Gender Disparities in AP Computer Science Exams: Analysis of Trends in Participation and top Achievement

Gender Gaps in CS Education

Research investigating the nature of gender disparities in the CS workforce identified comparable issues in the CS-related educational settings and attainments. For example, women represented only one in five CS graduates in 2019 (NCWIT, 2021). Similarly, as of 2016, fewer than 20% of CS doctorate degrees at PhD granting institutions were earned by women (Zweben & Bizot, 2018). National reports have suggested that the underrepresentation of women in computing might be associated with barriers and conditions that emerge during pre-college educational settings (Master et al., 2021; NAP, 2021). In line with these reports, a large body of research indicates that K–12 experiences contribute significantly to later tenacious disparities, given that these experiences determine most students’ career choices (Busch, 1995; Haller & Beyer, 2006; Joseph & Thomas, 2020; Madkins et al., 2019; Sonnert, 2009; Wang et al., 2015; Wilson, 2002).

Several studies associated female underrepresentation in CS with parenting-related factors (Bhanot & Jovanovic, 2009; Ing, 2014; Miller & Kimmel, 2012; Sax et al., 2020; Sonnert, 2009; Wang & Hejazi Moghadam, 2017; Wang et al., 2015). Traditionally, parents have perceived their daughters’ and sons’ abilities unequally and they also have had biases about guiding their kids to STEM careers (Sonnert, 2009). For example, parents usually undervalue their daughters’ abilities in STEM fields but they overvalue their sons’ abilities (Bhanot & Jovanovic, 2009). Moreover, girls had lower parental encouragement and support to pursue a career in CS, which resulted in lower interest in computing (Wang & Hejazi Moghadam, 2017). When supported by their parents, girls cultivated comparable interest and confidence in computing-related activities (Ing, 2014; Miller & Kimmel, 2012).

In a related vein, teachers and school settings have played a crucial role in helping girls participate and succeed in computing programs too. In particular, teachers have been very influential for students in choosing a career or major. However, teachers are less likely to encourage girls toward CS careers when compared to boys, which is another factor contributing to the underrepresentation of women in computing (Google & Gallup, 2016). Further, when girls were exposed to female role models such as teachers, they were more likely to visualize themselves in computing careers (Haller & Beyer, 2006). Moreover, girls, whose high school CS teachers were women, were significantly more likely to enroll in computing courses compared to their female peers whose CS teachers were male (Haller & Beyer, 2006).

In addition to the crucial role that teachers play, developing a sense of belonging in the classroom contributes to girls’ self-efficacy beliefs (Master et al., 2016; Medel & Pournaghshband, 2017; Tillberg & Cohoon, 2005). Girls were found to be unwilling to enroll in computing courses when they noticed that activities, projects, and lessons were prepared according to boys’ interests. According to Master et al. (2021), gender-related stereotypes about computing may contribute to gender imbalance too. For example, even having students presume “women are less interested in computing than men” may contribute to the gender disparity in computing jobs (Master et al., 2021). Likewise, when girls don’t see other girls in the CS classroom, they lack belonging and interest in computing (Wang et al., 2015). Relatedly, Webb and Miller (2015) and Cheryan et al. (2013) underscored that girls showed greater interest in computing when they felt belonging in computing.

Prior experiences in CS classes affect girls’ interest in computing too (Cohoon & Cohoon, 2006). A growing body of research indicated that participation in high school CS courses significantly predicted girls’ choices in majoring in CS-related fields (Babes-Vroman et al., 2017; Brown & Brown, 2020; Wang et al., 2015). For instance, analyzing 67,000 college students’ responses, Babes-Vroman et al. (2017) found that participation in AP CS programs during high school correlated positively with the decision to enroll in CS bachelor's degree programs. In another study, Brown and Brown (2020) conducted a 4-year longitudinal study and reported that enrollment in the AP CS programs predicted students’ college enrollment. According to the researchers, students were more likely to attend higher education if they participated in AP CS programs (Brown & Brown, 2020). Similarly, |Wang et al. (2015) indicated that taking an AP CS course in high school significantly affected girls’ decisions to study computing in college. More particularly, enrollment in AP CS A course during high school influenced nearly 4 out of 10 girls’ decisions to study computing-related majors (Wang et al., 2015). Concluding altogether, these findings draw attention to the importance of prior experience, in particular to the participation in AP CS courses, for the retention of girls in the computing disciplines (Babes-Vroman et al., 2017).

Despite the persistent gender disparities found in the computing field, researchers have consistently indicated that gender has no bearing on CS ability (e.g., Beyer, 2014; Boda & McGee, 2021). For example, Boda and McGee (2021) investigated gender disparities in AP CS A exam achievement. Using the AP exam scores of over 500 students from Chicago Public Schools, the authors examined the participation and achievement differences between girls and boys from 2016 to 2019. According to their findings, girls were equally as likely as boys to pass the AP CS A exam. They also underscored that student participation in a course taught by a CS instructor with a prior AP CS A teaching experience further predicted success in an AP CS A program (Boda & McGee, 2021).

