Spread of Online Chatting and Global Developments in Reading Literacy during 2000–2018

Abstract

This chapter addresses the relationship between increased online chatting during 2000–2018 and changes in reading literacy. The findings relate to per-country changes and involve 39 countries from five continents. The data was derived from the international PISA surveys. Two groups of countries are distinguished: those with a low prevalence of online chatting in 2009 and accelerating growth during 2009–2018 and those with high chatting prevalence in 2009 but decreasing growth afterward. The first group shows substantially improved reading literacy during 2000–2009. However, this positive trend changed around in the subsequent period when online chatting clearly accelerated in these countries. In contrast, the second group shows improvement during 2009–2018, when the growth in online chatting decelerated in these countries. By 2009, the prevalence of online chatting was already high in these countries. This coincided with a small decrease in reading literacy. Additional analyses indicate that during 2000–2009 the relationship between increased online chatting and decreased reading literacy is mediated by changing percentages of students reading fiction at least once a month. During 2009–2108, the relationship is mediated by changing awareness of useful reading strategies.

1. Introduction

The rise of the internet in the final years of the previous century has changed people’s reading habits profoundly. As the use of digital media spread, reading from paper lost ground to reading from computer screens and smartphone displays. Authors like Carr [1] and Wolfe [2] have argued that digital media use promotes superficial reading strategies, like skimming and browsing instead of more thorough and time-consuming approaches. For today’s youth, the development of more demanding reading strategies may be at risk. As a result, superficial reading may become the norm.

Before the internet, books, newspapers, and magazines were the main reading materials. In comparison to the usually short and fragmented texts on the internet, writings on paper are long but also well-structured and continuous in nature. Textual information on the internet is often scattered over various web pages in complex networks.

2. Focus on online chatting

Among several other types of reading-related ICT activities such as taking part in online group discussions, searching for information online, reading emails, and reading online news online chatting stands out. In the first two decades of the twenty-first century, it has become more popular than any of the aforementioned activities among 15-year-olds in dozens of countries [3]. In addition, the rate of growth varies considerably between countries. By 2018, a large majority (over 70%) of 15-year-olds in the 39 countries included in this study engaged in online chatting on a daily basis (see Table 1).

	Low chatting prevalence in 2009 (< 35%)		High chatting prevalence in 2009 (> 35%)		Overall across groups
	Mean	SD	Mean	SD	Mean	SD
Reading literacy
Change 2000–2009	21.2	16.9	−1.2	16.0	8.8	19.8
Change 2009–2018	−9.9	19.0	9.2	14.0	.7	18.8
Percentage of students reading fiction at least once a month
Change 2000–2009	15.3%	10.7	7.1%	4.6	10.8%	8.8
Change 2009–2018*	−2.5%	4.6	−3.9%	3.6	−3.3%	4.1
Awareness of useful reading strategies
Change 2009–2018	−.015	.124	.089	.094	.042	.119
Percentage of students chatting daily
Prevalence in 2009	18.4%	12.6	44.7%	8.3	32.9%	16.8
Change 2009–2018	53.8%	13.8	26.1%	8.3	38.5%	17.8

Table 1.

Reading literacy, reading fiction, reading strategies, and online chatting.

*

The difference between both groups is not statistically significant (p = .227).

Countries are weighed by the number of students they represent.

All other differences are statistically significant (p < .05; two-tailed).

Online chatting relates to the exchange of written messages over the internet in real time. The messages are usually short and informal. Together with the rapid exchange, this fosters a type of communication similar to spoken conversation. It also sets online chatting apart from exchanges via email or internet forums. The increased prevalence of online chatting is indicative of changed reading practices. Traditional long and well-structured texts on paper have lost ground to short fragments on screen.

This chapter describes the relationship between per-country patterns in the spread of online chatting and the development of reading literacy of 15-year-olds. It reports the findings from an analysis that exploits the variation in growth rates between countries.

