You are on page 1of 56

Enrollment Forecast Planning Book for Billings School District 2

Revision: Original

3/3/2011

This Page Intentionally Left Blank

Revision: Original

ii

3/3/2011

Enrollment Forecast Planning Book

Table of Contents

Table Of Contents
1. 2. 3. 4. 5. 6. 7. 8. Project Background .............................................................................................................. 1-1 DDP Consulting Services Contract ...................................................................................... 2-1 DDP Study Execution........................................................................................................... 3-1 Birth Rate Analysis ............................................................................................................... 4-1 Mobility Factors Analysis .................................................................................................... 5-1 Private and Homeschools Research and Analysis ................................................................ 6-1 Facilities Research ................................................................................................................ 7-1 Redistricting Modeling ......................................................................................................... 8-1 Enrollment Actuals Screen Shots ...................................................................... A-1

Appendix A:

Revision: Original

iii

3/3/2011

Enrollment Forecast Planning Book

Table of Contents

This Page Intentionally Left Blank

Revision: Original

iv

3/3/2011

Enrollment Forecast Planning Book

Project Background

1. Project Background
Billings School District 2 (SD2) operating efficiency and financial projection accuracy depend upon consistent enrollment projections and redistricting models. Overall School Board, Superintendent, Central Office Leadership Team (COLT), and departmental motivations and expectations regarding these projections/models and their respective contributions thereto must be queried and documented. These become both the foundational underpinnings and strategic directives for the overall capability and process. Production of these projections/models must follow documented, standardized procedures. SD2 has invested heavily in industryleading enrollment projection and redistricting modeling software and services, from Davis Demographics and Planning (DDP). DDPs Desktop Planning Suite is an annuallylicensed software product. It is layered upon ESRIs ArcGIS Geographic Information System. It offers the trained sophisticated user a comprehensive data management, enrollment projection and redistricting platform. As stated on ESRIs website: A geographic information system (GIS) integrates hardware, software, and data for capturing, managing, analyzing, and displaying all forms of geographically referenced information. GIS allows us to view, understand, question, interpret, and visualize data in many ways that reveal relationships, patterns, and trends in the form of maps, globes, reports, and charts. A GIS helps you answer questions and solve problems by looking at your data in a way that is quickly understood and easily shared. SD2 expensed some $40,000 over the last three (3) years, training two executive secretaries on the DDP software. The results and capabilities did not satisfy enrollment forecasting inquiries from two school board members, the CFO and his two analysts. Dr. Jack Copps (then Superintendent) brought DDP President Greg Davis to Billings in June, 2010, for executive and boardlevel discussions regarding consulting services (available from DDP) to augment SD2 capabilities. Mr. Davis applauded the secretaries efforts, but he concluded that his software requires extensive expertise for complete capability. The Superintendent, COLT, and the School Board depend upon such capability now more than ever, as they research, evaluate and plan myriad district facility/financial scenarios. Dr. Keith Beeman (present Superintendent) then requested that two specific SD2 employees negotiate a consulting services contract with DDP. Daniela Walsh, MBA, is a Financial Analyst for CFO Harper. Richard Dews is a computer scientist and software developer in Dr. Jensens Adult Education Department. Harper previously directed Dews and Walsh to research enrollment numbers and their data sources, since late 2009. They obtained historical OPI enrollment numbers by year, by school, and by grade. Dews developed an enrollment trends analysis model (using MS Access and programmatic code). See sample screens from this model in Appendix A. Bar graphs were generated to depict trends and anomalies, leading to questions of causality. Dews obtained a 30day free trial of the software from DDP, installed it on his PC, and requested data from Primary/Secondary. Through further research, Dews and Walsh determined: SD2 past enrollment numbers were available from different sources, and that substantial inconsistencies appeared in sameyear number sets, depending upon source and acquisition method used
1-1 3/3/2011

