abstract Summary of the argument in a scholarly work. Usually abstracts are limited to 100–200 words.

ad hoc model A mathematical model that is not derived from optimizing principles. Instead, the hypothesized relationships come from common sense or experience.

annotated bibliography List of references on a topic, each of which is briefly described and evaluated. An annotated bibliography also includes complete bibliographic information on each reference. Compare with critical review and research abstract.

annualized growth rate What the cumulative growth rate over a year would be of an actual growth rate over a shorter period of time. For example, what a monthly growth rate would be if it continued over twelve months.

argument An assertion or claim supported by reasons or evidence. See also sound argument.

auto, or serial, correlation Common problem of regression analysis. Autocorrelation means that the errors in each observation of the data sample are dependent on or correlated with each other. Autocorrelation artificially inflates the reported t scores, so that explanatory variables may appear to be statistically different from zero when they are not.






Boolean searching Keyword searching using the Boolean operators AND, OR, or NOT to narrow or widen the search.

browsing Searching for literature by manually examining documents or bibliographies for references. Contrast with keyword searching.






causal validity The degree to which a test indicates that a relationship is causal, rather than merely correlational. For example, to what extent can one be sure that X causes Y, rather than the converse, or that both X and Y are determined by some third variable? Instrument validity is one dimension of internal validity.

change, or first difference The difference in values of two adjacent observations. With time-series data, it is the subsequent value less the prior value.

coefficient estimates See estimated coefficients.

confounding, or control, variables "In a hypothesized relationship, confounding variables are other variables that affect the variable you are trying to explain. In order to correctly assess the hypothesized relationship, you need to control for the effects of those other variables. For example, if you believe that consumer spending depends on income, to assess that relationship, you must control for the effects of other variables, such as interest rates or savings, which might also affect consumer spending. See also multiple regression analysis. "

correlation A measure of the degree of linear association between two variables. The correlation coefficient ranges from positive one (a perfect positive relationship between the two variables) to zero (no relationship between the two variables) to negative one (a perfect negative relationship between the two variables.) Compare with covariance.

covariance A measure of how two variables vary together. It is related mathematically to the standard deviation and the variance. Compare with correlation.

critical review The summary and evaluation of the argument of a research report, along with complete bibliographic information to identify the report. Compare with annotated bibliography and research abstract.

critical thinking The evaluation of competing arguments based on their logic and evidence.

cross-section data Different observations of a variable at the same point in time; for example, average personal income across the fifty states in 2004. Contrast with longitudinal data and time-series data.






deductive reasoning Reasoning that starts from one or more general principles and derives specific predictions from them. See also valid deduction. Contrast with inductive reasoning.

dependent variable The concept we are trying to explain in a hypothesized relationship. Contrast with explanatory,or independent, variable.

derived data Data actually used in empirical testing, which have been manipulated from the source data. Contrast with raw data.

descriptive statistics Statistics that summarize the underlying data. Descriptive statistics include measures of central tendency, such as the mean, median, and mode, and measures of dispersion, such as the standard deviation and variance.

directories Hierarchically organized catalogs of information on the Internet on a variety of subjects. See also virtual libraries.

dummy variable A constructed variable in a regression analysis to capture qualitative, usually binary factors such as gender (male/female). Dummy variables have values of one (e.g., male) or zero (e.g., not male). Dummy variables can be used in the construction of an interaction variable.






economic (or scientific) significance The quality of a relationship that is large enough to matter in the real world. See also size of an estimated coefficient. Contrast with statistical significance.

editing Correcting the grammar, mechanics, and style of a piece of writing.

endogenous variable A variable that is determined inside the model being studied. Contrast with exogenous variable.

estimated coefficients The products of regression analysis, the estimated coefficients are estimates of the relationship between each explanatory variable and the dependent variable, ceteris paribus. See also sign of an estimated coefficient.

explanatory, or independent, variable A concept that is hypothesized to explain, affect, or cause the dependent variable. Independent variables can be exogenous or endogenous. Contrast with dependent variable; see also multiple regression analysis.