3. Prior research

Randomized controlled trials clearly indicate that reading comprehension is higher when texts are read from paper than from computer screens [4, 5, 6]. Still, large-scale survey findings fail to provide support for these findings. Data from the international PISA surveys indicate even slightly higher scores on the reading tests for 15-year-olds that frequently engage in online chatting [7]. Still, the notion that reading from screen hampers comprehension is supported by the finding that fewer students provide correct responses when test items are presented on computer screens instead in a paper format [8, 9, 10].

Unfortunately, the PISA surveys do not track individual reading literacy development over time. In each of the seven triennial surveys conducted during 2000–2018, the reading literacy of 15-year-olds in a particular year has been assessed for dozens of countries. Therefore, it is possible to map per-country changes in reading literacy. With regard to individual students, however, the surveys only yield cross-sectional data. Perhaps an effect of online chatting would be discerned if individual growth in reading literacy were assessed. Undoubtedly, there is an urgent need for research into the long-term effects of digital media on reading literacy [2].

4. The present study

The focus here is on per-country growth rates in online chatting and reading literacy trends. Their correlation points to an effect of changing environments. Superficial reading strategies may become the default for all individuals in a shared environment as online chatting spreads. In that case, everyone is affected to a similar extent. Such an effect would be similar to the effect of diesel and gasoline cars on health. Within a certain area, everyone is affected by exhaust gases. Individual variation in car use is hardly relevant.

The findings to be reported relate to changes during 2000–2018 and involve 39 countries from five continents. See Figure 1 for an overview. Table A1 provides a detailed list of the countries included. Special attention is paid to distinct patterns before and after 2009. The 2009 PISA survey was the first after the introduction of the iPhone in 2007. This innovation stimulated the exchange of information over the internet even further, in this case via compact hand-held devices. Moreover, in 2009 WhatsApp was launched. In a short time, this became the most popular online chatting tool in many countries across the world. In this study two groups of countries are compared:

1. Countries with high chatting prevalence in 2009 (more than 35% of 15-year-olds chatting daily) and decelerating growth afterward.

2. Countries with low chatting prevalence in 2009 (less than 35% of 15-year-olds chatting daily) and accelerating growth afterward.

Figure 1.

Countries included by chatting prevalence in 2009.

The first group includes 29 countries and represents 55.2% of all students across the entire set of 39 countries. The second group consists of 10 countries representing 44.8% of all students. Figure 1 shows that the group of countries with low chatting prevalence in 2009 mainly involves countries in East Asia, Southeast Asia, and Latin America. European and North American countries dominate the group with high chatting prevalence in 2009.

Both groups will be compared with regard to reading literacy trends during 2000–2009 and 2009–2018. In addition, correlations between per-country reading literacy trends and (changes) in chatting prevalence are reported. This study also aims to assess to what extent these correlations are mediated by changes in the frequency of reading fiction and awareness of useful reading strategies. It seems plausible that increases in online chatting coincide with decreases in the reading of fiction. Online chatting involves swift exchanges of short and informal messages, whereas reading fiction requires (much) more time, effort, and patience. Given that online chat messages are usually (very) short, it also seems likely that increased online chatting promotes superficial reading and hampers the development of time-consuming reading strategies that allow for deeper comprehension. Prior research does report higher levels of reading comprehension among students that engage frequently in reading fiction [11]. Likewise, individuals that use appropriate reading strategies perform better on reading tests [12]. However, the novelty of this study, however, is that it shows to what extent per-country changes in these variables coincide with changes in reading literacy over time. The correlations’ robustness is established through several additional data analyses controlling for potentially confounding variables.

4.1 The international PISA surveys

The Programme for International Student Assessment (PISA) is an Organisation for Economic Co-operation and Development (OECD) initiative. During 2000–2018 seven surveys have been conducted every 3 years. Due to the COVID pandemic, the survey planned for 2021 was postponed to 2022. Its primary objective is to assess the knowledge and skills of 15-year-olds in reading, mathematics, and science. In addition to cognitive tests, the data collection involves student questionnaires on school experiences and general background. PISA started in 2000 with 32 countries. Ten additional countries joined in 2002 [13].

