• Criteria for Useful Education Indicators
• The Condition of Education as a Resource
• The Need for Additional Indicators
• Questions of Cause and Effect
• Conclusion

The report is written in a style that should be accessible to a wide range of readers, from interested parents, to policy wonks, to education researchers. Plenty of tables without text are available in an appendix for those who love numbers; and for those who really love numbers, the standard errors are provided in another appendix as well. But the body of the report is in plain English with helpful charts and graphs.

Essentially, The Condition of Education is a book of indicators that can be used to monitor education in America over time. A comprehensive and consistent set of indicators must be based on a model of the education system that is sufficiently general, yet sufficiently complete that its validity will continue over time. At the crudest level, there need to be indicators of education inputs, processes, and outputs. Inputs include fiscal and other resources, teacher quality, student background, parent characteristics, and community norms. Processes include (a) organizational characteristics of schooling at the national, state, district, and school levels and (b) instructional characteristics of schooling (curriculum quality and teaching quality). Outputs include achievement, participation, attitudes, and aspirations. To the credit of NCES, the 2001 edition of The Condition of Education is comprehensive when judged against this model of inputs, processes, and outputs.

Education indicators, to be useful, must also meet several criteria:

First and most important, indicators must be easily understood by a broad audience. Indicators are not intended primarily to serve the needs and interests of researchers; rather, they are created to serve the information needs of policymakers and the public.

Second, even though the education system is extraordinarily complex—with various layers, levels, and participants—the number of education indicators must be relatively few. Organizing indicators by the main features of the model can help.

Third, indicators need to be stable over time, both in what is being monitored and how it is being measured. Without stability of definition and measurement, looking at trends over time is not possible; but trends are exactly what is of most importance. Is achievement getting better over time? Is the system growing in terms of numbers of students served? Are costs per pupil decreasing?

Fourth, indicators must be reported at a disaggregated level. In the United States, the main story about education is one of variability. For example, there is enormous variance between states in levels of student achievement. Some of our states achieve at levels comparable to the highest achieving countries in the world; other states achieve at levels comparable to the lowest achieving countries in the world. Variance is not only important for understanding student achievement; it is equally important for understanding inputs and school processes.

Fifth, education indicators must be accurate. There are two aspects of accuracy. One is an issue of measurement. For example, if an indicator is based on self-report questionnaire data, can self-reports be trusted to be valid? Sometimes the answer is yes and sometimes no. The second aspect of accuracy has to do with standard errors. When average student achievement is reported, for example, is it based on a sufficiently large sample that the number reported would have been likely to vary little, if at all, had a different sample been taken?

Sixth, indicators must be timely. Timeliness is always a problem for statistics. For example, the National Assessment of Educational Progress (NAEP) has assessed student achievement in mathematics no more often than every 2 years, and sometimes less frequently.

Timeliness leads to the seventh characteristic of indicators: they must be affordable. Obviously, there are tensions between the first six characteristics and affordability. At the current time, the United States invests only a tiny fraction of 1 percent of its total education budget on education indicators. Clearly, a larger investment is warranted.

When I review the 2001 edition of The Condition of Education, not only do I see that the volume and the indicators in it are driven by a comprehensive model of the education system; I also see that great attention has been paid to meeting the seven desirable characteristics of indicators.

One doesn’t sit down to read The Condition of Education from cover to cover. I use it as a resource to look up answers to questions I have about education. Sometimes these questions are stimulated by family conversations, other times by my teaching or research. Now that The Condition of Education is available on the NCES Web Site (http://nces.ed.gov), accessing the volume for answers to questions as they occur is easier than ever before.

In browsing through the 2001 edition, I couldn’t help being struck by a recurring theme. There are large and persistent inequities in our education system. These inequities are found in education inputs, processes, and outputs:

• White students continue to outperform Black and Hispanic students in reading, mathematics, and science. There was some small closing of these equity gaps in the 1970s, but the gaps persist and they are large.
• Among 25- to 29-year-olds, 6 percent of White students have failed to complete high school, compared with 13 percent of Black students and an alarming 37 percent of Hispanic students.
• A higher percentage of White high school graduates go immediately to college than do Black or Hispanic high school graduates. (Across all groups, 63 percent of high school graduates go immediately to college.)
• Among high school graduates, 36 percent of White students complete a bachelor’s degree, compared with 21 percent of Black students and 15 percent of Hispanic students.