Revision: Original

Enrollment Forecast Planning Book

Project Background

Yellowstone County birth rates could easily be broken down into eight (8) related ZIP codes, and such data was available from DHHS Vital Statistics Birth rate variability yeartoyear within any one of these ZIP codes was substantial, and myriad statistical methods could be considered for projections with large longrange differences DDP software uses a concept of Study Areas geographical regions wherein students are assigned attendance to a specific school (primary, intermediate, and secondary) o SD2 historical selection of Study Area boundaries and student/school assignments were often inconsistent in size and locality of associated schools DDP software allows a broad range of variables for birth rates and mobility factors, which may be entered as general averages or granular specifics by grade, year, per study area o Enrollment projection models will vary substantially, depending upon these entered variables

CFO Harper arranged to have Dews and Walsh present their findings to School Board chairperson Barbara Bryan. Dews proposed that, should DDP software continue to be used within SD2, a Planning Book should be developed (similar to the CFOs Budget Book) to standardize and document the enrollment projections process and methodology for comprehensiveness and consistency.

Revision: Original

1-2

3/3/2011

Enrollment Forecast Planning Book

DDP Consulting Services Contract

2. DDP Consulting Services Contract


Soon after Dr. Beeman assumed responsibilities as Superintendent of Billings Public Schools, he called a meeting to discuss the Districts enrollment forecasting needs and DDPs consulting services offerings. At his request, those in attendance included: Barbara Bryan: SD2 School Board Chairperson Dr. Keith Beeman: Superintendent Kathy Olson: Executive Director of Primary Education Marta McAllister: Primary Education Executive Secretary, DDP Software Trainee/User Thomas Harper: CFO Daniela Walsh: MBA, Financial Analyst for Mr. Harper Richard Dews: Adult Education Business Instructor, Computer Scientist

The attendees concurred that SD2s then-existing DDP software operation expertise was insufficient to produce comprehensive enrollment forecasts, to satisfy present and future District planning needs. It was decided that SD2 may further benefit from the objective third party expertise and recommendations supplied by DDP, a noted industry expert in the field of enrollment forecasting. Dr. Beeman requested that the DDP project results also include specific NCLB student attributes. These were not originally included in SD2 historical DDP software use. Dr. Beeman assigned Harper, Walsh and Dews the responsibility to negotiate a consulting services contract with DDP. This was done to ensure reliable tactical and strategic long-term enrollment forecasts and associated implications. As they negotiated a consulting services agreement with DDP, Dews and Walsh sought project details for the Planning Book, including data sources and formats, methodology, etc. Those details would enhance SD2s capability to maintain the resulting and future enrollment/redistricting models. SD2 retained DDP in August 2010 for a comprehensive enrollment forecast and buildout study. (See DDPs originally proposed services contract and SD2s negotiated and executed contract and services proposal). Upon contract execution, DDPs Project Manager David Kaitz requested specific information to support his efforts. Harper was assigned authority over project management and DDPs efforts, and Dews suggested crossdepartmental collaboration in the form of a Forecast Team to be staffed as follows: Gail Harms: Enrollment Data Acquisition (Technology) Marta McAllister and Kathy Gonzales: Existing SD2 DDP/GIS Configurations (Primary/Secondary) Daniela Walsh and Richard Dews: Project Management, Vendor Management, and Process Documentation (Business/Finance)

Revision: Original

2-1

3/3/2011

Enrollment Forecast Planning Book

DDP Consulting Services Contract

Dews suggested the following flowchart, along with standard contract and Statement of Work activities/content, for the collaborative project.