Reading literacy was the main focus in 2000/2002, 2009, and 2018. In other years, reading literacy was assessed as well. However, most test items involved either mathematics or science in these surveys. Only when the surveys focused on reading, the student questionnaires addressed reading attitudes and reading-related activities in addition to general background questions.

The PISA data collection is highly standardized and therefore well-suited for cross-national comparisons. Across countries, students in the same age range are targeted. Age rather than grade level determines whether students are included in the national samples.

The present study involves 39 countries, which all took part in the first PISA survey. Two countries that participated in PISA 2000 were excluded. The participation in PISA of Liechtenstein ended in 2012. The Netherlands failed to meet all sample requirements in 2000. Romania, which joined in 2002, was excluded because the age range of the sampled students was not in line with that of the other countries.

To account for the variation in size among countries weights are applied. The samples of countries like Brazil, Indonesia, and the United States represent over two million students per year. On the other hand, for countries like Iceland and Luxembourg, this number is approximately five thousand. Country weights are based on the total number of students the national samples represent, averaged across years. The weights are standardized in such a way that their average equals one. This precludes that the statistical significance of the findings is overestimated. In the absence of weights, the European countries would dominate the results although they account for only 31% of all students represented by the national samples. The study includes 24 European countries. This would amount to 61% according to a one-country one-vote principle. For more details, see Table A1.

4.2 Measurements

The cognitive tests in PISA take up approximately 2 hours. The reading test addresses four main reading processes (reading fluently, locating information, understanding, and evaluating/reflecting) and includes several types of texts (e.g., narrations, argumentations, instructions). Test development is based on a conceptual framework that has been adapted a few times over the years [14]. Attention to digital environments and multiple resources was new to the 2018 reading assessment. An additional adaptation involves the switch from paper-and- pencil tests to computer-based ones in 2015. As a result, adjustments for “mode effects” were needed to take into account that computer-based test versions usually yield fewer correct responses. Notwithstanding these changes, assessment of trends in reading literacy over time is possible [15]. The average reading score across OECD countries is approximately 500 with a standard deviation around 100 points.

For this study, country-specific reading literacy trends were estimated for 2000–2009 and 2009–2018. These linear trends are based on each country’s reading literacy averages in 2000, 2002, 2003, 2006, 2009, 2012, 2015, and 2018. A detailed account of reading trends, online chatting, and weights per country is provided in Table A2.

The main explanatory variable in the data analysis is per-country online chatting prevalence. In 2009 and 2018, students indicated how often they engaged in online chatting. The percentages of 15-year-olds that reported chatting several times a day are used to gauge changes in online chatting. Information on online chatting is lacking for 2000/2002. The 2009 percentages are considered to approximate growth during 2000–2009. It seems plausible to assume that at the turn of the century, the percentage of 15-year-olds chatting several a day was close to zero in each and every country. Chat platforms were around at the time, but it required much more effort and patience to get access to the internet than it does today. Connections were mainly made on desktop computers, via modems and analog telephone lines.

The student surveys in 2000/2002, 2009, and 2018 contained several items on reading activities. Using the answers related to reading fiction (including stories, novels, and narratives), per-country changes during 2000–2009 and 2009–2018 were calculated. These relate to the percentages of pupils who indicated that they read fiction at least once a month on a voluntary basis. For the countries that joined the first PISA survey in 2002, the difference between 2002 and 2009 was increased by ⁹/₇ in order to estimate change during 2000–2009.

Only the 2009 and 2018 questionnaires included items on reading strategies. As a result, changes in this variable can only be determined for the 2009–2018 period. Students were asked to rate the usefulness of six different strategies for understanding and memorizing a text (e.g., I underline important parts of the text). Next, it was established to what extent the student ratings concurred with a ranking established by reading experts [16]. Finally, this resulted in a score with zero mean and a standard deviation equal to one in OECD countries. High scores indicate student ratings that are in line with the experts’ ranking.

