COVID-19 and changes in content usage behavior: The case of South Korea

Abstract

This study examines changes in content usage time due to the COVID-19 pandemic in South Korea using Korean Media Panel data for the period 2011–2020 and explores the reasons for these changes. This study focuses on four principal contents: television programs, movies/videos/user-created content, traditional telecommunication services, and chatting/messenger/social network services. The empirical results indicate an increase in usage time for the four principal contents, as well as total content usage time because of the pandemic. The results also show that average Korean people stayed longer at home after the onset of the pandemic, leading to an increase in the time spent on all the principal contents, except for traditional telecommunication services, as well as an increase in total content usage time. Furthermore, this study suggests that whereas the effect of the pandemic on television program viewing time was mainly attributable to changes in time spent at home because of the pandemic, the effect on other contents was mainly caused by non-location-related factors. This study predicts changes in content usage time after the end of the pandemic and provides strategic suggestions.

1. Introduction

People have changed their media usage behavior along with technological developments such as the advent of new media that displace old media. Accordingly, the reasons for and the results of changes in media usage have been an important research topic in various social science fields. For example, previous studies have examined whether the use of the Internet has displaced traditional media such as television, newspapers, and radio. Many studies have examined how the diffusion of the Internet has reduced television viewing, people's single most important medium until now (Jang & Park, 2016; Lee, 2018; Lee & Leung, 2008; Liebowitz & Zentner, 2012). Recently, with the acceleration of broadband diffusion and the increasing popularity of over-the-top (OTT) services such as Netflix and YouTube, several studies have examined the relationship between online video distribution and traditional pay television (Cha, 2013; Cha & Chan-Olmsted, 2012; Lee, Lee, & Kim, 2016; Waldfogel, 2009).

In addition to replacing old media with new media, big events, along with changes in social trends, also change how people use media. In particular, the spread of infectious diseases such as COVID-19 has substantially changed people's media usage. Changes in media usage time due to the pandemic can be decomposed into two components: those because of changes in time spent at home (hereafter referred to as the location effect), and those attributable to all remaining factors other than location, which affect people's need for media usage (hereafter referred to as the non-location effect). The non-location effect may be partly caused by the pandemic's impact on people's psychological states1 and/or their desire to explore health-related information during the pandemic.

The term “media” has been used in various contexts and meanings. Media usage time can be measured in several aspects: media device, content or usage behavior, connection method, and so on. For example, the substitution of the Internet for television viewing frequently implies that access to online videos or user-created contents (UCCs) over the Internet has reduced time spent watching television programs, but it also means that more and more people are watching television programs on their smartphones or laptops instead of home television sets. The former refers to the competitive relationship between contents because of the development of the Internet, whereas the latter refers to the substitution between media devices for the use of the same content. Therefore, it is necessary to establish how media is defined and in what aspects media usage time is measured. This study measures media usage time as time spent using each content mainly because it focuses on changes in people's media usage behavior because of the pandemic. The pandemic was more likely to affect people's choice of content through location and/or non-location effects.

We examine changes in content usage time due to the pandemic in South Korea (hereafter referred to as Korea) and explore the reasons for these changes, with special attention paid to time spent at home. In particular, using reliable panel data, we address the following research questions: (i) For which content is usage time rising and declining? (ii) How has COVID-19 affected the usage time of individual contents? (iii) To what extent are these changes in content usage time attributable to location and non-location effects?

We performed an empirical analysis using Korea Media Panel (KMP) data for the period 2011–2020. The KMP, which is government-authorized and annually funded by a public research institute, the Korea Information Society Development Institute (KISDI), provides extensive and reliable panel data on the media usage behaviors of Koreans for a sample period (Jang & Park, 2016; Sung & Kim, 2020). We chose the four principal contents from the viewpoint of usage time, also contributing to the purpose of simplifying the presentation of empirical results: television programs, traditional telecommunication (telecom) services such as voice calls/text messages/e-mails, movies/videos/UCCs (hereafter referred to as MVUs), and chatting/messenger/social network services (SNSs) (hereafter referred to as CMSs). As indicated in Table 2, as of 2020, these four contents were the four most used contents.2 In addition, these four contents are not only essential in ordinary people's daily lives, but some of them have also attracted special attention from mass media and previous media studies. For example, the increasing popularity of OTT services and SNSs has sparked controversies regarding regulatory policy and media substitution (Peitz & Valletti, 2015; Stork, Esselaar, & Chair, 2017; Wellmann, 2019; da Silva & de Andrade Lima, 2022).

Table 2.

Changes in the average content usage time (unit: minutes).

	2011–2013	2014–2016	2017–2019	2020
All contents	1,195.5	1,203.8	1,246.5	1,333.5
Employed persons	1,134.6	1,141.0	1,204.4	1,305.5
Jobless persons	1,248.9	1,249.4	1,287.7	1,373.5
Students	1,272.8	1,319.4	1,326.1	1,381.6
Television programs	520.7	508.2	494.3	501.6
Radio/music channels	31.8	40.6	43.2	46.3
Movies/videos/UCC	10.2	15.2	30.9	90.6
Music/digital soundtracks	20.2	27.6	40.9	49.9
Newspaper articles	34.9	17.1	13.6	25.5
Books/magazines	145.3	134.8	121.7	74.5
Traditional telecom services	189.9	154.3	147.3	143.0
Chatting/messenger/SNS	38.5	78.3	92.8	114.4
Information content	52.6	57.3	55.3	60.8
Online commerce	8.4	5.9	9.1	12.6
Games	38.5	37.0	44.0	53.2
Work-related content	95.0	117.4	141.2	153.2
Cultural content	5.8	7.6	9.4	3.6
Others	3.6	2.5	2.6	4.4

Note: All numbers are based on the sum of primary and simultaneous usage time in three days. Television programs include both real-time programs and videos on demand. Employed persons include self-employed persons and jobless persons exclude students.

Without a crisis such as COVID-19, drastic changes in people's length of stay at home would not occur. In this sense, the pandemic provides a unique research opportunity to investigate the relationship between people's choice of places to stay and content usage behavior. In addition, to the best of our knowledge, few or no studies have focused on the effect of COVID-19 on content usage behavior, despite the extensive literature on its outcome. This study contributes to the COVID-19 literature by addressing this gap.

The remainder of this study is organized as follows. In Section 2, after reviewing the literature, we specify several research hypotheses and questions on changes in content usage time and their relationship with COVID-19. Section 3 presents the research method along with a detailed explanation of the data. Section 4 analyzes the changes in content usage time after the onset of COVID-19 and provides estimation results. Finally, this study provides a summary, along with its strategic implications and limitations.