Revision: Original

2-2

3/3/2011

Enrollment Forecast Planning Book

DDP Study Execution

3. DDP Study Execution


The DDP study was designed to loosely follow their included timeline. The Forecast Team used an augmented version of the timeline, which contained parallel efforts on some tasks for Forecast Team training and objective comparative analysis to DDPs submissions. See both timelines here. Initial Excel [Enrolled] Student Data Files for 20072009 were sent to DDP for analysis and consideration. DDP processed this data into single-year summary signoff sheets, and then they requested the Forecast Team to review the counts of students per grade per school. If validated, Walsh was to sign these sheets and authorize DDP to utilize the referenced data as foundational input for forecasting purposes. However, Dews found sizable anomalies in the counts, after analyzing the numbers through a spreadsheet. A lengthy iterative process ensued, wherein the Forecast Team refined and standardized the enrollments data for all requested years. Harms documented the final consistent Student Data File acquisition process. Final Excel Student Data Files were provided to DDP, and subsequent Student Data File Verification Forms were executed. With process input from Kaitz, McAllister and Gonzales regeocoded the 2009 enrolled students. The Forecast Team obtained one additional networked license for the DDP Schoolsite Desktop Planning Suite ($2,600), for these reasons: Facilitate Suite training, as suitable for Harms, Dews and Walsh Enable simultaneous geocoding of four (4) years of enrolled SD2 students o McAllister and Gonzales spent weeks regeocoding 2009 with Kaitz suggestions, but did not deem necessary the regeocoding of 2007 & 2008 o Harms, Dews and Walsh deemed necessary the regeocoding of 20072009 for: Training Harms, Dews and Walsh on DDP software and the geocoding process Ensuring data consistency across all affected years Incorporating requested NCLB student attributes Incorporate latest geocoding instructions from DDP Augmenting, if necessary, McAllisters geocoding process documentation Diversifying the objectivity of the geocoding efforts in SD2 Maintaining project schedule integrity without slippage, in case original geocoded files were found in any way inferior or unacceptable Mitigate financial impact of $5,600 per year for DDP to geocode additional years beyond the contracted 2010 obligation

Kaitz received and analyzed the Forecast Teams final geocoded files, and ultimately compared McAllisters and Dews 2009 files. Differences were deemed negligible, and Kaitz accepted Dews 20072009 geocoded files as final project input for consistency reasons. Through the above efforts, Dews demonstrated an acceleration of SD2s annual geocoding process from multiple weeks to less than one day. These efforts resulted in a finalized SD2 Geocoding Instructions document. Final DDPvalidated GIS Streets, Schools, Plats, Study Areas, and Geocoded Student Points (20072010) were then obtained from Kaitz. These were installed and layered for ease of use on Harms, Dews and Walsh PCs for training purposes. Dews conducted several training sessions for Harms and Walsh, both on ESRIs GIS system (ArcCatalog and ArcMap) and SD2s new
Revision: Original 3-1 3/3/2011

Enrollment Forecast Planning Book

DDP Study Execution

geocoding process. Dews desired to understand the scope of DDPs data acquisition, and the strategic impact their choices and methods made to their anticipated results report. He performed extensive analysis of the following mainstays of DDP data and methodology: Birth Rates Mobility Factors

Results of the above efforts, detailed in subsequent chapters of this book, allowed Dews to advise Harper on a broad range of historical data, implications of specific choices selected therefrom, and objective considerations when reviewing DDPs draft and final study reports. As a result, Harper queried an explanation from Kaitz regarding DDP initial draft study results incongruity with historical overall trends. That written query, combined with a subsequent teleconference including Dews and Walsh, resulted in Kaitz/DDP reducing their projection numbers substantially. A second draft report was produced by DDP, and it was distributed to COLT members during the week of January 17th, 2011. A third (more comprehensive) draft report was received from Kaitz by the Forecast Team on the evening of January 23, 2011. It was emailed to COLT members and a hard copy was hand delivered to Dr. Beeman on January 25, 2011. Shortly after the study commenced in August, David Kaitz held a meeting with the Forecast Team and several COLT members. During that meeting, and at multiple occasions since, Kaitz requested SD2 school facility capacities (optimum and maximum) preferably by board or district policy. To date, no such policy or statistics have been produced. Two SD2 internal (possibly vendor-supported) follow-on studies are recommended by Harper, Dews and Walsh: Facilities Study Redistricting Study

A DDP teleconference or videoconference is scheduled for February 4, 2011 facilitating draft report content questions and answers for all report recipients. Input from that pending interchange will allow DDP to produce their final report, due by mid-February 2011.