5. Findings

5.1 Comparison between the two groups of countries

Table 1 shows, by country group, the changes in reading literacy, percentages of students who read fiction at least once a month, awareness of useful reading strategies, and percentages of students who chat daily, both from 2000 to 2009 and 2009–2018. The bottom rows on online chatting show that the growth in online chatting strongly accelerated during 2009–2018 for the group of countries with low chatting prevalence in 2009. In contrast, the growth decelerated during 2009–2018 for the group with high chatting prevalence in 2009.

The countries with low chatting prevalence in 2009 showed substantial improvement in reading literacy during 2000–2009. The countries with high chatting prevalence in 2009 showed no improvement at all in the same period. The countries with low chatting prevalence in 2009 witnessed a strong acceleration in online chatting during 2009–2018. At the same time, reading literacy dropped. After a 21.2-point improvement during 2000–2009, reading literacy scores dropped by 9.9 points in the subsequent period. Online chatting growth decelerated during 2009–2018 in the countries with high chatting prevalence in 2009. Along with this deceleration, reading literacy improved by 9.2 points.

Percentages of students reading fiction at least once a month were on the rise in both groups of countries during 2000–2009. But, the increase was over twice as large in countries with low chatting prevalence in 2009 (15.3% points vs. 7.1%). The 2009–2018 period shows a modest decrease in reading fiction for both country groups. However, the difference between both groups is not statistically significant.

The countries with low chatting prevalence in 2009—and accelerated growth in the subsequent period—show a slight decline in awareness of useful reading strategies, whereas the countries with decelerating growth in online chatting during 2009–2018 show a clear improvement in awareness of useful reading strategies in that period.

5.2 Correlations between changes in the key variables

Table 2 shows the correlations (Pearson’s r) between changes in the key variables. The correlations below the diagonal involve changes during 2000–2009. The ones above the diagonal represent correlations between changes during 2009–2018. Figures 2 and 3 provide a graphical display of the correlations between online chatting and changes in reading literacy during 2000–2009 and 2009–2018. Both graphs also show the regression lines that indicate to what extent reading literacy declines as online chatting increases. The following equations describe the bivariate linear regression equations for 2000–2009 and 2009–2018:

	Percentage of students chatting daily	Reading literacy	Percentage of students reading fiction	Awareness of useful reading strategies
Percentage of students chatting daily	—	−.660	.052*	−.552
Reading literacy	−.574	—	.057*	.726
Percentage of students reading fiction	−.770	.588	—	−.098*

Table 2.

Correlations (Pearson’s r) between changes during 2000–2009 (below the diagonal) and 2009–2018 (above).

*

None of the correlations that involve reading fiction are significant for p < .05.

Countries are weighed by the number of students they represent.

Correlations for 2000–2009 are shown below the diagonal.

Correlations for 2009–2018 are shown above the diagonal.

Figure 2.

Online chatting prevalence 2009 and per-country reading change 2000–2009. The number of students per country determines the size of each bubble.

Figure 3.

Change chatting 2009–2018 and per-country reading change 2009–2018. The number of students per country determines the size of each bubble.

$$2000–2009:Changereadingliteracy=31.038-.676\times Chattingprevalence$$E1

$$2009–2018:Changereadingliteracy=27.633-.700\times Changeinchatting$$E2

The above equations show the unstandardized regression coefficients. In a bivariate regression analysis, the standardized regression coefficient (beta) is equal to the correlation between the explanatory variable and the outcome variable. These correlations are reported in Table 2.

The unstandardized coefficients indicate that reading literacy drops by about seven-tenths of a point at a 1% increase in online chatting. The constants (intercepts) in the equations denote the expected reading literacy increase in the absence of any change in online chatting. Both intercepts are close to 30. This suggests that without the rise of online chatting, reading literacy would have increased by about 30 points in both periods. It can also be inferred from the equations that a 46% chatting prevalence in 2009 would coincide with zero reading change during 2000–2009. Belgium fits this scenario closely. By 2009, 45.8% of the Belgian 15-year-olds were engaged in online chatting several times a day. The change in reading literacy between 2000 and 2009 for Belgium is minimal (0.7 points, see Table A2). During 2009–2018, a 39% increase in chatting would match a situation without any change in reading literacy. The findings for Chile are in line with this scenario. Reading literacy improved only marginally in Chile during 2009–2018 (3.6 points). At the same time, online chatting increased by 38.3 percent points.