2. Literature review

2.1. Trends in media or content usage time

People change their media usage time for several reasons, including changes in their socio-economic status. One of the main reasons for the change is the emergence of new media and the substitution of new media for old media. For the past several decades, previous studies on media substitution have explored whether new online media that emerged along with the spread of the Internet and digital devices replaced traditional media such as television viewing, newspapers, and traditional telecom services. This literature frequently regards media substitution (complementarity) as a negative (positive) relationship between the time expended on two media services (Sung & Kwack, 2016), assuming the principle of relative constancy in the time spent on media or a zero-sum principle. That is, the media substitution theory describes the time displacement effect, indicating a decrease in the time spent on old media due to an increase in the time spent on new media.3 In contrast, previous studies in the field of communication have frequently relied on the concepts of both the uses and gratifications theory and the niche theory as theoretical frameworks to explain the adoption and use of new media. The uses and gratifications theory postulates that an individual selects a medium over others to gratify specific needs, with special attention paid to the need for gratification provided by various media (Cha & Chan-Olmsted, 2012; Katz, Blumler, & Gurevitch, 1973; Ruggiero, 2000; Stafford, Stafford, & Schkade, 2004). The niche theory explores the competitive relationship between old and new media, such as the degree of competition (overlap) between the two media and consumers’ perceptions of satisfaction toward one medium over another (Dimmick, Chen, & Li, 2004; Ha & Fang, 2012).

Whatever theoretical framework is adopted, in general, empirical studies have identified an increase (decrease) in the usage time of new (old) media because of media substitution. For example, although several earlier studies found no or weak substitutability between online and offline newspapers (Dimmick et al., 2004; Kaye & Johnson, 2003; Nguyen & Western, 2006), more recent studies tend to support the displacement hypothesis (Gentzkow, 2007; Ha & Fang, 2012; Lee & Leung, 2008; Sung & Kim, 2020; Westlund & Färdigh, 2011) by showing an increase in the use of online newspapers rather than print newspapers. Most previous studies demonstrate that the diffusion of the Internet reduces time spent watching television (Jang & Park, 2016; Lee, 2018; Lee & Leung, 2008; Liebowitz & Zentner, 2012). Similarly, several studies assert that the diffusion of OTT services or online video distribution reduces conventional television viewing (Cha, 2013; Cha & Chan-Olmsted, 2012; Lee et al., 2016; Waldfogel, 2009).

Recently, as in other countries, OTT services are rapidly gaining popularity in Korea. KISDI (2021, p. 2021) indicates that Korea's OTT market sales (subscriptions) from 2016 to 2020 recorded an average annual growth rate of 27.5% (24.9%). According to the KMP, the proportion of people who have used OTT services in the past three months has increased from 41.0% in 2019 to 72.2% in 2020. Netflix has acquired millions of subscribers in a short period of time since its provision of online video streaming services in 2016. OTT platforms offer TV programs, but their killer content continues to include movies, videos, and UCCs (MVUs in this study). YouTube, which mainly provides UCCs, is the most popular OTT platform in Korea. Considering that this growing popularity of OTT services leads to an increase in MVU usage time, we propose the following hypothesis:

H1: The time spent on MVUs is increasing.

In contrast, the rapid growth of OTT services in Korea is threatening conventional television viewing. Lee et al. (2016) indicate that the time spent on online video services reduced the time spent watching terrestrial and pay televisions even in the early 2010s in Korea. This substitution of OTT services for television viewing may have intensified with the wide spread of OTT services. However, a reduction in television viewing, especially through terrestrial or pay television connections, does not necessarily mean a decrease in the use of television programs as content. More and more people watch television programs through smartphones, and television programs are popular items of OTT services, making it difficult to draw definitive conclusions about the usage time of television programs in Korea. Therefore, we propose the following research question.

RQ1: Is the time spent on television programs increasing or decreasing?

Over the last decade, more and more people have been using digital platforms, such as mobile messengers and SNSs, as a means of communication. As these digital platforms are offering the same staple services of voice, messaging, and video calls that used to be the domain of traditional telecom carriers, they threaten to cannibalize traditional telecom services (Mohr & Meffert, 2017). In fact, whereas mobile messengers have gained wide popularity among mobile users, text messages and telephone calls are declining in some countries, causing a decrease in sales or profits for telecom operators (Jirakasem & Mitomo, 2018; Stork et al., 2017; Wellmann, 2019). The development of mobile messengers and SNSs (CMSs in this study) has caused controversy over whether CMS replaces traditional telecom services (Gerpott, 2015; Gerpott & Meinert, 2016), raising policy concerns in the electronic communications markets (Peitz & Valletti, 2015).

In Korea, the increased use of CMSs has become a serious threat to traditional telecom services. Shin and Kim (2015) confirmed that mobile messengers reduced the use of voice calls and text messages during the period 2010–2013. KakaoTalk, the number one mobile messenger in Korea, became a dominant messenger in the early 2010s (Lee & Kim, 2016), indicating the substitution of mobile messengers for text messaging and voice calls (Rhee & Kim, 2014). Korea Press Foundation (2021) reports that as of 2021, 99.4% of Koreans have used KakaoTalk mainly for communication. The KMP indicates an increase in the ratio of SNS users to all sample respondents from 16.8% in 2011 to 52.4% in 2020 in Korea. Along with this increase in the use of CMSs, the decline in traditional telecom services is expected to become clearer. Therefore, we present the following hypothesis:

H2: Whereas the time spent on traditional telecom services is on a decreasing trend, the time spent on CMSs is on an increasing trend.

2.2. COVID-19 and content usage

Along with the worldwide spread of COVID-19, the literature on this pandemic has surged in a short period. As Brodeur, Gray, Islam, and Bhuiyan (2021) indicate, research on the pandemic is not limited to economic issues but encompasses a wide range of fields, from the measurement of the spread of COVID-19 to its effects on health, gender, racial inequality, and the environment. In contrast, little attention has been paid to the changes in the usage of media or content due to the pandemic. Some studies on media usage have examined the role of mass or social media in creating and disseminating unverified information and fake news during the pandemic. This is because the COVID-19 crisis has renewed concerns about the dangers of misinformation and its persuasive effects on behavior (Allington, Duffy, Wessely, Dhavan, & Rubin, 2020; Bursztyn, Rao, Roth, & Yanagizawa-Drott, 2020; Liu, Chen, & Bao, 2021; Simonov, Sacher, Dubé, & Biswas, 2020). Another stream of research explores various aspects of media use during the pandemic, such as the characteristics of people who easily share unverified information (Cato et al., 2021; Islam, Laato, Talukder, & Sutinen, 2020). Despite these previous studies on media usage, to the best of our knowledge, this study is the first to focus on changes in content usage due to the pandemic.