Lack of access to computing education and resources might be another factor causing later disparities in the labor market. This is particularly true in environments where girls may not have the same opportunities to engage with computing as their male peers. According to the Capacity for, Access to, Participation in, and Experience of equitable CS education (CAPE) framework (Fletcher & Warner, 2021), equity in computing education is an outcome of access to computing resources and education. The developers of the framework, Fletcher and Warner (2021), emphasized that “if students are to choose to participate in CS, they must first have equitable access to CS courses and programs” (p. 25). In addition, they added that the underrepresentation of women in computing is not a result of a mere factor but inequities in the entire ecosystem of CS education. The authors indicated that without placing a systematic perspective on disparities, understanding why girls avoid computing would not be possible.

AP Program

The AP program was founded by College Board (formerly known as College Entrance Examination Board), a not-for-profit educational organization, in 1955 to provide high school students with college-level courses. The AP program initially started with 12 AP courses and served approximately 1,000 students from 110 high schools during its first year. The number of students registered for the AP courses and the number of courses offered grew gradually over the years. In 2020, the number of students who took at least one AP exam increased to 2.64 million, and the number of high schools that offered AP programs reached 20,000 in the United States alone. The majority of U.S. high schools offer at least one AP course for their students; over 4.7 million AP exams were given in 2020 (College Board, 2021c).

Currently, the program offers 38 college-level courses in diverse subjects including math, science, CS, history, social sciences, art, English, and foreign languages. In most cases, the length of an AP course lasts for approximately two semesters; it starts in the fall and concludes with an official exam at the end of the spring semester. The exam scores range from one to five, where a score of “five” represents the highest possible score that is often equivalent to letter grade “A” at the college level. Some higher education institutions individually set minimum scores that could enable students to get college credit.

AP CS Program

The Advanced Placement Computer Science (AP CS) is a subset of the AP program. During the past several decades, the program has offered three CS courses and examinations, namely AP CS A, AP CS AB, and AP CS Principles. AP CS A and AP CS Principles programs are still in operation, but the AP CS AB program was discontinued in 2009. Detailed information about the AP CS A, AP CS Principles, and AP CS AB programs is presented below.

Present Study

One program that has been regarded as the industry standard for pre-college CS education is the AP Program (College Board, 2021b). It is widely acknowledged that AP courses play a critical role in majoring in computing-intensive fields and broadening the participation of women in the computing workforce (Brown & Brown, 2020; Howard & Havard, 2019; Sax et al., 2022; Sax et al., 2020; Wang et al., 2015). However, to the best knowledge of the authors, no current studies investigated the direction and magnitude of gender disparities in AP CS exams. Given that the AP CS program is a major pipeline for the computing workforce, such an investigation also helps to understand current and future directions of gender disparities in K–12 CS education and beyond (Sax et al., 2022).

Within this study, the authors examined publicly available data that belonged to three AP programs, AP CS A, AP CS AB, and AP CS Principles. Readers might question why data from a discontinued program, AP CS AB, and a very fresh program, AP CS Principles, were included in the study. We were aware that including data from these exams would not provide us with enough data points to investigate trends for each of these exams individually. However, including these data in aggregate AP CS data would enrich our findings and discussions and would also help us to document the differences among exams. In addition, including data from all exams would provide us with a more diverse student profile who ended up taking one of these exams.

The primary goal of this study was to investigate the trends in gender disparities in participation and top achievement among AP CS examinees. Researchers in other STEM areas, in particular mathematics, have suggested that differences in gender representation in the labor force were associated with gender disparities in high levels of achievement in early educational attainments (Lazear & Rosen, 1990; Makel et al., 2016; Wai et al., 2010). In line with this perspective, we analyzed the gender disparities in participation and top achievement separately. Our methodologies for identifying gender disparities in participation and top achievement among AP CS students are presented. The following research questions guided the present study:

What are the trends in gender disparity in participation in AP CS exams?

Research Design and Data Source

In this study, the authors used trend analysis as a research design methodology to answer the research questions. Trend analysis provides researchers with a robust method to understand how things have changed over time. Despite its unique strength to identify patterns existing within the datasets, it is rarely used in educational research (Rae, 2014). Through trend analysis, the authors analyzed the data obtained from the College Board's annual AP National Summary reports. The data drew from test scores of over 1,000,000 students who participated in one of the three AP CS exams (CS A, CS AB, and CS Principles) from 1997 to 2020. The AP National Summary reports included detailed information about exam participation and performance separated into several categories, including exam type, grade level, and gender (College Board, 2020). The number of participants and male-to-female ratios (MFRs) of participants in each of these three exams is shown in Table 1. Likewise, Table 2 shows the gender breakdown and MFR of top achievers who scored five on the exams over the years.

