The 2019-2024 Outlook for Special Education Software in the United States
- June 2018 •
- 592 pages •
- Report ID: 5449848 •
- Format: PDF
This study covers the latent demand outlook for special education software across the states and cities of the United States. Latent demand (in millions of U.S. dollars), or potential industry earnings (P.I.E.) estimates are given across some 12,600 cities in the United States.
For each city in question, the percent share the city is of its state and of the United States as a whole is reported. These comparative benchmarks allow the reader to quickly gauge a city vis-à-vis others.
This statistical approach can prove very useful to distribution and/or sales force strategies. Using econometric models which project fundamental economic dynamics within each state and city, latent demand estimates are created for special education software. This report does not discuss the specific players in the market serving the latent demand, nor specific details at the product level. The study also does not consider short-term cyclicalities that might affect realized sales. The study, therefore, is strategic in nature, taking an aggregate and long-run view, irrespective of the players or products involved.
This study does not report actual sales data (which are simply unavailable, in a comparable or consistent manner in virtually all cities in the United States). This study gives, however, my estimates for the latent demand, or potential industry earnings (P.I.E.), for special education software in the United States. It also shows how the P.I.E. is divided and concentrated across the cities and regional markets of the United States. For each state, I also show my estimates of how the P.I.E. grows over time. In order to make these estimates, a multi-stage methodology was employed that is often taught in courses on strategic planning at graduate schools of business.
Another reason why sales do not equate to latent demand is exchange rates. In this report, all figures assume the long-run efficiency of currency markets.
Figures, therefore, equate values based on purchasing power parities across geographies. Short-run distortions in the value of the dollar, therefore, do not figure into the estimates. Purchasing power parity estimates were collected from official sources, and extrapolated using standard econometric models. The report uses the dollar as the currency of comparison, but not as a measure of transaction volume. The units used in this report are: US $ mln.
1.3 THE METHODOLOGY
In order to estimate the latent demand for special education software across the states and cities of the United States, I used a multi-stage approach. Before applying the approach, one needs a basic theory from which such estimates are created.
In this case, I heavily rely on the use of certain basic economic assumptions. In particular, there is an assumption governing the shape and type of aggregate latent demand functions.
Latent demand functions relate the income of a state, city, household, or individual to realized consumption. Latent demand (often realized as consumption when an industry is efficient), at any level of the value chain, takes place if an equilibrium is realized.
For firms to serve a market, they must perceive a latent demand and be able to serve that demand at a minimal return. The single most important variable determining consumption, assuming latent demand exists, is income (or other financial resources at higher levels of the value chain). Other factors that can pivot or shape demand curves include external or exogenous shocks (i.e., business cycles), and or changes in utility for the product in question.
Ignoring, for the moment, exogenous shocks and variations in utility across geographies, the aggregate relation between income and consumption has been a central theme in economics. The figure below concisely summarizes one aspect of problem.
In the 1930s, John Meynard Keynes conjectured that as incomes rise, the average propensity to consume would fall. The average propensity to consume is the level of consumption divided by the level of income, or the slope of the line from the origin to the consumption function.
He estimated this relationship empirically and found it to be true in the short-run (mostly based on cross-sectional data). The higher the income, the lower the average propensity to consume.
This type of consumption function is shown as "B" in the figure below (note the rather flat slope of the curve). In the 1940s, another macroeconomist, Simon Kuznets, estimated long-run consumption functions which indicated that the marginal propensity to consume was rather constant (using time series data). This type of consumption function is show as "B" in the figure below (note the higher slope and zero-zero intercept). The average propensity to consume is constant. For a general overview of this subject area, see Principles of Macroeconomics by N. Gregory Mankiw, South-Western College Publishing; ISBN: 0030340594; 2nd edition (February 2002).
Is it declining or is it constant? A number of other economists, notably Franco Modigliani and Milton Friedman, in the 1950s (and Irving Fisher earlier), explained why the two functions were different using various assumptions on intertemporal budget constraints, savings, and wealth. The shorter the time horizon, the more consumption can depend on wealth (earned in previous years) and business cycles.
In the long-run, however, the propensity to consume is more constant. Similarly, in the long run, households with no income eventually have no consumption (wealth is depleted).