As noted previously, the spread of COVID-19 can affect people's content usage mainly via two routes: location and non-location effects. The location effect states that the pandemic affects people's choice of places to stay, which in turn leads to changes in content usage time. The non-location effect refers to all remaining factors that affect people's need for content usage. For example, the pandemic has affected people's need for content usage through its impact on their psychological state. In addition, people may spend more time searching for health-related information because of fear of infection during the pandemic, resulting in changes in content usage behavior.

People allocate their time to a variety of activities, both media- and non-media related, within a 24-h time budget. People devote some of their available free time, that is, hours of the day excluding the time required for employment, studies, housework, etc., to using various contents. Content usage time frequently depends on people's places of stay. For example, traditional content consumption, such as television viewing and newspaper reading, can mostly be regarded as home-based activities (Thulin & Vilhelmson, 2019). Previous media-related studies on the use of time and space have focused on the effect of people's extensive use of digital media on various offline activities, especially among young people (Cope & Lee, 2016; Robinson, 2011; Thulin & Vilhelmson, 2019). However, little attention has been paid to the effect of COVID-19 on people's choice of places to stay, as well as the relationship between location and content usage.

When working from home is forced because of voluntary or compulsory social distancing to contain the spread of COVID-19, it increases the time spent at home. A decrease in commuting time because of working from home may result in an increase in available free time, causing people to spend more time using content. However, the change in the usage time of individual contents due to the increase in the time spent at home may vary depending on the content's characteristics. For example, the majority of people tend to watch television programs on their television sets at home, indicating that television program viewing time increases with an increase in time spent at home. Lee (2017) indicated that most people still watch TV programs at home, and approximately half of them watch TV programs with other family members in Korea. Im and Ahn (2021) also suggested that real-time TV viewing takes place overwhelmingly at home. In addition, family co-viewing tends to increase individual viewers' TV consumption (Mora, Ho, & Krider, 2011), also increasing TV program viewing time.

Because people use MVUs for reasons that overlap with TV programs, we anticipate that people will use MVUs more frequently as their time spent at home increases. Using the KMP data, Seo (2019) showed that as people stay longer at home, they tend to spend more time on smartphones, leading to an increase in the usage of CMSs because CMSs are mainly based on smartphones. Finally, if people spend more time at home with their families, their need for voice calls and text messages is expected to decrease. Therefore, we propose the following hypothesis.

H3: More time spent at home leads to an overall increase in the use of contents.

H4: Along with the increase in time spent at home, people are expected to spend more time on television programs, MVUs, and CMSs but less time on traditional telecom services.

Contrary to the location effect, it is not clear whether the non-location effect increases or decreases individual content usage time. For example, the relationship between people's psychological state and their use of individual content is ambiguous, with few empirical studies on this relationship being reported. People may have become more dependent on traditional and/or emerging content to overcome their feelings of social isolation owing to the pandemic. By contrast, as people experience more anxiety and depression, they may be more reluctant to use content. Because the non-location effect is uncertain, the net effect of COVID-19 on the usage time of individual content is also uncertain. Therefore, we present the following research question:

RQ2: What is the net effect of COVID-19 on usage time of individual content?

3. Data and research method

3.1. Data

This study uses KMP data for the period 2011–2020. The KMP provides a complete picture of media usage and ownership in Korea by conducting repeated surveys on the same respondents sampled in 2011.4 However, some of the sampled respondents were no longer available over time for several reasons, such as moving to other locations or non-response, making it necessary to add 1,027 (972) households and 2,436 (2,355) individuals to the survey participants in 2019 (2020). In this sense, the KMP has unbalanced panel data. As of 2020, the KMP has collected responses from 4,260 households and 10,302 individuals, with various data collected on media ownership and connectivity, communication service subscription and expenditure, and media usage behaviors, along with demographic information for the sample households and individuals.

The KMP includes questionnaires for both households and individuals. In particular, the questionnaire for individuals contains a time-use diary, referred to as a media diary, in which an individual records information regarding their main and secondary (simultaneous) media usage for three whole days. The media diary divides the day into 15-min periods, and respondents are asked to describe what they are doing in each period. Specifically, in a media diary, each respondent reports their media usage with different combinations of four categories: 17 places, 42 media devices, 40 content or usage behavior, and 21 connection methods. For example, a respondent may report that they primarily watch an over-the-air broadcasting program (content) through a home television set (media device), which is connected using IPTV (connection method) at home (location), stating that they simultaneously watch real-time UCC (content) through a smartphone (media device), which is connected using a 5G mobile service (connection method) at home (location). Using data on places, we calculated the length of stay for each individual place. Forty types of content were regrouped into 14 contents to simplify the presentation of results, as shown in Table 2, with the four principal contents being the focus. To eliminate outliers, this study extracted only respondents aged 8 to 80, and excluded respondents whose home staying time or usage time of all contents is equal to zero.

As shown in Fig. 1 , Korea has experienced a short-lived spread of COVID-19 and a subsequent lull since the first confirmation of the infection in February 2020 and before the onset of the Omicron variant. In Korea, the actual number of confirmed infections before the onset of the Omicron variant has remained relatively low without official lockdowns compared to that in other countries because of effective policies such as aggressive testing, contact tracing, and strict quarantine on a targeted few (Aum, Lee, & Shin, 2021; Han, 2021). Additionally, to contain infectious diseases, the Korean government has imposed preventive social distancing measures such as bans on gatherings and restrictions on businesses (Cho, 2021). The KMP conducts a survey for 10 weeks between June and July every year, with the time of the survey being different for each individual. Fig. 1 indicates that this survey period corresponds to the incubation period immediately before the second widespread COVID-19 wave. During this period, the Korean government presented only personal and community guidelines to contain the pandemic, termed “distancing in daily life,” without any mandatory social distancing measures. Therefore, any changes in content usage time during this period should be attributable to people's voluntary choices to cope with the risk of infection.

Fig. 1.

Changes in the number of confirmed infections and deaths in Korea: https://covid19.who.int, Note: Weekly data are compiled Sunday to Sunday. Incomplete weeks are censored.

3.2. Measurement method of content usage time

Using the media diary data, an individual's usage time for each content is calculated by multiplying 15 min by the number of periods for which they record the content under main or secondary use. For example, an individual's time spent viewing TV programs is calculated by multiplying 15 min by the number of periods for which they report watching TV programs as their main activity or simultaneously with other activities, regardless of location, media device, or connection method. TV programs include not only terrestrial TV programs but also those provided by cable networks (called program providers in Korea). Similarly, the other three contents are calculated by focusing only on the time when the corresponding content is used, regardless of location, media device, or connection method.