exogenous variable A variable that is determined outside the model being studied. As such, it is treated as a given in the model. Contrast with endogenous variable.

experimental research Research methodology that is usually characterized by two sample groups: an intervention or treatment group, and a control group. The control group is designed to be in every way identical to the treatment group except for the treatment. Thus, if the experiment between the two groups yields different outcomes, then that difference can be ascribed to the treatment. Contrast with survey, or nonexperimental, research.

external validity The degree to which the results of a study can be generalized to other situations, applications, or circumstances. Contrast with internal validity.






focus A feature of good writing, where the point of the document is crystal clear to the reader and the argument does not wander from this point.






good argument See sound argument.

growth rate See rate of change.






heteroskedasticity A common problem of regression analysis that occurs when the errors in each observation of the data sample, while independent from one another, have different variances. Heteroskedasticity artificially inflates the reported t scores, so that explanatory variables may appear to be statistically different from zero when they are not.






implicit price deflator A price index constructed by dividing the real value of a concept into the nominal value, making it an implicit price index, or deflator. See also price index and unit price.

independent variable See explanatory variable.

inference A conclusion reached as the result of reasoning logically about facts and relationships. See also warranted inference.

index number A unit-free weighted average designed to compare trends in prices or quantities since some base period. The raw data are transformed so that the value of the base period is 100. Two examples of index numbers include price indices and quantity indices.

inductive reasoning Reasoning that infers a general rule from one or more specific cases or situations. Contrast with deductive reasoning.

instrument validity The degree to which a test instrument measures what it purports to. Instrument validity is one dimension of internal validity.

interaction term A constructed variable in a regression analysis designed to capture the effects of two explanatory variables, for example age and race, interacting. The interaction variable is constructed by multiplying the two explanatory variables together. See also squared term and dummy variable.

internal validity The degree to which an impact observed (Y) can be attributed to the study variable (X); in other words, given the assumptions and evidence, one can deduce that X causes Y. See also instrument validity, relationship validity, and causal validity. Contrast with external validity.






keyword searching Searching for literature using search engines on the Internet or on specialized electronic databases. Compare with phrase searching. Contrast with browsing.

knowledge The meaning or interpretation of evidence.






level of significance The risk that a researcher is willing to take that the null hypothesis will be rejected when it is true. Commonly accepted levels of significance are 1 or 5 percent. A 10 percent level of significance is about the highest most researchers will accept. See also Type 1 error and Type 2 error.

literature (on a subject) Studies to date on a subject, both published and in working papers.

literature survey or review The process of identifying the major studies to date on a subject, or a written summary of those studies.

logical fallacy An argument that is flawed because the conclusion does not actually follow from the reasons stated, even though the argument is phrased in a way that makes you think it does.

longitudinal, or panel, data A sample of data that is essentially a cross section followed over time, where strictly speaking the cross section contains the same individuals in each time period. Otherwise, the sample is described as a pooled cross section. Contrast with cross-section data and time-series data.






maintained hypothesis The theoretical prediction of a model; what one hopes to prove when conducting a statistical test. If the evidence rejects the null hypothesis, then it confirms the maintained hypothesis.

mean A measure of central tendency, computed as the arithmetic average. Contrast with median and mode. See also descriptive statistics.

median A measure of central tendency defined as the middle value of a data sample with half the observations above the median and half the observations below. Contrast with mean and mode. See also descriptive statistics.

mode A measure of central tendency defined as the most frequent value. Contrast with mean and median. See also descriptive statistics.

moving average data-smoothing technique that replaces the actual observation in each period with an average of the n – 1 preceding data points.

mathematical reasoning An approach to theory development based on mathematical manipulation of premises to derive formal hypotheses.

micro data set Observations of individual economic agents such as individuals, households, or firms.

multicollinearity A common problem of regression analysis that occurs when two or more explanatory variables are highly correlated. As a result, the reported t statistics are artificially low, making it more difficult to prove statistical significance.