Revision: Original

3-2

3/3/2011

Enrollment Forecast Planning Book

Birth Rate Analysis

4. Birth Rate Analysis


DDPs User Guide for enrollment forecasting (pages 37-39) directs that a specific process be followed when calculating birth rates and their impact upon kindergarten-specific enrollment projections. It states that most Kindergarteners are five (5) years old, and it offers a methodology for using past birth rates by ZIP code to formulate future birth rates. As stated in the User Guide: The first step in calculating birth factors is to obtain the number of births that occurred within the district for most recent five years. In most cases, you can obtain the number of births, collated by zip codes, from your State's Department of Health and Statistics. In this example, the district contains three zip codes - 92625, 92626, and 92627. After the number of births, by year, by zip code are obtained and totaled, the percentage of change from the base year (1997) is calculated by dividing future years (1997-2001) by the base year (2002). Because data for 2002 and beyond would not yet be available, birth factors would be estimated based on past trends. Here, we estimated a continued decline in births at 93% of the base year.

The Forecast Team obtained audited ZIP code-based annual births for each SD2-covering Yellowstone County ZIP code, from Bruce Schwartz at the Montana Department of Health and Human Services (DHHS) Vital Statistics unit. The Team analyzed data for all associated ZIP codes from 1989 through 2009, producing group and individual charts. Kaitz obtained the same information, but due to DDPs methodology he used only a subset of recent years (2004 through 2009) and primary ZIP codes (59101, 59102, 59105, and 59106 combining 59102 and 59106).

Revision: Original

4-1

3/3/2011

Enrollment Forecast Planning Book

Birth Rate Analysis

These numbers, and their source methodology of divination, are validated by Kaitzs actual birth rate charts included in his December 9, 2010 draft projections document (see below). Kaitzs future years are divined through a rolling four (4) year average, reseeding more averaged years with each iteration. This causes a normalization effect, or flattening.

Revision: Original

4-2

3/3/2011

Enrollment Forecast Planning Book

Birth Rate Analysis

Utilizing DDPs documented methodology for devising birth rates, we can predict Kaitzs results even beyond 2020:

Revision: Original

4-3

3/3/2011

Enrollment Forecast Planning Book

Birth Rate Analysis

In the following charts, you can see the effects of this normalization graphically. Next to each DDP projection is a historical 21-year chart for the same ZIP-oriented births, with a linear trend line added for clarity.

Revision: Original

4-4

3/3/2011

Enrollment Forecast Planning Book

Birth Rate Analysis

Revision: Original

4-5

3/3/2011

Enrollment Forecast Planning Book

Birth Rate Analysis

Dews continued beyond birth rate divination into kindergarten student counts analysis. Below, see the chart of actual enrollment counts of kindergarten children for each SD2 and outlying elementary school, from 1993 through 2009. Schools are alphabetized into rows, and yearly counts are colored in gradients from green (lowest count for that school) to magenta (highest count for that school) for the range of years.

Revision: Original

4-6

3/3/2011

Enrollment Forecast Planning Book

Birth Rate Analysis

These numbers allow us to graph the historical kindergarten counts. Below, see the bar chart depicting SD2, outlying, and summed actual kindergarten historical totals, and deduce your own trend interpretation(s).

Revision: Original

4-7

3/3/2011

Enrollment Forecast Planning Book

Birth Rate Analysis

Finally, we can derive actuals/trend charts which paint a broader picture still.

Revision: Original

4-8

3/3/2011

Enrollment Forecast Planning Book

Birth Rate Analysis

We observe from analysis of the data, obviated through these charts, that: SD2 kindergarten counts are only now returning to where they were twenty (20) years ago SD2 percentage of five (5) year olds arriving in our classrooms is declining A substantial downturn in SD2 kindergarten counts existed for a decade Major difference in trend and enrollment projections would be possible if we only looked back five (5) years, instead of 20 years, possibly indicating a falsely exaggerated growth trend