The report is so full of useful information that it’s difficult to know where to stop. Here are a few additional statistics that caught my eye:

• Over 50 percent of parents report that they are “very satisfied” with their children’s schools. I guess they don’t blame the schools for the large inequities, since the results varied hardly at all by ethnicity.
• A surprisingly large 8 percent of students report being threatened or injured with a weapon during the previous 12 months. That’s nearly 1 out of 10; as a parent of five, I find that scary.
• In constant 2000 dollars, the median earnings of 25- to 34-year-olds (a) decreased over the past 20 years for males, with the percentage decrease being largest for the least educated, and (b) increased over the past 20 years for females, with the percentage increase being largest for the most educated. It’s hard to be happy about males losing ground, but it’s definitely encouraging to see females gaining ground.

Having established that the 2001 edition of The Condition of Education provides a great deal of timely and useful information about education in the United States, I must also point out that there are some key pieces of information missing. The reasons for these missing indicators undoubtedly relate to issues of cost and technical capability. Nevertheless, there are areas where NCES needs to move in the future.

Perhaps the most glaring omission from the volume is information about instructional practices. For example, in my work I have found that the content of instruction delivered by teachers is the single most powerful school-controlled predictor of differential gains in student achievement. In short, what students have had the opportunity to study determines in important ways what they learn. At one level, this sounds obvious; yet careful studies of algebra I in high school reveal enormous differences in the delivered curriculum, even across algebra I classes in the same school. Differences in the content of instruction across classes for first-year high school biology are even greater. Issues of equity could be better understood if indicators of instructional practice were included in the volume. For example, do Black students have access to the same content of instruction as White students?

Even though I identify the need for education indicators of instructional practice, and especially the content of instruction, I recognize that there are significant challenges to producing high-quality indicators. I have had good luck using teacher self-report on surveys, but there are limits to the validity of such data. Teachers may not see and understand their own practice in the same way as others would. Observation studies can solve some of this problem, but to understand the content for a school would require more observations than could be afforded. Work needs to be done to develop valid survey techniques to supply the needed indicators.

A second omission is in the area of teacher knowledge. There are tests of teacher knowledge for entering teachers, but data are not available on a national probability sample of teachers. Further, these entry-to-the-profession measures of teacher knowledge, while quite good, need further development to capture what has come to be known as teacher pedagogical content knowledge, the intersection between content knowledge and pedagogical practice informed by how students learn. At what level of knowledge are U.S. teachers operating in the various academic subjects and at the levels of elementary, middle, and high school? Do teachers in schools serving high concentrations of students from low-income families have the same content and pedagogical content knowledge as teachers serving more affluent populations? At present, we simply don’t know the answers to these basic but important questions.

In almost all conversations about education, it is not long before interest turns to questions of cause and effect. For example, in the case of teacher knowledge, do teacher content and pedagogical content knowledge explain differences in gains in student achievement? What are the effects on student achievement of increasing high school graduation requirements? What are the effects of requiring low-performing students to go to summer school? There is no end to these pressing and important questions about what works. Although education indicators do an excellent job of providing descriptive information about the inputs, processes, and outputs of education, they can only serve to stimulate hypotheses about what works. Asking questions about cause and effect pushes one toward intervention studies. If teacher knowledge is increased, what are the effects on gains in student achievement?

Although experiments provide the strongest evidence of cause and effect, they are expensive and difficult to conduct. Excellent information about cause and effect can also be obtained through sophisticated statistical analyses of longitudinal studies of students. Fortunately, once again we can turn to NCES for the necessary data. The first national longitudinal study, called the National Longitudinal Study of the High School Class of 1972, surveyed students from this class in their senior year of high school, and in subsequent years as well. Since then, there have been several national longitudinal studies, a few extending down into the lower high school grades and, most recently, one extending from preschool through the first several grades of elementary school. While the NCES indicators reported in The Condition of Education cannot provide answers to our “what works” questions of cause and effect, NCES national longitudinal studies have generated a great deal of insight.

Let me close with a couple of comments about NCES itself. I have chaired the Advisory Council on Education Statistics since 1994; and as long ago as 1985, I served on a National Academy of Sciences committee to evaluate the agency. I have worked directly with two NCES commissioners and one acting commissioner.

In 1985, NCES was struggling. I’m happy to report that the agency has addressed its earlier weaknesses and now produces timely and accurate education statistics on a wide variety of important issues. It has gained the respect of its sister statistical agencies in the government. Careful studies reveal the agency’s products are seen as useful by a wide variety of audiences. The agency’s longitudinal studies are of excellent quality and heavily used by researchers and policy analysts. The Condition of Education is one of the longest term continuing publications of the agency and a jewel in the crown of education statistics.

A strong government statistical agency requires political independence. Statistical facts must not be subject to political bias. At the same time, the agency has the responsibility to provide timely, accurate, and useful statistics, leaving interpretation of the causes behind these statistics to others. This is a delicate balance. The 2001 Condition of Education appears to be a product where the balance is working.