The figures in Table 2 confirm that in both periods the correlation between (changes in) the percentage of students chatting online and reading literacy changes is strong and negative (−.574; −.660). The correlations suggest that during 2000–2009 change in reading fiction might present an intermediary factor in the relationship between online chatting and reading literacy. Change in reading fiction correlates strongly with both online chatting and reading literacy. Change in awareness of useful reading strategies seems to assume this role during 2009–2018.

5.3 The search for mediating and confounding variables

A number of regression analyses have been conducted to further explore the conjecture that reading fiction and reading strategies function as mediating factors. If the relation between online chatting and reading literacy is (partially) mediated by reading fiction or reading strategies, multiple regression analyses with both online chatting and reading fiction/reading strategies as explanatory variables will show a considerably reduced coefficient for online chatting. The high correlations between online chatting and changes in reading fiction (during 2000–2009) or reading strategies (during 2009–2018) suggest strong relationships between online chatting and both variables. If multiple regression analysis fails to show strong coefficients for online chatting, the findings may still be in line with an indirect effect. Online chatting is clearly related to decreases in percentages of students reading fiction and awareness of useful reading strategies. In turn, these variables may show strong coefficients in a multiple regression analysis.

In the next step, it is investigated to what extent 15 possibly confounding variables can account for the correlations in Table 2. They include societal changes (e.g., regarding parents’ education levels, internet availability), educational changes (e.g., student-teacher ratios), changes in sample characteristics (e.g., coverage of all 15-year-olds), and additional changes in reading activities and attitudes (e.g., enjoyment of reading). See Table A3 for a complete list of the control variables. By means of stepwise regression, it is established to what extent they are related to changes in reading literacy, reading fiction, and awareness of reading strategies. With regard to reading fiction and reading strategies, changes in reading attitudes and reading activities were not included as control variables. These variables can be either effects or causes of reading fiction and awareness of useful reading strategies.

Table 3 shows the results of the first set of multiple regression analyses. For the 2000–2009 period, the findings show that the coefficient of online chatting prevalence in 2009 with change in reading literacy is no longer statistically significant when change in reading fiction is included as well. The p-value for the regression coefficient of reading fiction can still be considered significant at the .05 level in a one-tailed test. This conclusion is warranted as the coefficient indicates a positive relationship, which is in line with expectations. Regarding the 2009–2018 period, both the regression coefficients of change in online chatting and awareness of useful reading strategies reach statistical significance. However, also in this case the standardized regression coefficient of online chatting is considerably smaller than the bivariate correlation in Table 2 (−.373 vs. -.660).

	beta	p-value	R^2
Outcome: Change reading literacy 2000–2009
Prevalence online chatting 2009	−.297	.156
Change reading fiction 2000–2009	.359	.089	.382
Outcome: Change reading literacy 2009–2018
Change online chatting 2009–2018	−.373	.004
Change reading strategies 2009–2018	.520	<.001	.624

Table 3.

Multiple regression analysis—Standardized coefficients (beta) of online chatting and reading fiction/reading strategies.

Countries are weighed by the number of students they represent.

p-values are based on two-tailed tests.

Table 4 shows the results of the stepwise regression for the 2000–2009 period. With regard to change in reading fiction, stepwise regression fails to produce any significant coefficients in addition to the prevalence of online chatting in 2009. None of the possibly confounding factors that were considered can account for changes in the percentages of students reading fiction at least once a month.

	beta	p-value	R^2
Outcome: Change reading literacy 2000–2009
Prevalence online chatting 2009	—	—
Change reading fiction 2000–2009	.414	<.001
Change student-teacher ratio 2000–2009	−.413	<.001
Change coverage all 15-year-olds 2000–2009	.322	.003
Change frequency reading newspapers 2000–2009	.228	.028	.751
Outcome: Change reading fiction 2000–2009
Prevalence online chatting 2009	−.770	<.001	.594

Table 4.