A more detailed analysis of the chart on page 4-5 is necessary. Before alarm is raised over magenta segments towards the right of any school row, one must factor in facility capacity, recent or planned annexes or additions, and load balancing with other facilities. Facility capacities were requested by Kaitz early in this study, directly to the CFO and Executive Directors of Primary and Secondary Education. To date, no such information has been finalized, established as policy, or provided to Kaitz for impact upon his analyses and recommendations. Additionally, look back along school rows to note the similarity of general range in the last two decades. Perhaps that facility has housed successfully a similar number of kindergarten students before. With a little effort, a process may be developed to produce kindergarten student counts and individual/averaged ages (see following page). Charts may be colored in gradients to depict student quantity trends, and age ranges for study areas feeding specific schools. The charts on the following page depict kindergarten student counts and ages within study areas feeding SD2 elementary schools, from 2007 through 2010. These are just a few of the myriad examples of interesting ways to look at our demographic GIS data. This research and analysis provided greater understanding of overall birth rates and kindergarten trends. This led to positive collaboration with and management of the vendor, when reviewing their submitted draft projections.

Revision: Original

4-9

3/3/2011

Enrollment Forecast Planning Book

Birth Rate Analysis

Revision: Original

4-10

3/3/2011

Enrollment Forecast Planning Book

Mobility Factors Analysis

5. Mobility Factors Analysis


DDPs User Guide for enrollment forecasting (pages 40-42) directs that a specific process be followed when calculating mobility factors the net effect of the migration of students into and out of the district from existing homes. The guide instructions continue: These factors help account for housing resales, renter migration, private school transitions, and high school drop-outs. If you have at least four consecutive years of geocoded student data (which represents three years of enrollment change), the mobility factors can be calculated down to the study area level. The factors are derived from comparing student counts in study areas where there has been no new residential development for the past four years. Mobility is calculated for each grade in the district. Comparison is made between students in one grade with the numbers of students in the next grade the following year. For example, if your historical data begins in 1997, you would compare the total Kindergarten students of 1997 with that of 1st grade students of 1998 (the next year of students). All things being equal and with no new development in the area of comparison, the number of 1st grade students should be equal to the Kindergarten class from the year before assuming they all moved up a grade and stayed in the same area. NOTE: In general, mobility factors should have a tendency to vary slightly higher or lower than 1.00 unless there are extreme drop-outs or movement from private schools or other factors. If your calculated mobility factors are extremely higher or lower than 1.00, you may wish to check your enrollment counts and/or factor calculations. During the course of the DDP study, the Forecast Team developed consistent processes for creating annual enrolled student data files and associated geocoded data sets. These were subsequently provided to DDP in two forms: Excel-based Student Data Files (SDF), containing enrolled student demographic information, for October 2007, 2008, 2009, and 2010. See detailed Forecast Team evolution of related data acquisition experiences and finalized instructions for SDF creation: o Harms DDP Project Background o Student Data File Data Collection Process o Self-Contained Student Data Collection Process ArcGIS (ArcEditor and ArcCatalog) based Geocoded Student Files, derived from the above Student Data Files, for 2007-2009 (DDP generated, by contract, the 2010 geocoded student file)

The process of importing a single annual Excel Student Data File into ArcGIS, and converting its contents into meaningful geospatial points, is geocoding. An ArcGIS address locator is defined, associating map-referenced address components with source data file fields or columns. That address locator is used by ArcGIS in a largely automated process of lookup, matching student addresses with known locations on a map. However, the base street/address reference does not always match the source data field contents especially when street names are misspelled, apartment or trailer numbers are used, etc. Those exceptions are handled manually by the ArcGIS
Revision: Original 5-1 3/3/2011