While the debate surrounding beliefs about how income and consumption are related is interesting, in this study a very particular school of thought is adopted. In particular, we are considering the latent demand for special education software across the states and cities of the United States.
The smallest cities have few inhabitants. I assume that all of these cities fall along a "long-run" aggregate consumption function. This long-run function applies despite some of these states having wealth; current income dominates the latent demand for special education software. So, latent demand in the long-run has a zero intercept. However, I allow different propensities to consume (including being on consumption functions with differing slopes, which can account for differences in industrial organization, and end-user preferences).
Given this overriding philosophy, I will now describe the methodology used to create the latent demand estimates for special education software. Since ICON Group has asked me to apply this methodology to a large number of categories, the rather academic discussion below is general and can be applied to a wide variety of categories and geographic locations, not just special education software in the United States.
1.3.1 STEP 1. PRODUCT DEFINITION AND DATA COLLECTION
Any study of latent demand requires that some standard be established to define "efficiently served". Having implemented various alternatives and matched these with market outcomes, I have found that the optimal approach is to assume that certain key indicators are more likely to reflect efficiency than others.
These indicators are given greater weight than others in the estimation of latent demand compared to others for which no known data are available. Of the many alternatives, I have found the assumption that the highest aggregate income and highest income-per-capita markets reflect the best standards for "efficiency".
High aggregate income alone is not sufficient (i.e. some cities have high aggregate income, but low income per capita and cannot be assumed to be efficient). Aggregate income can be operationalized in a number of ways, including gross domestic product (for industrial categories), or total disposable income (for household categories; population times average income per capita, or number of households times average household income).
Latent demand is therefore estimated using data collected for relatively efficient markets from independent data sources (e.g. Euromonitor, Mintel, Thomson Financial Services, the U.S. Industrial Outlook, the World Resources Institute, the Organization for Economic Cooperation and Development, various agencies from the United Nations, industry trade associations, the International Monetary Fund, and the World Bank). Depending on original data sources used, the definition of special education software is established. In the case of this report, the data were reported at the aggregate level, with no further breakdown or definition. In other words, any potential services that might be incorporated within special education software fall under this category. Public sources rarely report data at the disaggregated level in order to protect private information from individual firms that might dominate a specific product-market. These sources will therefore aggregate across components of a category and report only the aggregate to the public. While private data are certainly available, this report only relies on public data at the aggregate level without reliance on the summation of various category components. In other words, this report does not aggregate a number of components to arrive at the "whole". Rather, it starts with the "whole", and estimates the whole for all states and cities in the United States (without needing to know the specific parts that went into the whole in the first place).
Given this caveat, in this report we define the sales of special education software as including all commonly understood products falling within this broad category, such as web-enabled software programs, customizable educational content, assistive technologies, and applications for assisting in learning, irrespective of product packaging, formulation, size, or form. Companies participating in this industry include Scantron, Crick Software, Merit Software, DynaVox Mayer-Johnson, SunGard, and Monarch Teaching Technologies.
In addition to the sources indicated below, additional information available to the public via news and/or press releases published by players in the industry was considered in defining and calibrating this category. All figures are in a common currency (U.S. dollars, millions) and are not adjusted for inflation (i.e., they are current values). Exchange rates used to convert to U.S. dollars are averages for the year in question. Future exchange rates are assumed to be constant in the future at the current level (the average of the year of this publication’s release in 2018).
This report was prepared from a variety of sources including excerpts from documents and official reports or databases published by the World Bank, the U.S. Department of Commerce, the U.S. State Department, various national agencies, the International Monetary Fund, the Central Intelligence Agency, various agencies from the United Nations (e.g. ILO, ITU, UNDP, etc.), and non-governmental sources, including ICON Group Ltd., Euromonitor, the World Resources Institute, Mintel, the U.S. Industrial Outlook, and various public sources cited in the trade press.
1.3.2 STEP 2. FILTERING AND SMOOTHING
Based on the aggregate view of special education software as defined above, data were then collected for as many geographic locations as possible for that same definition, at the same level of the value chain. This generates a convenience sample of indicators from which comparable figures are available.