OPEN IN VIEWER

Table 1.

Number of Participants and Male-to-Female Ratios (MFRs) in Advanced Placement (AP) Computer Science (CS) Exams.

Year	AP CS A			AP CS AB				AP CS Principles				All AP CS		All AP (CS and non-CS)
	M	F	MFR	M	F	MFR	M	F	MFR	M	F	MFR	M	F	MFR
1997	5584	1408	3.97	3841	526	7.3				9425	1934	4.87	419328	480135	0.87
1998	5000	1144	4.37	3430	485	7.07				8430	1629	5.17	459458	532494	0.86
1999	9834	1959	5.02	5839	611	9.56				15673	2570	6.10	520845	601569	0.87
2000	10979	2180	5.04	5951	719	8.28				16930	2899	5.84	574905	667419	0.86
2001	12637	2618	4.83	6589	814	8.09				19226	3432	5.60	634469	745677	0.85
2002	12651	2492	5.08	6811	758	8.99				19462	3250	5.99	706509	842490	0.84
2003	11903	2224	5.35	6164	704	8.76				18067	2928	6.17	776053	929154	0.84
2004	11620	2252	5.16	5291	628	8.43				16911	2880	5.87	838849	1013851	0.83
2005	11233	2218	5.06	4454	500	8.91				15687	2718	5.77	927049	1137996	0.81
2006	11613	2495	4.65	4238	499	8.49				15851	2994	5.29	1018477	1247561	0.82
2007	11864	2665	4.45	4272	591	7.23				16136	3256	4.96	1110599	1366363	0.81
2008	12225	2789	4.38	4200	615	6.83				16425	3404	4.83	1199064	1475232	0.81
2009	12965	3096	4.19	4268	632	6.75				17233	3728	4.62	1288661	1572251	0.82
2010	15664	3726	4.20							15664	3726	4.20	1417052	1717634	0.83
2011	17139	4000	4.28							17139	4000	4.28	1525335	1840282	0.83
2012	20147	4635	4.35							20147	4635	4.35	1630791	1963905	0.83
2013	24070	5485	4.39							24070	5485	4.39	1751004	2073687	0.84
2014	29869	7458	4.00							29869	7458	4.00	1844208	2212075	0.83
2015	36202	10142	3.57							36202	10142	3.57	1971809	2371738	0.83
2016	41737	12642	3.30							41737	12642	3.30	2048230	2511043	0.82
2017	42921	13167	3.26				30608	13172	2.32	73529	26339	2.79	2155059	2648363	0.81
2018	45651	14389	3.17				48544	22320	2.17	94195	36709	2.57	2206930	2716142	0.81
2019	48498	15699	3.09				63465	30895	2.05	111963	46594	2.40	2204426	2725721	0.81
2020	48698	16246	3.00				75293	38672	1.95	123991	54918	2.26	2038212	2548888	0.80

Note. F = Female; M = Male.

OPEN IN VIEWER

Table 2.

Number of top Scorers and Male-to-Female Ratios (MFRs) in Advanced Placement (AP) Exams.

Year	AP CS A			AP CS AB			AP CS Principles			All AP CS			All AP (CS and non-CS)
	M	F	MFR	M	F	MFR	M	F	MFR	M	F	MFR	M	F	MFR
1997	624	102	6.12	1216	128	9.50				1840	230	8.00	68033	61949	1.10
1998	477	65	7.34	1121	128	8.76				1598	193	8.28	76569	65228	1.17
1999	1667	203	8.21	1890	131	14.4				3557	334	10.7	88569	76449	1.16
2000	1654	195	8.48	2070	213	9.72				3724	408	9.13	96997	85332	1.14
2001	2359	293	8.05	2287	243	9.41				4646	536	8.67	102243	86671	1.18
2002	2564	347	7.39	2387	195	12.2				4951	542	9.13	118224	103938	1.14
2003	2069	324	6.39	2334	239	9.77				4403	563	7.82	127305	110219	1.16
2004	2264	298	7.60	1462	135	10.8				3726	433	8.61	140588	126570	1.11
2005	2064	312	6.62	1392	162	8.59				3456	474	7.29	145910	128931	1.13
2006	2610	450	5.80	1466	142	10.3				4076	592	6.89	164690	142835	1.15
2007	2386	396	6.03	1433	179	8.01				3819	575	6.64	179800	158402	1.14
2008	2749	462	5.95	1628	243	6.70				4377	705	6.21	198185	172753	1.15
2009	3114	576	5.41	1690	237	7.13				4804	813	5.91	227924	200717	1.14
2010	4248	822	5.17							4248	822	5.17	244395	215701	1.13
2011	4458	771	5.78							4458	771	5.78	256464	228845	1.12
2012	4897	916	5.35							4897	916	5.35	287855	253635	1.13
2013	6511	1252	5.20							6511	1252	5.20	284028	245912	1.15
2014	6463	1370	4.72							6463	1370	4.72	290843	260201	1.12
2015	8970	2109	4.25							8970	2109	4.25	287479	270229	1.06
2016	8880	2252	3.94							8880	2252	3.94	317355	303990	1.04
2017	10411	2809	3.71				4370	1615	2.71	14781	4424	3.34	320402	294390	1.09
2018	11534	3225	3.58				6885	2921	2.36	18419	6146	3.00	346863	325288	1.07
2019	13163	3799	3.46				8844	3930	2.25	22007	7729	2.85	347362	332732	1.04
2020	12548	3848	3.26				7083	5323	1.33	19631	9171	2.14	333600	373209	0.89

