number of topics in STM.

https://doi.org/10.1371/journal.pone.0305238.g006

We constructed each topic with high scored words. We calculated word score by giving higher weight to words that appear less frequently in other topics. Score divides the log frequency of the word in the topic by the log frequency of the word in other topics. The word composition of the eleven topics is shown in Table 6. We assigned the patent to one topic with the highest probability of being found among the 11 topics. Overall, the proportion of topic 6 (skin trouble care products) was the highest, as shown in Fig 7. This, again, shows that the cosmetics regarding skin trouble care have gained wide attention by inventors regardless of gender. On the other hand, topics 1, 7, 5, 3 are observed in relatively small proportions.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. Proportion by topics.

https://doi.org/10.1371/journal.pone.0305238.g007

Download:

• PNG
  larger image
• TIFF
  original image

Table 6. Topics found in cosmetic patents.

https://doi.org/10.1371/journal.pone.0305238.t006

With the powerful features of STM, we can estimate how much each covariate is associated for each topic, as shown in Fig 8. In topics 3, 6, 8, 9 and 10, it is observed that more proportion of patents is expected for those involving female inventors only. We can conclude that female inventors are more likely to invent the technologies related to comprehensive skincare, haircare, and makeup, which is widely used for female compared to male consumers due to different level of understanding on various skin types by gender and the traditional perception that improving beauty is feminine [44]. On the other hand, the proportions of topics 2, 4, 5, 7, and 11 decrease for patents with women participation. This shows that the participation of male inventors is with the inventions of well-being and personal care, which has greater demand than other cosmetic products by men customers in cosmetics market [45]. Especially, the collaborations between different genders had the lowest proportions in topics 2 and 7. Interestingly, the inventor group of mixed gender has highest proportion only in the topic 1. These trends in beauty-related topics may indicate the different interests in beauty products between genders, which may result in less collaborations between male and female inventors in patenting activities.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 8. Individual topic proportion by gender.

https://doi.org/10.1371/journal.pone.0305238.g008

We estimated the prevalence for the topics, which had highest proportions, using the linear regression according to the interaction between gender and year, as shown in Fig 9. Over time, there was a significant increase in Topic 6, regardless of gender of inventors. In Topic 10, the proportion of patents with female shows a marked increase over those with other gender compositions.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 9. Topic proportion by gender over period on each topic by gender and year.

https://doi.org/10.1371/journal.pone.0305238.g009

The performance of cosmetics patents

First, the most widely used indicator of performance evaluation of a patent is forward citation, which refers to the citations received from newer patents. The patents that have greater forward citations have the technological importance for the evolution of subsequent inventions [28, 29]. Hence, the number of forward citations has been used as a measure of inventive quality in previous studies [34, 46, 47].

We observed the number of patents from 0 to 100 cited by gender, as shown in Fig 10. The number of forward citations for patents follows the zero-inflated Poisson distribution, which is mostly concentrated in the value of zero. More patent with female inventors were less cited than those with male.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 10. The distribution of citation by gender.

https://doi.org/10.1371/journal.pone.0305238.g010

Each patent was assigned to the topic with highest probability among 11 topics that it can have. 932 patents without abstract information in the PATSTAT database were assigned topics based on their IPC codes. The applicant affiliation consists of company, hospital, individual, non-profit organization, university, and unknown.

We conducted the zero-inflated Poisson regression to observe how gender differences affect the number of patent forward citation predictions. Of the total patents, 4% of patents with more than 100 citations were considered as outliers and were excluded from our analysis. Variables for predicting the number of forward citations for a patent include categorical variables, which are patent inventor’s gender (base level: female), patent’s topic (base level: topic 1), and affiliation inclusion (base level: no applicants assigned to each affiliation), and continuous variables, which are the number of claims, family size, the number of backward citations, the length of grant lag, the number of inventors, and the number of assigned 8-digit IPC. The number of inventors is related to the solid knowledge foundation [28, 29], whereas the number of IPCs represents the technology scope, which indicates how diversified or broad technical features a patent composed of [25, 28, 29]. These variables are proposed to have a relationship with the patent value, which is approximated by forward citation in this study. We also used the age of the patent as an offset variable to reflect the characteristic that the number of citations increased over time. The zero-inflated Poisson distribution is determined by two parameters λ and π as follows:

Forward citation has a value of zero with the probability of π, and follows a Poisson distribution with a probability of 1−π having a parameter of λ. Parameters λ and π are estimated as follows:

Table 7 shows the exponential values and p-value of the parameters λ and π. The significant variables for parameter π is the gender of inventor, which affiliation the patent applicant belongs to, and which topic the patent is allocated to. The variable that shows significant interaction with parameter λ most is which topic the patent is allocated to, and there is little gender difference. Inventor groups with male-only and mixed gender have positive relationships with the receiving of at least one forward citation compared to the female-only group. However, the parameter π of the patents of mixed gender was not statistically significant. For male, the λ was 0.945 times of that of female, which indicates the inventions of males get slightly fewer forward citations compared to those of females, whereas the λ for mixed gender group was 1.052 times of that of female. These indicate that the probability of not being cited is significantly lower in male than in female, despite the fact that the number of forward citations of already cited patents is lower than that of female inventors’ patents. Meanwhile, topics from 2 to 11 have higher probability of being cited than topic 1 (Fragrance Related Products). Moreover, they have a positive relationship with the received number of forward citations, except for topic 4 (UV Protection Products).

Download:

• PNG
  larger image
• TIFF
  original image

Table 7. The exponential values and p-value of the coefficients.

https://doi.org/10.1371/journal.pone.0305238.t007

Second, patent maintenance term can also be used as performance indicators. In general, when a patent is granted, the owner of the patent must pay maintenance costs to be legally protected. Hence, the period during which patent holders pay a fee for the patent maintenance indicates the implied value of a patent. We assume that the maximum lifetime of a patent is 20 years from application filling date when there are no specific legal events for the patent. The lifetime of a patent can be further extended through legal events such as Term Extension of Rights. In addition, the lifetime of a patent can be terminated by a legal event such as Lapse and Withdrawal and can be restored through a Reinstatement event.

The lifetime of the patent was censored as of January 1, 2018, as we used the PATSTAT’s Spring version of 2018. We conducted a survival analysis to observe if there was any difference in the lifetime of patents by gender. First, based on the data, we observed how the survival rate of patents changes according to gender through Kaplan-Meier survival analysis, as shown in Fig 11. The survival rate of those with only female participation is lower than that of male and mixed genders starting at 2000 days. In 20 years of patent maintenance term, the difference in survival rate between male and female is about 8%.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 11. Kaplan-Meier curve of patents by gender.

https://doi.org/10.1371/journal.pone.0305238.g011

In addition, we fitted the cox proportional hazard model considering the several variables from those used in the previous analysis. The cox proportional hazard model estimates hazards, which indicate the risk that the life of a patent will be terminated at a particular point in time. The hazard function of the patent k at time t is defined as follows.

The result of the Cox proportional hazard model is shown in Table 8. First of all, hypothesis test regarding the proportional hazard based on Schoenfeld residuals has p-value greater than 0.05, which supports the use of the proportional hazards model. In addition, very small p-values associated with the likelihood ratio, Wald, and score tests back up the use of X variables in the proportional hazard model. The most important factor influencing the hazard of the patent was the number of claims and the composition of inventors. The hazard that the patents by male will be terminated is 0.96 times of that of female or mixed, which indicates that the patents invented by male-only teams have slightly lower risk of termination than those invented by teams of female or mixed gender. Patents with greater number of claims have a slightly higher risk of termination, but this may be different by technological industries [48]. However, there was no statistically significant difference in risk by affiliations or topics.