We paid special attention to the following two problems when comparing content usage times across individuals and over time. First, the KMP selected original and additional survey respondents using a stratified multi-stage sampling design drawn from the Population and Housing Census of Statistics Korea. Owing to this sampling design, the mean content usage time should be calculated using individual cross-sectional weights. Second, the time reported in the media diary should be adjusted to control for the effects of weekends. The available free time varies greatly between weekdays and weekends. An individual's time spent at a specific place on weekends is substantially different from that on weekdays. For example, people usually spend more time at home on weekends than weekdays. In the KMP, each survey respondent fills out a media diary at different times each year, making it difficult to compare the staying time over time. Therefore, for year-by-year comparison, we converted an individual's weekend usage time for each content into the corresponding value for weekday usage time by employing the average ratio of their weekday to weekend usage time during the analysis period.

3.3. Econometric model

Two issues must be considered while identifying the effects of COVID-19 on an individual's content usage time. First, an individual's content usage time depends on their socioeconomic status. For example, if a student finds a job after graduation, they are likely to spend less time on music or videos. Second, an individual's content usage time is likely to display an upward or downward trend over time owing to several reasons, some of which we explain in the previous section. Therefore, we control for an individual's socioeconomic (demographic) variables and time trend in the estimation. Reflecting on this discussion, we specify the econometric model as follows:

$$CU{T}_{it}^m={\alpha }_0+{\alpha }_{1}TM+{\alpha }_{2}COVID+\sum _{j=1}^J{\beta }_j{x}_{it}^j+u_i+{\epsilon }_{it}$$ (1)

$$CU{T}_{it}^m={\alpha }_0+{\alpha }_{1}TM+{\alpha }_{2}HS{T}_{it}+{\alpha }_{3}COVID+\sum _{j=1}^J{\beta }_j{x}_{it}^j+u_i+{\epsilon }_{it}$$ (2)

Here, the dependent variable is individual i's usage time of the m-th content in year t ($CU{T}_{it}^m$). Specifically, we used both the total content usage time and usage time of the four principal contents as dependent variables. TM and COVID refer to the time trend and dummy variable for the COVID-19 outbreak period, respectively. That is, COVID equals unity for 2020, and zero otherwise. HST refers to the time spent at home, and $x^j$ is the $j$-th demographic variable. $u_i$ and ${\epsilon }_{it}$ are the unobservable time-invariant individual-specific effects and conventional time-variant error terms, respectively.

In Equation (1), we place special attention to the regression coefficient of COVID, which measures the effect of COVID-19 on the usage time of each content with a time trend and demographic variables controlled for.5 In other words, the coefficient assesses whether content usage time has risen above or has fallen below the trend line after the outbreak of COVID-19. Equation (2) differs from Equation (1) in that it includes the HST variable. That is, the regression coefficient of COVID in Equation (2) (${\alpha }_3$) measures the effect of the pandemic on content usage time, with changes in the time spent at home due to the pandemic controlled for, that is, the non-location effect.

The demographic variables used in the estimation include age, years of schooling, personal income, the number of family members (family size), job status (employed, jobless, and students), sex, marital status, and type of household (urban vs. rural). A dummy variable is specified for the last four variables. Specifically, a dummy variable equals unity for the jobless, students, males, married persons, and urban households, respectively, and zero otherwise. The KMP measures the educational level (graduation or enrollment status) in six (five) sections, from uneducated (enrollment) to graduate school or higher (graduation), and measures the income level in 18 sections, from no income to 8 million Korean won (KRW) or more. To reduce the number of control variables, we converted these interval scales into continuous values. For example, we regarded the years of schooling for high school graduates as 12 years and calculated the income of a person with an income level between 3 million KRW and 3.5 million KRW as 3.25 million KRW. Since the marginal effect of age on content usage time may increase or decrease with age, age squared is added as a control variable. For example, the usage time of MVUs or CMSs may increase with age but may decrease after a certain age. We also considered the interaction term between years of schooling and income as a control variable because the marginal effect of one variable on content usage may depend on the other variable.

People's socioeconomic status is predetermined before their choice to allocate time to the use of each content. Therefore, all demographic variables can be regarded as exogenous variables, making it appropriate to specify Equation (1) as a random-effects model. Furthermore, when between-variations dominate within-variations, as in this study, the random-effects estimator is more efficient than the fixed-effects estimator. In contrast, the time spent at home (HST) may be associated with the unobserved individual-level random effect ($u_i$), indicating that HST may be an endogenous variable. Therefore, when HST is included in Equation (2), we utilize the Hausman-Taylor estimator. The Hausman–Taylor estimator is one of the random-effects models, which allows some of the explanatory variables to be correlated with individual random effects ($u_i$). This estimator, originally proposed by Hausman and Taylor (1981) and developed by Amemiya and MaCurdy (1986), is based on instrumental variables using the information included in the model. Additionally, the use of the Hausman–Taylor estimator makes it possible to estimate coefficients on time-invariant explanatory variables.

4. Results

4.1. Changes in home staying time and content usage time due to COVID-19

Table 1 presents the changes in the average time spent at various places over three days for the entire sample. In Table 1, the analysis period is divided into three sub-periods of three years and 2020 to simplify the table and identify the changes before and after the outbreak of the pandemic. For reference, Appendix Table 1 shows the annual changes in the average time of staying at places. Fig. 2 depicts annual changes in the average time spent at home for the entire sample and three employment types: employed, jobless (e.g., housewives and job seekers), and students.

Table 1.

Changes in the average time spent at various locations (unit: minutes).

	2011–2013	2014–2016	2017–2019	2020
Home	2,680.2	2,634.0	2,584.1	2,741.2
Employed persons	2,309.6	2,292.0	2,285.6	2,344.0
Jobless persons	3,517.7	3,431.9	3,349.2	3,559.4
Students	2,492.4	2,550.3	2,545.8	3,073.1
Workplaces	799.5	828.7	817.9	818.1
Schools	309.5	256.0	229.7	112.6
Transportation	209.0	228.0	244.6	230.8
Restaurants	62.0	84.2	112.9	98.3
Commerce facilities	58.7	66.1	68.2	59.7
Places for leisure	50.5	57.2	61.4	41.4
Others	111.1	103.0	109.7	78.8

Note: All numbers are based on the sum of the time spent in three days. Transportation includes both public and private transit. Places for leisure include entertainment, recreation, sports and cultural facilities. Employed persons include self-employed persons and jobless persons exclude students.

Fig. 2.

Changes in the average time spent at home (unit: minutes), Note: All numbers are based on the sum of the time spent in three days. Transportation includes both public and private transit. Places for leisure include entertainment, recreation, sports and cultural facilities. Employed persons include self-employed persons and jobless persons exclude students.