Enrollment Forecast Planning Book

Mobility Factors Analysis

operator, searching the roads dataset for possible matches or likely typos, etc. The intent is to match as many of the students with actual address points as possible. Some students are out of the area, under protective orders, or using P.O. boxes, etc. In those cases, the respective records are tagged with codes representing Un-Matched or Out-of-Area, etc. The process may be further improved through experience, with the operator noting occasional consistencies in the types of student address corrections required. Those may be corrected in the original data collection process, while still in Excel format, or they may be handled during the geocoding steps either programmatically (with lookup conversion/replacement tables), etc. Increased automation would be the result of such data-cleaning efforts. Even without such expanded automation, however, Dews was able to accurately geocode three (3) years of student enrollment records (roughly 15,000 students per year) in less than one (1) day with no training. Clearly there are vast ranges of productivity possible when using such sophisticated software, depending upon the background, experience, education and resulting expertise of the operator. As stated in Chapter 1, Project Background, several members of the Forecast Team desired to geocode 2007 through 2009 for experience, education, and objective process/results comparison reason. We desired to obtain a level of understanding for methodology and data involved, allowing us to better manage the vendor and evaluate and question their draft findings. DDPs Kaitz decided to develop student mobility factors based upon actual student points within the combined geographical study areas (defined regions of land) feeding SD2s four (4) middle schools (Will James, Lewis and Clark, Riverside, and Castle Rock). Dews desired to further explore, learn, and evaluate the methods for divining mobility factors, so he tasked himself with creation of averaged mobility factors for the grouped feeder areas of the twenty-two (22) SD2 elementary schools. Starting from scratch, with no DDP or ArcGIS training, Dews first decided that a scope of analysis should be set at the elementary school feeding study areas the most granular without going to the individual 200+ study areas. He exported a subset of the geocoded student files for 2007-2010 (2010 obtained from Kaitz), and drew them into a new Microsoft Access database. From there, using SQL queries, he produced simple lists of student counts by grade by study area (for each year). He then wrote an Excel program to parse through these lists and accumulate total students per grade per study area for each of the four years. He then wrote a program to calculate the percentage of difference from year to year for subsequent grades (mobility factors). These mobility factors for each year-to-year matriculation of student counts by incremental grades were then averaged and formatted into cumulative matrices with color gradient for high-level analysis and pictorial review. The results were stunning and educational, even provocative. Such pictures allow viewers to be drawn to obvious trends and conclusions of student movement from grade to grade within these 22 groupings of land.

Revision: Original

5-2

3/3/2011

Enrollment Forecast Planning Book

Mobility Factors Analysis

Below, we see the resulting 22 areas of land feeding SD2 elementary schools and the averaged mobility factors of student movement from grade to grade within them.

With heavier gradients of growth depicted in deepening shades of green, and heavier gradients of loss depicted in deepening shades of red, look across any given school feeder area row to deduce highest and lowest four year averaged mobility factors for that geographic region. Averages are made for K-6, 7-8, 9-12 and total (TTL) for trend analysis. For example, although Arrowhead Elementary does not presently host kindergarten students, the nineteen (19) defined geographical study areas that feed elementary students to Arrowhead contain students ranging from kindergarten through high school seniors. Simple color observation naturally draws the eye to anomalous areas of the chart, where percentage of growth or loss can then be considered and compared. We will replicate the chart on the next page, with call-outs posing review possibilities.

Revision: Original

5-3

3/3/2011

Enrollment Forecast Planning Book

Mobility Factors Analysis

c d

a) The Arrowhead region stands out with a 21.8% average growth in student counts when moving from kindergarten to 1st grade, with Poly and Boulder running close behind b) Boulder (7.7%), Arrowhead (7%), and Meadowlark (5.7%) regions rank highest in averaged elementary student resident growth, while Washington (-7%) leads declines c) Substantial influxes of enrolled students from 8th to 9th grades within these areas indicate growth from external sources such as outlying schools, private schools and homeschools d) Substantial exits of enrolled students from 9th to 10th and 10th to 11th indicate loss to drop-out, homeschools, private schools, etc

Revision: Original

5-4

3/3/2011

Enrollment Forecast Planning Book

Mobility Factors Analysis

Coloring by gradient by grade vertically, with highest growth in deepest red (space requiring) and highest loss in deepest green (space creating), shows another picture (see above). When Dews started down the path of such data analysis, strategic questions of What should be considered? and What will those considerations require in data sources, processing, and formatting require/allow? were accommodated. A planned side-effect of this kind of research and resulting expertise/perspective was Dews ability to reverse engineer close approximations to DDPs divined mobility factors. As soon as Dews received Kaitzs first draft projections, he went to work reverse engineering DDPs birth rate and mobility factors divination. With Dews ability to calculate actual averaged mobility factors from the prior four (4) years, he factored that perspective in with his research and trends analysis of actual counts from the prior twenty-one (21) years. This provided better insight with which to analyze DDPs draft projections and advise Harper.