Note. CS = Computer Science; F = Female; M = Male.

Procedures for AP CS Exams

Data Analysis

Disparity index calculations

Researchers have used different types of disparity indexes to explore disparities in educational settings (Bahar, 2021a; Bahar, 2021b; Ellison & Swanson, 2010; Hyde et al., 1990; Hyde et al., 2008; Lindberg et al., 2010; Makel et al., 2016; Olszewski-Kubilius & Lee, 2011; Robinson & Lubienski, 2011; Wai et al., 2018). In this study, similar to some prior studies (Bahar, 2021a; Makel et al., 2016; Wai et al., 2018), we compared the number of males to females in AP exam participation and top achievement and we reported an MFR as a measure of gender disparity index. The MFR index can be presented as

M F R = # o f m a l e s # o f f e m a l e s

The MFR index value briefly indicates a parity or disparity between the number of male and female students. An absolute parity occurs when the value of the MFR is equal to 1, which means the numbers of male and female students are equal to each other (Bahar, 2021a). When the number of males exceeds that of females, the MFR value becomes larger than 1, which is an indicator of a disparity favoring male students. Similarly, when females outnumber males, the MFR value is expected to be smaller than 1, which means the disparity favors female students (Bahar & Maker, 2020). The operational definitions of gender disparity in participation and top achievement are presented below:

Index of Gender Disparity in Participation: In this study, we defined the gender disparity in participation as a measure of AP exam participation. We reported the index using the MFR among the number of students who participated in the exam.

Index of Gender Disparity in Top Achievement: Researchers who investigated the disparities in top achievement generally looked into the performances of top achieving students (Bahar, 2021a; Benbow & Stanley, 1983; Makel et al., 2016; Wai et al., 2010; 2018). Given that AP exams are criterion-referenced and student exam scores range from 1 to 5 with 5 being the highest possible score, we identified a score of 5 as top achievement. This is in line with the AP Score Scale table published by the College Board, which recommended a score of 5 as “extremely well qualified” with a college course grade equivalent of A + or A (College Board, 2021a). According to the College Board reports, roughly 20% of the participants received a score of 5in the past AP CS exams. In this study, similar to the participation index defined earlier, we also reported the top achievement index in terms of the MFR among the students who scored five on any of these AP CS exams.

Research Question 1

The findings indicated that the MFRs among students who participated in any type of AP CS exam declined from 4.87 to 2.26 between 1997 and 2020 (Table 1). According to the MK test, the MFR trend was downward, statistically significant (p < .001), and favored females (Table 3). The Sen's slope estimator procedures showed that the magnitude of the change in the trend line (slope) was −0.179 per year (Figure 1), which means the decline in the MFR was roughly one point in every 5–6 years. This shows that if the current trend remains steady, the number of males and females who participate in all types of AP CS exams will reach parity in less than two decades.

OPEN IN VIEWER

Figure 1.

Trends in male-to-female ratios (MFRs) among all participants and Sen's slopes.

OPEN IN VIEWER

Table 3.

Mann-Kendall (MK) two-Tailed Trend Test Summary Statistics.

Exam	Variable	Obs	Mina	Maxa	M	SD	Kendall's Tau	S	Var(S)	p-value	Slope Value	Intercept Value
AP CS A	Participation	24	2.998	5.352	4.257	0.715	−0.609	−168.000	1625.333	<.001	−0.096	197.935
	Top Achievement	24	3.261	8.482	5.741	1.570	−0.833	−230.000	1625.333	<.001	−0.223	452.951
AP CS AB	Participation	13	6.753	9.556	8.052	0.919	−0.231	−18.000	268.667	.306	−0.116	239.480
	Top Achievement	13	6.700	14.427	9.646	2.065	−0.359	−28.000	268.667	.100	−0.270	549.567
AP CS Principles	Participation	4	1.947	2.324	2.125	0.162	−1.000	−6.000	8.667	.083	−0.123	250.681
	Top Achievement	4	1.331	2.706	2.161	0.587	−1.000	−6.000	8.667	.083	−0.404	816.821
All AP CS	Participation	24	2.258	6.170	4.550	1.222	−0.761	−210.000	1625.333	<.001	−0.179	363.616
	Top Achievement	24	2.141	10.650	6.207	2.295	−0.877	−242.000	1625.333	<.001	−0.325	659.291
All AP (CS and non-CS)	Participation	24	0.800	0.873	0.831	0.020	−0.565	−156.000	1625.333	<.001	−0.002	5.497
	Top Achievement	24	0.894	1.180	1.113	0.060	−0.551	−152.000	1625.333	<.001	−0.005	10.302