Stepwise regression analysis regarding changes during 2000–2009.

Countries are weighed by the number of students they represent.

p-values are based on two-tailed tests.

For change in reading literacy, the stepwise regression first of all yields no significant coefficient for online chatting. This implies that when controlling for possibly confounding factors, no direct effect on online chatting can be discerned. On the other hand, the results do show a significant coefficient for reading fiction. This coefficient is smaller than the bivariate correlation (.414 vs. .588), but still substantial and statistically significant. Additional variables with a significant coefficient are changes in student-teacher ratios, coverage of all 15-year-olds, and frequency of reading newspapers.

All in all, the findings indicate a strong and negative effect of online chatting on changes in reading fiction, which in turn shows a substantial effect on changes in reading literacy during 2000–2009. In other words, the negative relationship between increased online chatting and reading literacy runs via decreased percentages of students reading fiction during 2000–2009.

The findings for 2009–2018 are reported in Table 5. With regard to change in reading strategies, the stepwise regression yields one additional variable with a significant coefficient, namely change in coverage of all 15-year-olds in the national samples. Awareness of useful reading strategies decreased in countries with increasing coverage during 2009–2018. However, the analysis also yields a significant and negative coefficient for change in online chatting. It differs only slightly from the bivariate correlation between online chatting and reading literacy (−.533 vs. -.552).

	beta	p-value	R^2
Outcome: Change reading literacy 2009–2018
Change online chatting 2009–2018	—	—
Change reading strategies 2009–2018	.584	<.001
Change parents’ occupation 2009–2018	.372	<.001
Change reading enjoyment 2009–2018	.281	.005	.700
Outcome: Change reading strategies 2009–2018
Change online chatting 2009–2018	−.516	<.001
Change coverage all 15-year-olds 2009–2018	−.553	<.001	.407

Table 5.

Stepwise regression analysis regarding changes during 2009–2018.

Countries are weighed by the number of students they represent.

p-values are based on two-tailed tests.

Regarding changes in reading literacy during 2009–2018, the stepwise regression again does not yield a significant effect of online chatting. But, the analysis does show a strong coefficient for changes in reading strategies. It is not as strong as the bivariate correlation (.584 vs. .726), but definitely substantial. Additional variables showing significant coefficients are changes in parents’ occupation and reading enjoyment.

Also for the 2009–2018 period, the findings are in line with the notion that the negative relationship between online chatting and changes in reading literacy is indirect. Increases in online chatting coincide with decreasing awareness of useful reading strategies. Subsequently, changing awareness of useful reading strategies coincide with changes in reading literacy.

6. Conclusion and discussion

The findings show that developments in reading literacy levels of 15-year-olds across the world in the first two decades of the twenty-first century are clearly linked to the speed with which online chatting has spread. Reading literacy markedly improved during 2000–2009 in countries with low percentages of 15-year-olds chatting on a daily basis. In the subsequent period, the prevalence of online chatting increased rapidly in these countries. This coincided with a clear decrease in reading literacy. In contrast, countries with high chatting prevalence in 2009 showed no improvement during 2000–2009. When the growth of online chatting clearly decelerated in these countries during 2009–2018, reading literacy improved. The relationship between online chatting and reading literacy was found to be mediated by changes in reading fiction during 2000–2009 and by changes in awareness of useful reading strategies during 2009–2018. The finding that these factors mediate the relationship between online chatting and reading literacy is clearly in line with theoretical considerations on the relationship between digital media use and reading literacy.

By 2009, online chatting had already become popular among 15-year-olds, especially in Europe and North America. In East Asia, Southeast Asia, and Latin America, the percentages of students chatting on a daily basis were still fairly moderate in 2009. It seems that the launch of the iPhone in 2007 and WhatsApp in 2009 triggered the rise of online chatting in those parts of the world.