Table 1 indicates that the average Korean population spent more time at home after the onset of the pandemic. Specifically, average Korean people spent approximately 157 min (=2741.2–2584.1) longer at home over a span of three days in 2020 than in the previous sub-period (2017–2019). Moreover, Table 1 and Fig. 2 show that the length of stay at home in 2020 increased compared with the previous periods, regardless of employment type. For example, the jobless, on average, spent approximately 59 h (3559.4 min) at home in 2020, that is, approximately 210 min longer than in the previous sub-period. On average, students stayed at home for approximately 527 min longer in 2020 than in the sub-period just before the pandemic. This largest increase among the three employment types appears to be mainly because of the parallel between offline and online classes. However, average Korean people spent less time at restaurants, commerce facilities, and places for leisure in 2020 than in the period 2017–2019. This decrease is unambiguously attributable to the fear of infections, considering that the time spent at these locations displayed a steadily increasing trend before the pandemic.

Table 2 displays changes in the average usage time of each content for the entire sample across the sub-periods and 2020, whereas Appendix Table 2 shows its annual changes. Fig. 3 depicts annual changes in the average usage time of the four principal contents for the entire sample and for three employment types. All numbers are based on the sum of main and simultaneous usage time in three days.

Fig. 3.

Changes in the average usage time for four principal contents (unit: minutes), Note: All numbers are based on the sum of primary and simultaneous usage time in three days. Television programs include both real-time programs and videos on demand. Employed persons include self-employed persons and jobless persons exclude students.

Table 2 demonstrates the tendency of average Korean people to spend more time on the usage of all contents after the pandemic. In particular, the total content usage time, on average, drastically increased in 2020 regardless of employment status, mainly because of a large jump in the usage time of MVUs and CMSs after the pandemic. For example, the average value for total content usage time of the jobless increased by 38.3 min from the second sub-period (2014–2016) to the third sub-period (2017–2019) and soared by 85.8 min in 2020 from the third sub-period. As shown in Fig. 3, this large increase was mainly attributable to an increase in the usage time of MVUs and CMSs by the jobless. In particular, the average time spent on MVUs by students increased by more than three times in 2020 compared with that of the previous sub-period. This is mainly because online classes are classified as videos in the KMP.

As shown in Fig. 3 and Appendix Table 2, the average television program viewing time displayed no clear pattern during the analysis period, including 2020, regardless of the employment type. In contrast, Korean people, on average, spent less time on traditional telecom services over time and even after the pandemic. Another noteworthy finding is that the average time spent on reading newspaper articles continued to decrease before the pandemic but nearly doubled in 2020 from the previous sub-period. Similarly, the average time spent on information content increased from 55.3 min for the sub-period 2017–2019 to 60.8 min in 2020. These phenomena appear to be because of the increasing need for information, especially COVID-19-related information, following the pandemic.

4.2. Estimation results using unbalanced panel data

Table 3 reports the summary statistics for the variables used in the estimation. All summary statistics can be interpreted conventionally. For example, sampled respondents, on average, expended 590.0 min, 26.1 min, 154.9 min, and 67.4 min on television programs, MVUs, traditional telecom services, and CMSs, respectively, in three days during the analysis period. The sample averages for age, years of schooling, and income are 44.2 years, 11.5 years, and 12.3 hundred thousand KRW, respectively. The proportion of males, married people, and urban residents is 46.3%, 61.1%, and 88.0%, respectively.

Table 3.

Summary statistics.

	Cases	Mean	Std. Dev.	Min	Max
Dependent variables (Time spent on) (unit: minutes)
All contents	96,930	1,260.0	603.2	15.0	6,480.0
Television programs	96,930	590.0	422.4	0.0	3,300.0
MVUs (movies/videos/UCCs)	96,930	26.1	106.4	0.0	3,180.0
Traditional telecom services	96,930	154.9	145.6	0.0	3,870.0
CMSs (chatting/messenger/SNSs)	96,930	67.4	132.1	0.0	3,555.0
Independent variables
Time spent at home (minutes)	96,930	2,817.7	699.1	30.0	4,320.0
Age (years)	96,930	44.2	19.3	8.0	80.0
Years of schooling (years)	96,930	11.5	4.1	0.0	18.0
Income (hundred thousand KRW)	96,909	12.25	14.90	0.0	82.5
Family size	96,930	3.411	1.207	1.0	10.0
Dummies for job status (Base category = Employed persons)
Jobless persons	96,930	0.288	0.453	0.0	1.0
Students	96,930	0.192	0.394	0.0	1.0
Dummy for males	96,930	0.463	0.499	0.0	1.0
Dummy for the married	96,930	0.611	0.488	0.0	1.0
Dummy for urban residents	96,930	0.880	0.326	0.0	1.0

Note: Television programs include both real-time programs and videos on demand.

Table 4, Table 5 present the estimation results for Equations (1), (2), respectively. As indicated previously, the dependent variables are the usage time of all contents (Model 1), television programs (Model 2), MVUs (Model 3), traditional telecom services (Model 4), and CMSs (Model 5). Because most parameter estimates are statistically significant, we mention statistical significance only when necessary. In addition, the parameter estimates of the demographic variables are interpreted in Table 5, partly because of the small differences in their signs and statistical significance between the two tables.

Table 4.

Estimation results for Equation (1).

	Model 1	Model 2	Model 3	Model 4	Model 5
Dep. Var.	All contents	Television programs	MVUs	Telecom services	CMSs
Time trend (TM)	10.589***	−1.472***	4.027***	−8.654***	9.150***
	(0.868)	(0.519)	(0.139)	(0.209)	(0.164)
Dummy for COVID-19 (COVID)	85.652***	7.047*	59.048***	15.927***	2.673*
	(6.026)	(3.640)	(1.965)	(1.370)	(1.524)
Age	0.477	8.907***	0.482**	2.099***	0.851***
	(1.266)	(0.838)	(0.232)	(0.268)	(0.270)
Age squared	−0.028**	−0.025***	−0.016***	−0.021***	−0.020***
	(0.013)	(0.009)	(0.002)	(0.003)	(0.003)
Years of schooling	16.862***	−11.838***	1.578***	4.622***	7.238***
	(1.114)	(0.739)	(0.218)	(0.234)	(0.253)
Income	−7.552***	−3.340***	0.142	1.982***	1.280***
	(0.960)	(0.607)	(0.128)	(0.273)	(0.171)
Schooling years*Income	0.771***	0.173***	−0.028***	−0.082***	−0.102***
	(0.066)	(0.039)	(0.010)	(0.018)	(0.012)
Family size	−3.869	−12.048***	0.867**	−0.533	1.850***
	(2.821)	(1.788)	(0.388)	(0.673)	(0.533)
Dummy for the jobless	215.507***	246.251***	11.417***	−12.936***	−0.849
	(7.623)	(5.340)	(1.222)	(1.858)	(1.361)
Dummy for students	243.981***	5.803	16.511***	−21.082***	24.795***
	(16.434)	(7.254)	(3.807)	(3.210)	(4.471)
Dummy for males	4.019	−53.876***	5.413***	−7.719***	−19.615***
	(7.264)	(4.049)	(1.288)	(1.473)	(1.439)
Dummy for the married	−40.504***	52.872***	−19.474***	14.277***	−28.272***
	(8.234)	(5.559)	(1.257)	(1.877)	(1.592)
Dummy for urban residents	12.385*	22.815***	−1.383	3.153*	−2.942**
	(7.478)	(4.856)	(0.979)	(1.839)	(1.181)
No of observations	96,909	96,909	96,909	96,909	96,909

Note: Standard errors are in parentheses.