Note. AP = Advanced Placement; CS = Computer Science; LB = lower bound; M = Mean; Obs = observations; SD = standard deviation; UB = upper bound.

^a

Min and Max values refer to MFR values.

The results of the MK test for the AP CS A exam were similar to that of all-AP CS exams. As the MFR values declined from 3.97 to 2.97 (Table 1), a significant decreasing trend existed (p < .001) between 1997 and 2020 (Table 3). The Sen's slope estimator procedures showed that the magnitude of the change in the trend line was −0.097 per year (Figure 1), which means the decline in MFR was roughly one point in every 10 years. This shows that if the current trend stays steady, the number of males and females who participate in the AP CS A exam will reach parity in less than two decades.

For CS AB and CS Principles exams, the MK test could not detect significant trends, which might be a result of a small data set (Figure 1). However, similar to CS A, the Sen's slope estimators presented a decline in the MFR slope for CS AB (slope value: −0.116) and CS Principles exams (slope value: −0.124).

Research Question 2

The trend in gender disparity in top achievement in AP CS exams showed similar behaviors to the trend in participation. The findings indicated that the MFRs among students who scored five (top score) on any type of AP CS exam declined rapidly from 8.00 to 2.14 between 1997 and 2020 (Table 2). According to the MK test, the MFR trend was downward, statistically significant (p < .001), and favored females (Table 3). The Sen's slope estimator procedures showed that the magnitude of the change in the trend line was −0.325 per year (Figure 2), which means the decline in MFR was roughly one point in every 3–4 years. This shows that if the current trend stays steady, the number of males and females who scored five in all types of AP CS exams will reach parity in several years.

OPEN IN VIEWER

Figure 2.

Trends in male-to-female ratios (MFRs) among top scorers (score of 5) and Sen's slopes.

The results of the MK test for the AP CS A exam were similar to that of all AP CS exams. As the MFR values declined from 6.12 to 3.26 (Table 1), a significant decreasing trend existed (p < .001) between 1997 and 2020 (Table 3). The Sen's slope estimator procedures showed that the magnitude of the change in the trend line was −0.223 per year (Figure 2), which means the decline in MFR was roughly one point in every 4–5 years. This shows that if the current trend stays steady, the number of males and females who scored five in the AP CS A exam will reach parity by around a decade.

Similar to the MK test results in participation, the MK test could not detect significant trends for top achievement in CS AB and CS Principles exams (Figure 2). However, similar to CS A, the Sen's slope estimators presented sharp declines in the MFR slope for both CS AB (slope value: −0.270) and CS Principles exams (slope value: −0.404).

Trends in Gender Disparity in Participation for AP CS Exams

There is an increase in the number of programs that attempt to broaden the participation of girls in CS education in order to expand their access, interest, and attitude toward computing (Cummings et al., 2021; Ericson et al., 2016; McAlear et al., 2019; Michell et al., 2017; Scott et al., 2017). In spite of these efforts, nationwide reports emphasize that gender disparity continues to exist (Google & Gallup, 2015; 2016). Generally, our findings are consistent with prior studies, which have identified gender disparities, in favor of males, in participation in CS programs (Boda & McGee, 2021; Ericson & McKlin, 2018). As shown in Table 1, the number of AP CS exam takers was in favor of males for all three exams. According to the data we evaluated, the MFRs among any kind of AP CS exam takers in 2020 was roughly 2.26, whereas the MFR in the same year for all AP exams was around 0.80. Overall, comparing the MFR values among AP exam participants, the data indicated that the percentage of female students in a non-CS AP exam room was three times more than those of females in an AP CS exam room. Our findings confirm that the gender disparities in AP CS programs are similar to those in the computing workforce.

Although these numbers presented clear evidence for female underrepresentation in AP CS programs, the trend analysis provided hope for closing the gender disparities in the future. From 1997 to 2020, the MFR value declined from 4.87 to 2.26 for all AP CS exam takers. The trend analysis indicated a significant decreasing trend in MFR values across years, which has been in favor of females. Further, according to Sen's slope estimator procedures, if the current trend stays steady, the number of males and females who participate in all types of AP CS exams will reach parity in less than two decades. All of these promising findings show that the percentage of female students participating in AP CS classrooms is increasing steadily.