The relationship between online chatting and reading literacy suggests a large effect. During 2000–2009, countries with low chatting prevalence showed an improvement of 21.2 points. Countries with high prevalence experienced a minute decrease (−1.2, see Table 1). This amounts to a difference of over 20 points on the PISA reading literacy scale. In the subsequent period, the difference between both groups amounts to 19 points (−9.9 vs. 9.2). For both periods, this implies a difference of about two-tenth standard deviation in the distribution of individual-level reading scores. In the field of education, this may be considered a large effect [17].

The data analysis also indicates that a moderate increase in online chatting does not necessarily pose a threat to reading literacy. An increase below 40 percentage points over a nine-year period was found to coincide with zero reading literacy change. This implies that at a limited increase per year (4–5%) reading literacy may remain stable. In most countries included in the present study, the percentage of students chatting several times a day already exceeded 60% in 2018. The Russian Federation and Peru are the only exceptions. This suggests that there may be not much room left for further increases. Per-country percentages of students chatting several times a day have proven to be a useful indicator to gauge the impact of digital media use on reading literacy in the first two decades of the present century. However, it seems likely that future research will need alternative and more fine-grained measures.

The present study involves correlations between changes in country aggregates over time. Therefore, the conclusions relate exclusively to the country level. The findings do not show that individuals that frequently engage in online chatting are poor readers. This study only shows that per-country increases in online chatting in the first two decades of the twenty-first century coincided with declining reading literacy trends. At the country level, the PISA data allow for an analysis of changes over time, but this is not possible for the individual-level PISA data. The correlations reported in this chapter point to the effects of changing environments, like the effect of car use on the prevalence of lung-related health problems. However, the present study stands out as it shows that per-country changes in online chatting, frequency of reading fiction, and awareness of useful reading strategies coincide with changes in reading literacy. The findings are based on correlations. As a result, it cannot be precluded that other variables that were not controlled for in data analysis could account for the correlations that were reported. The variables that were controlled for showed some additional correlations with reading literacy, frequency of reading fiction, and/or awareness of reading strategies. However, this did not lead to substantially different conclusions. In addition, it also possible that country-specific circumstances might account for changes in reading literacy, e.g., major curriculum changes or local teacher shortages.

The research findings indicate how the diffusion of online chatting has shaped per-country reading literacy developments. This faces the education sector with new challenges. In comparison to a few decades ago, today’s youth acquire fewer reading skills outside school. Even more time and effort will have to be put into reading education than before by schools and teachers to prevent further decline in traditional reading skills. Still, the impact of online chatting may turn out trivial compared to recent developments in the field of Artificial Intelligence (AI). Currently, chatbots are improving rapidly and can generate texts that are barely distinguishable for human readers from texts written by fellow humans. In addition, they are very well capable of responding to prompts and questions that are posed in natural language. AI is getting better and better both at generating and interpreting natural language. This raises the question of which reading and writing skills people in the future will still need. It seems conceivable that in the future exchange of information and ideas will largely take place via ICT tools developed through machine learning (i.e., computer algorithms that are fine-tuned to the extreme by computer algorithms). Human input may become more and more redundant.

It goes without saying that the spread of ICT has had an enormous impact on people’s lives all over the world. In many respects, life has become easier and more enjoyable thanks to countless innovative ICT applications. Still, every major transition comes with certain drawbacks. Loss of traditional reading literacy skills seems to be one of them. It would be unfortunate if thorough and traditional reading skills were lost. For the moment, old-school reading comprehension skills still prove to be very useful when searching and interpreting online information [18, 19].