*p < 0.10, **p < 0.05, ***p < 0.01.

Table 5.

Estimation results for Equation (2).

	Model 1	Model 2	Model 3	Model 4	Model 5
Dep. Var.	All contents	Television programs	MVUs	Telecom services	CMSs
Time trend (TM)	14.499***	−0.410	4.133***	−8.789***	9.522***
	(0.983)	(0.511)	(0.163)	(0.216)	(0.170)
Time spent at home (HST)	0.101***	0.223***	0.016***	−0.014***	0.007***
	(0.005)	(0.003)	(0.001)	(0.001)	(0.001)
Dummy for COVID-19 (COVID)	66.882***	−38.769***	53.801***	18.788***	1.272
	(6.069)	(3.426)	(1.924)	(1.386)	(1.539)
Age	3.431*	4.409***	7.006***	0.410	3.611***
	(1.819)	(0.914)	(0.338)	(0.314)	(0.324)
Age squared	−0.076***	0.019**	−0.094***	−0.002	−0.049***
	(0.020)	(0.010)	(0.004)	(0.003)	(0.003)
Years of schooling	10.837***	−8.787***	3.390***	5.190***	6.267***
	(1.763)	(0.771)	(0.373)	(0.256)	(0.287)
Income	−4.312***	−1.691***	0.964***	1.631***	1.090***
	(1.183)	(0.635)	(0.152)	(0.284)	(0.187)
Schooling years*Income	0.495***	0.154***	−0.087***	−0.072***	−0.086***
	(0.086)	(0.041)	(0.012)	(0.019)	(0.014)
Family size	−10.851**	−14.843***	0.357	−1.902***	3.327***
	(4.608)	(2.015)	(0.664)	(0.717)	(0.623)
Dummy for the jobless	129.344***	73.615***	1.922	−4.849**	−4.916***
	(9.397)	(5.591)	(1.717)	(2.071)	(1.714)
Dummy for students	196.431***	−43.257***	5.317	−22.949***	32.125***
	(19.874)	(7.289)	(4.623)	(3.626)	(5.146)
Dummy for males	33.372***	−33.914***	8.381***	−9.525***	−16.542***
	(7.712)	(4.067)	(1.467)	(1.539)	(1.432)
Dummy for the married	−24.625*	35.603***	−21.451***	20.099***	−33.242***
	(13.068)	(6.100)	(1.906)	(2.126)	(1.904)
Dummy for urban residents	28.716***	36.578***	−3.285***	0.116	0.509
	(10.277)	(5.194)	(1.267)	(1.978)	(1.329)
No of observations	96,909	96,909	96,909	96,909	96,909

Note: Standard errors are in parentheses.

* p < 0.10, ** p < 0.05, *** p < 0.01

In Table 4, Table 5, the parameter estimate of a time trend (TM) is positive in Models 1, 3, and 5, and is negative in Model 4. These findings indicate that the use of all contents, MVUs, and CMSs shows an upward trend, whereas the use of traditional telecom services shows a downward trend, supporting H1 and H2. For example, Table 5 indicates that people's total content usage time tends to increase by approximately 15.0 min each year, with other variables being controlled for. In contrast, the parameter estimate of TM is negative but is not statistically significant in Model 2 of Table 5, providing clues to RQ1. That is, it indicates that the usage time of television programs did not show any decreasing trend over time in Korea, partly because of the advent of a new connection method such as online video distribution, and partly because an increasing number of people are accustomed to watching television programs on their smartphones.

In Table 5, the parameter estimate of the HST is always highly significant and has a positive sign in all models, except for Model 4. This finding implies that an increase in the time spent at home led to the increasing use of television programs, MVUs, and CMSs and the decreasing use of traditional telecom services. For example, an increase of 1 h in time spent at home, on average, led to an increase of approximately 13.4 min (=60 min ⅹ 0.223) in time spent watching television programs. In addition, the results show that the more time Korean people spend at home, the higher their total content usage time. These results are consistent with H3 and H4.

As indicated previously, we should place special attention to the parameter estimates of the dummy variable for the COVID-19 outbreak (COVID). In Table 4, COVID has a statistically significant and positively signed parameter estimate in all the models. Specifically, the results indicate an increase of around 85.7 min in the total content usage time, as well as an increase of 7.0 min, 59.0 min, 15.9 min, and 2.7 min in the usage time of television programs, MVUs, traditional telecom services, and CMSs, respectively, after the onset of COVID-19. That is, for whatever reason, average Korean people increased their use of these contents after the pandemic. Interestingly, average Korean people increased their usage time of MVUs, including Netflix and YouTube, more than other principal content, as they spent more time at home because of the pandemic. This finding reflects the increasing popularity of OTT services in Korea after the pandemic. In addition, to avoid face-to-face contact with others, average Koreans have become more dependent on traditional telecom services than before.

When the time spent at home (HST) is considered in Table 5, the parameter estimate for COVID in Model 1 changed from 85.652 in Tables 4 to 66.8826 In other words, the non-location effect due to the pandemic is likely to be greater than the location effect. In other words, average Korean people have changed their total content usage time more because of changes in non-location factors, presumably changes in their psychological state, and/or more search for health-related information, than because of changes in places to stay due to the pandemic. Similarly, a relatively small change in the parameter estimate of COVID in Models 3 and 4 between the two tables suggests that the effect of COVID-19 on the usage time of MVUs and traditional telecom services is mainly attributable to non-location factors. In contrast, the parameter estimate of COVID in Model 2 becomes negative in Table 5, suggesting that the effect of COVID-19 on television program viewing time is mainly attributable to its impact on the time spent at home. Finally, in Table 5, the results indicate no statistically significant changes in the usage time of CMSs due to the pandemic, implying that the increase in the usage time of CMSs in 2020, confirmed in Table 2, is in line with the previous serial trend and not the effect of COVID-19.