According to the data, while in operation between 1997 and 2009, the AP CS AB program could not succeed to gain sufficient interest from students; more specifically, there was a lack of interest from female students. During these 13 years, the average number of female participants in the AP CS AB exam was barely over 500 for any year. Meanwhile, the MFR value for the same exam was roughly 8.00, which was twice more than that for the AP CS A exam during the same period. After the AP CS AB exam was discontinued following the 2009 exam, the AP CS A served as the only CS course in the AP program until 2017. The trend analysis indicated that MFR values for AP CS A exam have significantly decreased, in favor of females, between 1997 and 2020. Further, the findings showed that, even without the addition of the AP CS Principles course in 2017, the MFR among participants in AP CS A exams would reach parity in the next two decades. However, it is evident that reaching parity in MFR values is shortened by half with the addition of the AP CS Principles course.

Investigating the factors that affected the decline in MFR values and significant decreasing trends in gender disparities was beyond the goal of this study. However, we argue that the addition of the AP CS Principles course to the AP CS program was one of the factors that accelerated the decline of the MFR in the overall AP CS program. Given that no trend in MFR values existed among participants in AP CS Principles, readers can question how the addition of the AP CS Principles accelerated the decline of MFR in the overall AP CS program. One of the facts behind our argument was that, as of 2020, the MFR among participants in AP CS A exam was 2.97, whereas it was only 1.94 for the AP CS Principles exam. Moreover, the number of girls who participated in the AP CS Principles exam jumped from 13,000 to 39,000 in just 3 years. Considering the number of female students in the entire AP CS program never reached 13,000 in any year before 2016 even when AP CS A and AB courses were combined, it is reasonable to assume that the AP CS Principles course played a role in the decline of the MFR value among AP CS participants.

Although it is still far from having parity in MFR values among participants, the AP CS Principles course has been the most successful AP CS program to reach females. One can question what makes AP CS Principles more successful than AP CS A and CS AB courses aimed at reaching females. We believe that this is a very valid and important question to unearth why females are underrepresented in AP CS programs and even further in the CS labor force. At this point, we pivot our point of view on the matter and suggest looking into why females did not favor AP CS A or AB programs compared to the AP CS Principles. Given that average MFR values in AP CS A program has been almost double of that in AP CS Principles, understanding the factors that contribute to such high MFR values in a different CS program might provide researchers with significant clues to answer the question. However, this is an essential question left for future researchers.

Trends in Gender Disparity in top Achievement for AP CS Exams

One purpose of this study was to explore the trends in gender disparity for top achieving students who scored “five” at the AP CS exams. As seen in Table 2, the number of male students who received “five” at the AP CS programs was higher than the girls for all AP CS exams. According to the findings, the average MFR among students who scored 5 from any kind of AP CS exam between 1997 and 2020 was roughly 6.2, whereas the MFR value for the same period for all other AP exams was around 1.11. Similar to the findings associated with AP CS participation, the data indicated that the percentage of top-scoring female students (scoring five out of five) in a non-CS AP exam room was almost six times more than those of females in an AP CS exam room. Overall, these findings were consistent with earlier research, which reported that AP CS programs mostly favored male students in regards to top achievement (Boda & McGee, 2021; Brown & Brown, 2020; Ericson & Guzdial, 2014). However, we found that, from 1997 to 2020, the MFR value among top scorers declined from 8 to 2.14 for all AP CS exam takers. Regarding the decline in MFR values, the trend analysis indicated significant downward values across years, which has been in favor of females. Comparing the numbers over years, the decline in MFR values has been a significant accomplishment for gender equity.