multiple regression analysis A statistical method of estimating hypothesized relationships while controlling for the effects of confounding variables. Regression is a technique for estimating the independent influences of each explanatory variable, and thus statistically maintaining ceteris paribus.






narrative reasoning An approach to theory development based on brainstorming about the topic.

nominal data Values expressed in terms of actual market prices. Nominal is also known as current dollars. Contrast with real data.

nonscience Fields of study in which arguments are assessed through logic and normative standards rather than empirical evidence. Contrast with science.

null, or statistical, hypothesis The hypothesis that is actually examined in a statistical test, for example, that there is no relationship between two variables. Contrast with maintained hypothesis.






optimizing model A mathematical model derived explicitly by assuming that economic agents maximize or minimize an objective function such as profits, utility, or costs.

organization A feature of good writing, where a series of nested arguments lead logically to the thesis as a conclusion.






parameter estimates See estimated coefficients.

per capita Per person. Some quantity (e.g., GDP) divided by the population gives the quantity per capita (e.g., GDP per capita).

percentage "The proportion, or share, expressed as a fraction of one hundred. For example, one-half is 50/100, or 50 percent. "

percent change "A change expressed as a percentage. See also rate of change. "

plagiarism Taking credit for someone else’s words or ideas, even when it’s unintentional.

prewriting Developing a written argument by exploring ideas and relationships. Also known as exploratory writing.

phrase searching Keyword searching on an exact phrase.

popular literature Summaries of scholarly studies, often written by journalists. Contrast with scholarly literature.

power of a statistical test The probability of correctly rejecting a null hypothesis when it is not true.

price index The index number that is essentially a weighted average of the prices of some type of good or service, where the weights tend to be either the corresponding quantities or the budget shares in a given year. See also implicit price deflator. Contrast with quantity index.

primary data source The person or organization that collected or constructed a data set, often using surveys. See also secondary data source.

proportion The share or fraction of a whole. See also percentage.






quantity index The index number that is essentially a weighted average of the quantities of some type of good or service, where the weights are the corresponding prices in a given year. Contrast with price index.






random sample of data A sample for which every item in the population has an equal chance of being selected. An equivalent definition is that every possible sample of the same size n has an equal chance of being the sample selected. A nonrandom sample of data doesn’t correctly reflect the underlying population of data; for example, there could be more above- or below-average observations than occur proportionately in the population. See also sampling error.

random variation "The variation in some data that is not systematically explainable. "

range A measure of dispersion of a data sample indicated by the lowest value in the sample and the highest value. See also standard deviation, variance, and descriptive statistics.

raw data The data as they exist at the source. Contrast with derived data.

rate of change A change expressed as a percentage; for example, the change divided by the initial value. See also percent change.

real data Values expressed in terms of prices as they existed in a base period. Real data exclude the effects of inflation since the base period. Real dollars are also known as constant dollars, or 1992 dollars, where 1992 is the chosen base year. Contrast with nominal data.

relationship validity The extent to which one can conclude on the basis of an empirical test that there is a statistical relationship. Relationship validity is one dimension of internal validity.

research The creation of knowledge.

research abstract A summary of the argument made by a research report, along with complete bibliographic information to identify the report. See also annotated bibliography and critical review.

research topic The broad subject area a research project will investigate.

research question The specific focus of a research project.

research hypothesis The proposed answer to a research question. The hypothesis is the result deduced from a researcher’s analysis. It is what is tested in an empirical research project.

revising Reviewing the information in an argument, rethinking the way it is organized, and looking for new patterns of meaning; more than just editorial correction.






sampling error The result of a nonrandom sample of data where the above-average values of some observations are not offset by the below-average values of other observations. As a consequence, the data appear to show a relationship that does not in fact exist. See also random sample.

scholarly literature Studies done by experts in a field and published in professional journals or books. Contrast with popular literature.

science Fields of study in which arguments are empirically testable. Contrast with nonscience.