The estimated parameters for the demographic variables are mostly consistent with a priori expectations. We present some notable results, particularly the confirmed patterns, below. The parameter estimates of age and age squared in Models 1, 3, and 5 of Table 5 demonstrate that the amount of time people spent on all contents, MVUs, and CMSs tends to increase with age, but decreases after a certain age. For example, the average Korean spends more time on all contents as they get older up to the age of 45, but less time thereafter. In contrast, the average Korean spends more and more time on television programs as they get older. The use of traditional telecom services does not depend on age. Table 5 shows that the more educated people are, the less (more) time they spend on television programs (all the remaining principal contents). In addition, the higher the income level, the more (less) time is devoted to MVUs, traditional telecom services, and CMSs (television programs). The parameter estimates of the interaction term between schooling years and income indicate that educational level offsets the negative or positive marginal effect of income on content use.

The dummy variable for students in Model 1 has a larger parameter estimate than that for the jobless, with both parameter estimates being positive, indicating that the total content usage time is higher in this order: students, jobless, and employed. The results indicate that jobless people tend to spend more (less) time on television programs (traditional telecom services and CMSs) than others do. Males spend more (less) time on MVUs (television programs, traditional telecom services, and CMSs) than females do. Married persons spend more time on television programs and traditional telecom services than single people do. Finally, urban residents spend more time on television programs than rural residents do. The other parameter estimates are interpreted similarly.

4.3. Estimation results using balanced panel data

The use of unbalanced panel data may produce biased results, especially because the respondents added in 2019 and 2020 have different content usage patterns from the existing respondents. Table 6 reports the estimation results using data from survey participants whose records exist for all analysis years. That is, the data used in Table 6 are balanced panel data, the cases of which have been reduced to 43,740. The results demonstrate almost no differences in the sign and statistical significance of the parameter estimates between Table 5, Table 6, confirming the previous results. For example, the parameter estimate of a time trend (TM) in Model 1 is positive and statistically significant in the two tables. By contrast, as expected, the size of the parameter estimates changes between the two tables, albeit slightly in many cases. In particular, the parameter estimates of the dummy variable for the COVID-19 outbreak period (COVID) in Table 6 decreased in absolute value compared to those in Table 5, except for Model 5. The balanced panel data exclude all respondents added to the KMP in 2019 and 2020, indicating the likelihood that these additional respondents may experience different location effects from existing respondents.

Table 6.

Estimation results for Equation (2) using balanced panel data.

	Model 1	Model 2	Model 3	Model 4	Model 5
Dep. Var.	All contents	Television programs	MVUs	Telecom services	CMSs
Time trend (TM)	19.779***	0.096	3.327***	−8.937***	10.939***
	(1.503)	(0.814)	(0.192)	(0.326)	(0.253)
Time spent at home (HST)	0.100***	0.225***	0.012***	−0.013***	0.005***
	(0.007)	(0.004)	(0.001)	(0.002)	(0.001)
Dummy for COVID-19 (COVID)	20.034**	−18.266***	32.947***	15.699***	−5.460***
	(8.118)	(4.954)	(2.211)	(1.973)	(2.014)
Age	−3.219	5.851***	2.707***	−2.180***	2.412***
	(2.820)	(1.366)	(0.359)	(0.494)	(0.438)
Age squared	−0.031	−0.003	−0.038***	0.031***	−0.045***
	(0.031)	(0.015)	(0.003)	(0.005)	(0.004)
Years of schooling	6.727**	−10.403***	2.376***	6.282***	4.674***
	(3.020)	(1.212)	(0.372)	(0.465)	(0.413)
Income	−5.129***	−1.929**	0.712***	1.041**	1.081***
	(1.698)	(0.869)	(0.178)	(0.422)	(0.297)
Schooling years*Income	0.553***	0.164***	−0.062***	−0.034	−0.075***
	(0.127)	(0.059)	(0.014)	(0.030)	(0.021)
Family size	−22.223***	−23.026***	1.084*	−2.066*	2.568***
	(6.698)	(3.001)	(0.555)	(1.140)	(0.907)
Dummy for the jobless	125.912***	79.760***	−0.321	−4.591	−4.380*
	(12.797)	(7.608)	(1.850)	(2.882)	(2.263)
Dummy for students	226.233***	−40.921***	5.065	−33.775***	25.430***
	(30.761)	(11.224)	(6.293)	(5.065)	(7.524)
Dummy for males	39.770***	−25.346***	5.434***	−9.965***	−16.323***
	(12.968)	(6.971)	(1.635)	(2.695)	(2.240)
Dummy for the married	−22.616	30.549***	−20.386***	18.769***	−20.901***
	(18.661)	(9.082)	(2.295)	(3.709)	(2.672)
Dummy for urban residents	49.750***	36.267***	−2.365**	2.507	3.258**
	(13.013)	(6.887)	(1.175)	(2.616)	(1.656)
No of observations	43,740	43,740	43,740	43,740	43,740

Note: Standard errors are in parentheses.

* p < 0.10, ** p < 0.05, *** p < 0.01

5. Conclusions

Using KMP data for the period 2011–2020, this study examined changes in content usage time due to the onset of COVID-19 in Korea and explored the reasons for these changes. We have decomposed changes in content usage time due to the pandemic into two components: location and non-location effect. The empirical results indicate an increase in the usage time of the four principal contents as well as the total content usage time because of the pandemic. The results also demonstrate that average Korean people spent approximately 157 min longer at home in three days after the onset of the pandemic in 2020 than in the period 2017–2019, leading to an increase in the time spent on all the principal contents except for traditional telecom services, as well as an increase in the total content usage time. When this location effect is considered, the study presents the likelihood that the pandemic had an impact on an individual's total content usage time, mainly through its impact on non-location factors, such as their psychological state and/or need for health-related information. The same is true for the usage times for MVUs and traditional telecom services. In contrast, this study suggests that the effect of the pandemic on television program viewing time was mainly caused by its impact on the time spent at home.

These results provide a clue to understanding the changes in content usage behavior after the pandemic. First, since the total content usage time increased abnormally because of the pandemic, it is likely to temporarily decrease thereafter. The same pattern is predictable for the usage time of individual contents. Second, we anticipate the usage time of MVUs to increase after their temporary decrease, whereas that of traditional telecom services will experience a steady decrease over time. Finally, we predict a steady increase in the use of CMSs, regardless of COVID-19, showing the largest growth rate among the major media. Based on these predictions, we can present strategic suggestions for the parties involved. Television programs will still maintain their position as the most important medium even after the end of the pandemic, but television program viewing time is expected to decrease among some consumer groups, such as students, those in their 20–30s, and the highly educated. These consumer groups tend to spend more time on MVUs and CMSs. In addition, the downward trend in the use of traditional telecom services has been temporarily halted because of the pandemic but is anticipated to decline again after the end of the pandemic. Therefore, broadcasters and telecom companies need to establish their own media strategies in response to this trend.