Although overall trends for top achievement and participation were similar, the MFR values among top scorers were higher than those among all participants for overall AP CS data. For example, the average MFR among students who scored five on any kind of AP CS exam between 1997 and 2020 was roughly 6.2, whereas the MFR value among all AP CS participants was around 4.55. This is an interesting finding because factors associated with the underrepresentation of females in AP CS participation were not reflected at the same rate in top achievement. Several prior studies pointed out similar problems in mathematics and spatial thinking domains as the gender gaps were found to be widening significantly at the right tail of the achievement curve (Bahar, 2021a; Benbow, 1988; Gallagher et al., 2000; Wai et al., 2009; Webb et al., 2007). Among these researchers, some of them associated higher gender disparities among top achievers with differences in cognitive abilities between males and females in these domains (Benbow, 1988; Gallagher et al., 2000; Wai et al., 2009; Webb et al., 2007). Although this claim may seem to offer a plausible explanation regarding higher gender disparities among students scoring five in the AP CS exams, we think the consistent decline in the disparities over the years suggests a different explanation. Supporting our perspective, studies in CS education revealed that females could achieve as high as males when interventions were specifically designed for their needs and interests. For example, Boda and McGee (2021) also acknowledged that girls who participated in the Chicago Alliance for Equity in Computer Science (CAFÉCS) program performed as well as males in the AP CS A programs in Chicago. CAFÉCS is a Researcher-Practitioner Partnership that places the accent on equity in CS education in Chicago Public Schools. The goal of the organization is to ensure that all students regardless of gender, ethnicity, and ability have access to engaging and rigorous CS courses. CAFÉCS offers teacher professional development and assists students for courses like AP CS. In another study, Ericson et al. (2016) demonstrated that girls who attended the Sisters Rise Up 4 CS program performed better in the AP CS A exam than boys. Ericson and her colleagues founded Sisters Rise Up 4 CS in Georgia and the program has been a part of the Expanding Computing Education Pathways program, which provides assistance, tutoring sessions, role models, and engaging technology experiences for girls who are taking AP CS exams. Ultimately, the program aimed to increase the number of girls passing AP CS exams, and therefore increase the number of women pursuing computing fields. Looking into the commonalities in these programs, the availability of mentors who were near-peer role models along with increased attention to equity, diversity, and inclusion made a difference (Cummings et al., 2021; Ericson & McKlin, 2012; 2018; Ericson et al., 2016). Additionally, introduction to computing through culturally and socially relevant programs, implementing inclusive pedagogy, and incorporating socially meaningful activities into the courses might have led girls to higher achievement in AP CS programs (Boda & McGee, 2021; Christensen et al., 2021; Ericson & McKlin, 2018; Goode & Chapman, 2011; Marcher et al., 2021; Margolis et al., 2008).

Moreover, if the high MFR values among top achieving students were related to cognitive abilities, it would make sense to expect similar findings across different countries and cultures. However, females have outperformed males in CS programs in England (Kemp et al., 2019). Similarly, in Malaysia, females dominated the computing workforce (Mellström, 2009). Melstrom has associated these findings with the cultural and social settings instead of cognitive abilities. Similar to what previous research suggested, social barriers might be playing an important role with regard to top achievement. In addition, stereotypes that depict computing as a masculine field, and a lack of belongingness, encouragement by parents and teachers, and role models in media and classroom in addition to an unsophisticated understanding of computing careers, and naive perceptions about computing might cause females to think that they do not have sufficient abilities to earn top scores in CS (Master et al., 2021; Wang & Hejazi Moghadam, 2017; Wang et al., 2015). In turn, this might lead young women to fail to perform at their highest level and may cause them to achieve lower scores than males. However, revealing possible factors for these disparities in top achievement are beyond the scope of this study; therefore, this is an important endeavor for future researchers.

Implications for Practice

Our findings have important implications for educators to ameliorate the gender differences in K–12 CS programs. First, our trend analysis indicated that large gender gaps, in favor of boys, emerged in CS education; however, the disparities have been declining significantly over the years. In particular, the addition of the AP CS Principles course has accelerated the period to reach parity in participation and top achievement. This implies that a people-oriented, socially meaningful, culturally relevant, and responsive computing curricula that cover a broader range of computing topics that are designed to encourage females to participate in CS education would definitely make a difference (Christensen et al., 2021; Marcher et al., 2021). On a parallel track, many experts in the field of CS education advocated for altering the programming language in AP CS A from Java to Python and incorporating topics more appealing to girls, such as data science, as Barbara Ericson and Barbara Liskow discussed during their keynote talk at the SIGCSE 2022 event. In regards to this fact, prior research and anecdotal evidences suggested that practices and strategies, which were designed to promote belonging and interest such as creative and expressive computing, have been successful in expanding girls’ participation and top achievement in CS courses (Barker et al., 2002; Guzdial, 2021; Master et al., 2016). At this point, we stress that educators, in particular CS teachers and school counselors, play a crucial role in encouraging the participation of females in CS programs. Teachers and counselors can contribute to the recruitment of a higher number of females to CS courses through initiatives that are designed to foster female students’ self-efficacy such as formal and informal school or after-school activities. Furthermore, like prior studies suggested (Ericson & McKlin, 2012; 2018; Ericson et al., 2016), increasing the appearance of female role models at the school by inviting successful female professionals as guest speakers and providing networking opportunities with peers who major in CS programs would allow females to observe others and increase their self-efficacy toward CS occupations too.