scientific method The set of procedures for drawing valid, reliable, and objective conclusions about scientific hypotheses.

seasonal adjustment Data-smoothing technique by which normal seasonal fluctuations in the data have been removed. The results show any divergence from the normal seasonal pattern.

secondary data source An individual or organization that publishes data that it did not produce. Often the data are easier to use than the primary data. Contrast with primary data source.

sign of an estimated coefficient Important for testing hypotheses about relationships between variables, the (algebraic) sign shows whether a relationship is positive or negative. See also estimated coefficient and size of an estimated coefficient.

significance (of a research topic or question) The scientific importance or weight of a research topic or question.

significance level See level of significance.

simultaneous equations bias A common problem of regression analysis. In an ordinary least squares regression, all the explanatory variables are assumed to be exogenous to the model. In other words, the explanatory variables are hypothesized to affect the dependent variable, but the converse is not true: the dependent variable should have no effect on the independent variable. If causation runs both ways, the result is simultaneous equations bias and the coefficient estimates will be biased.

size of an estimated coefficient The magnitude of a hypothesized relationship; the size is important for assessing the economic significance of a relationship. See also estimated coefficient and sign of an estimated coefficient.

Social Science Citation Index A bibliographic search tool that allows backward and forward searches chronologically. That is, it identifies all the studies cited by a given study as well as all the studies that cite it.

sound argument A valid deduction in which all the premises are true. See also argument.

source data See raw data.

specification error A common problem of regression analysis, where the empirical model is incorrect, for example, because of omitting relevant variables or including irrelevant ones. In the former case, the estimated coefficients will be biased and the t statistics will be artificially reduced; in the latter case, the coefficients are unbiased but the t statistics are inflated.

squared term A constructed variable in a regression analysis used to turn a quadratic relationship that illustrates increasing or diminishing returns into a linear relationship. The variable is constructed by squaring the base factor, for example, Age_squared = age * age. See also interaction term.

standard deviation standard deviation—A measure of dispersion of a data sample defined approximately as the average amount by which a data point differs from the mean value of the sample. See also variance and descriptive statistics.

standard error A measure of dispersion in a data sample, normalized by the size of the sample. The standard error is defined as the ratio of the standard deviation to the square root of the sample size.

statistical hypothesis See null hypothesis.

statistical hypothesis testing Using methods of statistical inference to test hypotheses about relationships between variables. Typically, statistical methods are used to differentiate between alternatives called the null hypothesis and the maintained hypothesis. Commonly used tests include the t-test and the F-test. See also statistical significance.

statistical significance The quality of a statistical relationship that is strong enough to reject the conclusion that it is spurious, due entirely to random chance in the data. See also statistical hypothesis testing. Contrast with economic significance.

survey, or nonexperimental, research Survey research involves the passive observation and analysis of events as they occur in nature, while using statistical methods to account for outside factors. Contrast with experimental research.






time-series data Different observations on the same variable at different points across time; for example, U.S. personal income for the period 1954–2004. Contrast with cross-section data and longitudinal data.

time trend A constructed variable in regression analysis used to capture the effects of technological improvement or other trends over time.

Type 1 error Wrongly rejecting the null hypothesis. Contrast with Type 2 error.

Type 2 error "Wrongly accepting the null hypothesis. Contrast with Type 1 error. "






unit value An estimate of the price of something defined as the value of shipments divided by the volume of shipments, or the nominal value divided by the real value. Compare with implicit price deflator.






valid deduction A deduction in which the conclusion must follow from the premises.

validity See internal or external validity.

variance A measure of dispersion of a data sample defined as the square of the standard deviation. See also descriptive statistics.

virtual libraries Internet directories compiled by librarians or other experts. As such, they tend to have better sources of information than directories.






warranted inference A conclusion that follows from the evidence. See also inference.

warrants Unstated or underlying assumptions on which an argument stands.

writing to learn The idea that writing is more than a product; it is also a process or a tool for working through and making sense out of ideas you don’t fully understand.