Appendix

	Low chatting prevalence (< 35%) in 2009	High chatting prevalence (> 35%) in 2009
Africa
No countries included	—	—
Asia
6 countries	4 countries	2 countries
25.4% of all students	24.6% of all students	0.8% of all students
	Indonesia, Japan, South Korea, Thailand.	Hong Kong, Israel.
Australasia
2 countries	1 country	1 country
1.5% of all students	0.3% of all students	1.2% of all students
	New Zealand	Australia
Europe
24 countries	2 countries	22 countries
31.1% of all students	0.4% of all students	30.6% of all students
	Albania, Ireland	Austria, Belgium, Bulgaria, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Iceland, Italy, Latvia, Luxembourg, Norway, Poland, Portugal, Russian Federation, Spain, Switzerland, Sweden, United Kingdom
North America
3 countries	1 country	2 countries
25.0% of all students	6.2% of all students	18.8% of all students:
	Mexico	Canada, United States
South America
4 countries	2 countries	2 countries
17.0% of all students	13.3% of all students	3.7% of all students
	Brazil, Peru	Argentina, Chile
Total	10 countries	29 countries
	44.8% of all students	55.2% of all students

Table A1.

Countries with low vs. high chatting prevalence by continent.

	Reading literacy		Online chatting		Country Weight
	Change 2000–2009	Change 2009–2018	Percentage 2009	Change 009–2018
Asia
Hong Kong	28.3	−14.7	56.6	25.2	.132
Indonesia	41.2	−28.2	4.2	62.7	4.956
Israel	36.5	5.4	42.8	38.1	.199
Japan	11.2	−6.8	9.0	76.5	2.302
South Korea	18.4	−36.8	29.3	53.4	1.100
Thailand	8.3	−29.7	14.8	58.0	1.230
Australasia
Australia	−15.1	−11.3	39.9	25.6	.475
New Zealand	−8.4	−13.6	25.4	49.0	.102
Europe
Albania	48.3	21.9	20.7	42.6	.066
Austria	−27.8	10.6	46.3	37.2	.158
Belgium	0.7	−11.8	45.8	37.8	.227
Bulgaria	5.4	1.5	73.1	−2.1	.122
Czech Republic	−9.5	10.3	65.0	13.9	.207
Denmark	−1.1	7.4	54.3	22.6	.112
Finland	−11.9	−17.3	48.5	26.5	.116
France	−4.4	1.8	46.4	25.6	1.414
Germany	17.8	1.7	55.7	35.1	1.565
Greece	8.1	−17.1	38.4	36.8	.202
Hungary	12.8	−19.3	57.5	8.2	.192
Iceland	−9.3	−16.7	65.0	14.6	.008
Ireland	−19.8	17.8	27.4	57.8	.109
Italy	2.3	−1.1	52.7	35.1	.991
Latvia	22.1	−8.0	48.6	22.0	.046
Luxembourg	32.5	−11.3	48.4	28.7	.010
Norway	−2.0	8.2	60.0	5.7	.109
Poland	25.8	1.4	62.3	21.2	.860
Portugal	16.7	9.4	49.6	20.7	.188
Russian Federation	6.1	34.7	42.0	15.9	3.003
Spain	−14.7	38.2	47.9	39.6	.753
Sweden	−26.6	8.6	52.7	14.5	.201
Switzerland	11.2	−18.8	44.9	40.4	.157
United Kingdom	−28.2	12.7	51.7	26.2	1.302
North America
Canada	−8.6	−2.0	47.9	16.7	.676
Mexico	8.6	3.3	22.7	47.1	2.432
United States	−4.9	6.2	36.5	25.0	6.645
South America
Argentina	−27.0	15.1	39.3	40.4	1.022
Brazil	10.0	5.7	34.9	39.0	4.404
Chile	47.2	3.6	38.5	38.3	.437
Peru	57.9	31.8	17.6	30.0	.772

Table A2.

Key data per country.

Societal developments	Educational changes	Changes in sample characteristics	Changes in reading behavior & attitudes
Percentage immigrants	Student-teacher ratio	Percentage female students	Reading enjoyment
Parents’ education	Grade level of 15-year-olds	Coverage of all 15-year-olds	Frequency reading magazines
Parents’ occupation		Student age	Frequency reading comic books
Internet availability			Frequency reading non-fiction books
			Frequency reading newspapers
			Time reading for enjoyment

Table A3.

Control variables in stepwise regression analyses.