Given that early exposure to introductory and inclusive computing programs in elementary, middle, or early high school might motivate students to take AP CS courses (Tsan et al., 2016), school curricula should be enriched with a higher number of pre-AP CS courses. Similarly, research suggested that students who participated in the AP CS Principles program are more likely to enroll in the AP CS A program. For example, Brown and Brown (2020) found that females who participated in the AP CS Principles program were more motivated to take the AP CS A exam and likely to go to college when compared to students who didn’t participate in the AP CS Principles programs. Similarly, Sax et al. (2022) found positive associations between participation in AP programs and majoring in computing fields for girls. In another study, Boda and McGee (2021) found that students who enrolled in early CS programs increased girls’ participation in AP CS programs. Combining them with the findings of our study, we suggest educators reach and recruit female students earlier through these introductory courses to increase their perseverance. Making students feel a sense of belonging to the CS classroom through engaging them with relevant projects, ensuring that they are safe in CS classrooms, and helping them understand that computing is a highly creative, collaborative, and socially meaningful and people-oriented occupation might influence girls’ participation and achievement in AP CS programs.

Lastly, in line with prior studies (Google & Gallup, 2016; Haller & Beyer, 2006), we believe teachers play a crucial role in increasing girls’ participation in CS education and workforce. If we would like to increase females’ participation and achievement in CS programs, we will need to equip our teachers with equitable and inclusive pedagogical knowledge as well as subject knowledge. However, most of the AP training initiatives, including College Board's AP Summer Institutes for teachers, are designed to increase teachers’ content knowledge. We recommend that professional development programs for in-service CS teachers should be designed in a way that considers equity as a priority. Relatedly, pre-service CS teacher education programs should draw attention to equity problems in computing so that teacher candidates graduate with awareness and competency to address the issue proactively and offer more inclusive CS courses and programs (Brown & Brown, 2020; Gray et al., 2020; Master et al., 2021). Finally, as Fletcher and Warner (2021) suggested, we recommend that teachers, administrators, and policymakers are encouraged to utilize the CAPE framework to identify antecedents to systemic gender inequities (Fletcher & Warner, 2021). Having such a paradigm will support educators and policy makers to approach the disparity problem from a broader exosystemic perspective, and we believe this will ease the process of designing efficient interventions for K–12 settings.

Implications for Future Research

According to the authors’ best of knowledge, no prior studies to date have investigated the association between disparity trends in participation and top achievement for AP CS exams. Thus, there is no similar study to compare and contrast the findings of our study. Nonetheless, it might be inferred from these findings that increasing females’ participation in AP CS programs will also positively influence the percentage of top-scoring females at AP CS programs. This is encouraging, particularly for the increased representation of women in the computing workforce. Given that prior studies have associated the disparities in the labor force with gaps in high levels of achievement in early educational attainments (Lazear & Rosen, 1990; Makel et al., 2016; Wai et al., 2010), declining MFR values in participation will not only positively reflect in top achievement in AP programs but also reflect in later job opportunities for females. However, we also assert that reaching parity in MFR participation values might not be enough to reach parity among top achievers. Examining the factors that contribute to lower MFR values in top achievement is still an important question for future researchers. It would also be appropriate to utilize qualitative methods to answer some of the follow-up questions that we suggested in this study.

Limitations

The empirical findings reported herein should be considered in the light of some limitations. First, we investigated the gender disparities among students who took AP CS exams, rather than those for students who enrolled in AP CS programs. Students are not mandated to take the corresponding AP exam after completing the course, and the datasets we obtained from College Board only represent the data belonging to examinees rather than students who enrolled in the program. For example, as of 2018–2019 school year, out of 117,787 high school students who enrolled in any AP program in the State of Georgia (Georgia Department of Education, 2021), only 84,239 of them chose to take an AP exam. Although we do not have national data available to us, from the example of Georgia, we conclude that roughly 1 out of 4 students who enrolled in an AP course does not take the AP exam followed by the course. This means that our findings allowed us to make discussions only about AP exam takers rather than all students enrolled in AP CS courses.

The second limitation concerns the data that belonged to the 2020 examination year. Given that 2020 testing was compromised by COVID-19 shutdowns that began about a couple of months before the AP exam period in May, the exam scores from 2020 might have shown discrepancies from the regular trends. We could have taken out the 2020 exam data from the dataset but we did not for a couple of reasons. First, we did not have any evidence from the literature to claim whether pandemic conditions might have impacted male and female participation and achievement differently in 2020. Second, the impact of COVID-19 was not temporary, and it is still prevalent all over the world. Taking out the 2020 data from the analysis would look like skipping a reality and this would be misleading for readers so we chose to include it in the dataset. However, these results must be interpreted with caution by readers and future researchers.

Lastly, another limitation was about the potential threats with the longitudinal data that might have masked possible measurement variances over the dataset's 24-year period. Considering the rapid growth of AP CS program over the years, some readers can question whether AP CS exams can be considered as a stable indicator of gender gaps in computer education. Although we find this as a relevant concern, we do not have any information/evidence to reject this practice. However, we believe further qualitative investigations are needed to disclose these potential threats, if they exist, so that we recommend future researchers to investigate these points.