For decades, media research has explored why people use specific mediums or content and what gratifications they receive from it. In particular, the uses and gratification theory holds that people choose media to meet their desires and needs to satisfy gratification, focusing on people's needs as drivers for media consumption (Cha & Chan-Olmsted, 2012; Katz et al., 1973; Ruggiero, 2000; Stafford et al., 2004). As new digital technologies present people with more media choices, motivation and satisfaction become crucial components of media analysis (Ruggiero, 2000), accelerating the application of uses and gratification theory to research on substitution between traditional and new media. Similar to new digital technologies, the spread of COVID-19 has changed people's need for content through its impact on time spent at home and non-location-related factors, which requires further research. In contrast, despite the presence of extensive studies from various perspectives on the effects of COVID-19, no or few studies have been reported regarding its effect on the usage of contents to date. We performed this study to fill this gap. This study contributes to the literature on COVID-19 and media usage by examining the pandemic's effects on content usage behavior. The decomposition of the change in content usage time into location and non-location effects can also be regarded as a new contribution to the literature on content usage. By analyzing changes in the usage time of several principal contents, this study provides new empirical evidence for media substitution.

However, as with other studies, this study has some limitations. Above all, this study does not provide a sufficient explanation for the reasons or motivations for the use of each content in the location and non-location effects. For example, the study indicates that the more time people spend at home, the more time they watch television programs, but the reason for this is not explained. This may be because of an increased need for entertainment or relaxation, or an increased demand for information to communicate with others. That is, the same content can be used for different needs depending on people's previous experiences. Therefore, research on the reasons for or drivers of content usage using more abundant data is necessary. Second, the study relies on one-year data on media usage after the COVID-19 outbreak, which may produce biased results. In addition, the data used in the analysis were collected during the incubation period with a small number of confirmed infections, implying no drastic changes in either the time spent at various locations or media usage time. If the data are collected during a widespread period with a large number of confirmed cases, different results can be obtained. Finally, this study does not sufficiently explore the factors causing non-location effects because of the lack of data. Therefore, further research on the non-location effect, such as the relationship between people's psychological states and content usage, deserves attention.

Funding source

This work was supported by the 2022 Research Fund of the University of Seoul.

Appendix Table 1 Annual changes in the average time spent at various locations (unit: minutes)

	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020
Home	2,715.8	2,695.2	2,629.4	2,632.3	2,630.9	2,638.9	2,574.6	2,607.7	2,570.2	2,741.2
Employed persons	2,339.6	2,320.4	2,268.6	2,266.7	2,304.8	2,304.5	2,274.3	2,303.0	2,279.4	2,344.0
Jobless persons	3,536.8	3,551.3	3,465.0	3,447.1	3,427.9	3,420.7	3,365.5	3,353.7	3,328.5	3,559.4
Students	2,487.5	2,532.8	2,456.8	2,533.9	2,530.5	2,586.4	2,488.2	2,617.2	2,532.1	3,073.1
Workplaces	752.5	816.6	829.5	834.8	808.7	842.7	849.3	801.4	803.1	818.1
Schools	338.5	298.3	291.7	268.0	249.5	250.5	247.5	225.3	216.4	112.6
Transportation	204.4	195.4	227.1	231.4	228.4	224.3	234.7	247.1	252.0	230.8
Restaurants	53.3	60.1	72.5	83.3	85.9	83.5	103.6	115.4	119.8	98.3
Commerce facilities	53.9	60.1	62.2	70.5	66.0	62.0	68.8	70.0	65.8	59.7
Place for leisure	50.1	49.5	51.7	57.3	59.9	54.3	58.5	64.5	61.3	41.4
Others	122.0	103.9	107.5	101.0	98.9	109.2	101.5	104.5	123.0	78.8

Note: All numbers are based on the sum of the time spent in three days. Transportation includes both public and private transit. Places for leisure include entertainment, recreation, sports and cultural facilities. Employed persons include self-employed persons and jobless persons exclude students.

Appendix Table 2 Annual changes in the average content usage time (unit: minutes)

	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020
All contents	1,203.2	1,182.0	1,201.4	1,188.3	1,229.3	1,193.8	1,208.9	1,290.9	1,239.5	1,333.5
Employed persons	1,148.3	1,126.2	1,129.2	1,104.3	1,182.3	1,136.6	1,165.1	1,235.6	1,212.4	1,305.5
Jobless persons	1,270.1	1,225.8	1,250.8	1,239.7	1,281.8	1,226.7	1,255.1	1,318.9	1,289.0	1,373.5
Students	1,236.1	1,259.9	1,322.6	1,345.3	1,296.2	1,316.6	1,286.0	1,427.7	1,264.7	1,381.6
Television programs	538.2	512.9	510.9	506.7	519.0	499.1	486.2	509.2	487.5	501.6
Radio/music channels	26.0	28.3	41.0	32.2	48.7	40.8	39.4	46.8	43.3	46.3
Movies/videos/UCC	10.3	10.1	10.2	11.9	15.6	18.2	20.9	32.3	39.6	90.6
Music/digital soundtracks	21.4	15.1	24.2	24.9	27.4	30.4	37.8	42.2	42.7	49.9
Newspaper articles	42.8	36.8	25.0	21.6	16.3	13.3	13.4	12.5	15.0	25.5
Books/magazines	137.7	146.1	152.2	140.0	132.0	132.3	125.8	135.0	104.4	74.5
Traditional telecom services	212.6	204.3	153.0	156.5	155.0	151.4	146.6	153.2	142.0	143.0
Chatting/messenger/SNS	17.2	33.0	65.4	70.4	81.9	82.6	90.1	98.3	89.9	114.4
Information content	50.7	49.3	57.8	60.8	60.2	50.8	54.0	55.5	56.5	60.8
Online commerce	9.2	9.9	6.0	5.9	6.3	5.4	8.9	8.6	9.8	12.6
Games	39.7	36.1	39.7	37.6	37.7	35.8	37.4	45.9	48.8	53.2
Work-related content	88.4	90.5	106.3	107.2	120.5	124.6	136.9	139.0	147.8	153.2
Cultural content	5.5	5.4	6.5	9.2	6.9	6.6	8.5	10.3	9.5	3.6
Others	3.5	4.2	3.2	3.3	1.8	2.3	2.9	2.2	2.7	4.4

^{Note: All numbers are based on the sum of primary and simultaneous usage time in three days. Television programs include both real-time programs and videos on demand. Employed persons include self-employed persons and jobless persons exclude students.}