Download:

• PNG
  larger image
• TIFF
  original image

Table 8. Results of the cox proportional hazard model.

https://doi.org/10.1371/journal.pone.0305238.t008

Discussion

With the advent of the modern era, as gender discrimination has been greatly resolved, women’s participation in the STEM field is rapidly increasing. In particular, women inventors have participated more in the chemical and health areas [4, 5], which are highly relevant to cosmetics technology. However, existing research on cosmetics hardly investigated the technological development regarding the gender of the inventor group. To understand the phenomena caused by these changes, this study identifies the contribution made by female expressed in cosmetic patents in terms of the three research questions.

With regard to the first research question, we observed the difference in application and grant according to women’s participation in cosmetic patent invention. In the past, most of the cosmetic patent applications were made only by male inventors, but over time, collaboration between male and female inventors has created many cosmetic patent applications. This result is in line with previous studies that demonstrated the increasing participation of female inventors in innovation, especially related to women’s life sciences. Although women’s participation in cosmetic patent applications is increasingly active, their participation in patent grants is negatively associated in cosmetics as in other sectors [8, 17].

For the second research question, we observed the difference in technological characteristics according to women’s participation in cosmetic patent invention. The collaboration between male inventors and female inventors has a positive impact on the number of backward citations and family size in cosmetic patents. Also, in terms of patent claims, the patents with participation of female had greater number of claims than the patents of male inventors. The inventor group with gender diversity also develops technology with both conventional and novel components in terms of technical scope, which can provide groundbreaking and practical solutions. These results show the technological improvement of cosmetic patents with the participation of different genders. Additionally, we observed differences in terms of the technical field by inventor groups. Within the detailed technical field extracted using STM, we notice some variation of topic prevalence depending on women’s participation. Specifically, the differences in the proportion of topics by inventors’ gender may correspond to their interests; the make-up-related topics are more likely to be invented by female inventors, while the topics regarding dental care products are more likely to be invented by male inventors [44, 45, 49]. Skin care-related topics have the highest proportions across all topics regardless of gender, of which products meeting the preferences of diverse consumer groups. These results show that the gender composition of cosmetic patents demonstrate various changes in technical characteristics.

In relation to the third research question, we observed the difference in performance according to women’s participation in cosmetic patent invention. There was little difference in the number of citations between the male inventor’s patent and the female inventor’s patent. In addition, female inventor or mixed-gender inventor group’s patent tends to be terminated slightly earlier than a male inventor’s patent. These results show that women’s participation still has a negative relationship with patent performance.

Conclusion

This study investigates the gender diversity that affects innovation characteristics and performance in cosmetics technology. The results of the study propose that women’s participation in cosmetic inventions is becoming active and has brought about many changes in technical characteristics, but in terms of performance, it is still sluggish.

Our research contributes to the research area of cosmetics and innovation by gender. This is one of few studies that explore the patenting activities by gender in more balanced fields with less gendered stereotypes involved. Specifically, we delve into the technological characteristics and performance in various perspectives to enhance the understanding of inventors’ gender and gender diversity affecting the development of cosmetics technology. The findings of the study provide the evidence of the thematic differences and characteristics of cosmetics technology by gender participation. The stakeholders can leverage the information to plan their R&D strategies considering the landscape and market demands to better address the various needs of different genders. Specifically, based on the proposed results, the direction of technology development and collaboration strategies can be established.

As to suggestions for future research, instead of classifying the gender characteristics to three levels (male, female, mixed gender), more detailed examinations implicating the proportion of female inventors may deepen the experiment. In addition, one may investigate the differences between the patent applications and granted patents of female inventors in depth, which provide insight into topical variation or factors that decide the quality of inventions. While this study focuses on documented inventions in cosmetic technology fields and its relationship with gender composition of inventor teams as the research field entails women as the mainstream customers, generalization on the research domain is possible. The research framework provides a baseline for extending the study to other fields of gender-oriented industries, which may contribute to better understanding the complexity of gender characteristics reshaping technology development.