If the series in question do not reflect the same accounting period, then adjustments are made. In order to eliminate short-term effects of business cycles, the series are smoothed using a 2-year moving average weighting scheme (longer weighting schemes do not substantially change the results).
If data are available for a geographic region, but these reflect short-run aberrations due to exogenous shocks (such as would be the case of beef sales in a state or city stricken with foot and mouth disease), these observations were dropped or "filtered" from the analysis.
1.3.3 STEP 3. FILLING IN MISSING VALUES
In some cases, data are available on a sporadic basis. In other cases, data may be available for only one year.
From a Bayesian perspective, these observations should be given greatest weight in estimating missing years. Assuming that other factors are held constant, the missing years are extrapolated using changes and growth in aggregate national, state, and city-level income.
Based on the overriding philosophy of a long-run consumption function (defined earlier), states and cities which have missing data for any given year, are estimated based on historical dynamics of aggregate income for that geographic entity.
1.3.4 STEP 4. VARYING PARAMETER, NON-LINEAR ESTIMATION
Given the data available from the first three steps, the latent demand is estimated using a "varying-parameter crosssectionally pooled time series model". The interested reader can find longer discussions of this type of modeling in Studies in Global Econometrics (Advanced Studies in Theoretical and Applied Econometrics V. 30) , by Henri Theil, et al., Kluwer Academic Publishers; ISBN: 0792336607; (June 1996), and in Principles of Econometrics, by Henri Theil John Wiley & Sons; ISBN: 0471858455; (December 1971), and in Econometric Models and Economic Forecasts by Robert S. Pindyck, Daniel L. Rubinfeld McGraw Hill Text; ISBN: 0070500983; 3rd edition (December 1991). Simply stated, the effect of income on latent demand is assumed to be constant unless there is empirical evidence to suggest that this effect varies (i.e., the slope of the income effect is not necessarily same for all states or cities). This assumption applies along the aggregate consumption function, but also over time (i.e., not all states or cities in the United States are perceived to have the same income growth prospects over time). Another way of looking at this is to say that latent demand for special education software is more likely to be similar across states or cities that have similar characteristics in terms of economic development.
This approach is useful across geographic regions for which some notion of non-linearity exists in the aggregate cross-region consumption function. For some categories, however, the reader must realize that the numbers will reflect a state’s or city’s contribution to latent demand in the United States and may never be realized in the form of local sales.
1.3.5 STEP 5. FIXED-PARAMETER LINEAR ESTIMATION
Nonlinearities are assumed in cases where filtered data exist along the aggregate consumption function. Because the the United States consists of more than 15,000 cities, there will always be those cities, especially toward the bottom of the consumption function, where non-linear estimation is simply not possible.
For these cities, equilibrium latent demand is assumed to be perfectly parametric and not a function of wealth (i.e., a city’s stock of income), but a function of current income (a city’s flow of income). In the long run, if a state has no current income, the latent demand for special education software is assumed to approach zero. The assumption is that wealth stocks fall rapidly to zero if flow income falls to zero (i.e., cities which earn low levels of income will not use their savings, in the long run, to demand special education software). In a graphical sense, for low-income cities, latent demand approaches zero in a parametric linear fashion with a zero-zero intercept. In this stage of the estimation procedure, a low-income city is assumed to have a latent demand proportional to its income, based on the cities closest to it on the aggregate consumption function.
1.3.6 STEP 6. AGGREGATION AND BENCHMARKING
Based on the models described above, latent demand figures are estimated for all major cities in the United States. These are then aggregated to get state totals.
This report considers a city as a part of the regional and national market. The purpose is to understand the density of demand within a state and the extent to which a city might be used as a point of distribution within its state.
From an economic perspective, however, a city does not represent a population within rigid geographical boundaries. To an economist or strategic planner, a city represents an area of dominant influence over markets in adjacent areas.
This influence varies from one industry to another, but also from one period of time to another. I allocate latent demand across areas of dominant influence based on the relative economic importance of cities within its state. Not all cities (e.g. the smaller towns) are estimated within each state as demand may be allocated to adjacent areas of influence. Since some cities have higher economic wealth than others within the same state, a city’s population is not generally used to allocate latent demand. Rather, the level of economic activity of the city vis-à-vis others is used. Figures are rounded, so minor inconsistencies may exist across tables.