Revision: Original

5-5

3/3/2011

Enrollment Forecast Planning Book

Mobility Factors Analysis

Kaitzs first draft projection for SD2 enrollment counts included the following table.

Utilizing the experience and expertise gained from the prior mobility factors research, Dews reverse-engineered an approximation of the birth rates and mobility factors at play in the above overall projection.

Revision: Original

5-6

3/3/2011

Enrollment Forecast Planning Book

Mobility Factors Analysis

Reverse-engineered birth rates and mobility factors from DDPs first draft projections:

Revision: Original

5-7

3/3/2011

Enrollment Forecast Planning Book

Mobility Factors Analysis

Knowing Kaitzs birth rate projection methodology and values, and understanding DDPs software (with Kaitzs validation) design to apply birth rates to the seed year (not a rolling year-to-year application), we can reverse engineer the general impact of his projected residential development. Knowing the basis of matriculating students flowing through a grade-to-grade year-to-year matrix, we can reverse-engineer generalized mobility factors for Kaitzs projections, too. All of the comprehensive preceding research, and its resulting expertise and perspective, extended Dews ability to advise Harper on DDPs first draft projections. As a result, Harper sent a letter of inquiry to Kaitz and held a subsequent teleconference to question DDPs projected student count growth rates. Here is the charted historical count of district-wide students from 1993 through 2010, compared with Kaitzs first draft projections from 2011 through 2020 (see below). One can easily see a trend anomaly, when comparing the left side of the arc (reality) to the right side (forecast).

Revision: Original

5-8

3/3/2011

Enrollment Forecast Planning Book

Mobility Factors Analysis

After some discussion and his independent subsequent review, Kaitz responded with a modified second draft projection. From his December 30, 2010 email: As promised, we spent part of the past week reviewing various ways to revise the original draft 10-year projections we sent to you on December 9th to better reflect a more realistic pattern for the next ten years. The attached District-wide projection is our next and hopefully most acceptable version. We wound up dropping the birthrates down [3]%; this is justified mainly due to the fact that many of the children born in your area simply do not show up 5 years later in your schools. This adjustment helps to address this issue. In addition, DDP reduced ALL of the Mobility Factors by 1%. This was mainly done to show a more realistic increase in the projections (not the large, unprecedented increases we were projecting in our original draft). I think, from DDPs perspective, we can live with these adjustments. Tell me what you think. If these are also acceptable on your end, then notify me as soon as possible and I can follow-up with the attendance area projections to review. David Kaitz, Project Manager See the impact of these changes, in the chart below.

For a complete comparison of each grade series from first draft to second draft, click here.

Revision: Original

5-9

3/3/2011

Enrollment Forecast Planning Book

Mobility Factors Analysis

Reverse-engineered birth rates and mobility factors from DDPs second draft projections:

Revision: Original

5-10

3/3/2011

Enrollment Forecast Planning Book

Mobility Factors Analysis

If we compare the reverse-engineered DDP first draft rates and factors with their second draft, here is the actual difference.

Revision: Original

5-11

3/3/2011

Enrollment Forecast Planning Book

Mobility Factors Analysis

Given the above validation, we can be objectively sure that Kaitz did in fact reduce birth rates by precisely 3% across the board. Additionally, it does appear that the lions share of mobility factors have been reduced by roughly 1%. His were taken at a more granular level, resulting in the tiny discrepancies seen above when rolling up into summary counts. However, why the resulting increase in reverse-engineered 9th grade mobility factors? Additionally, one may ask the following strategic questions: Realizing that the standard DDP methodology, as stated in their software User Guide, incorporates 4-5 year look-backs on actual data, why were the available 21 years of birth rates and 18 years of enrollment actuals not used for trend analysis and comparison at DDP? After Dews performed such analysis and shared the findings with Kaitz, why was the response a simple flat decrease of birth rates by 3% and mobility factors by 1%? What is the probability that the generally-reduced second draft enrollment projection is more or less accurate than simple trend continuance? Based upon what data, methodology, and formulation? What would have been the SD2 response to the original DDP first draft projection, if not for this additional research and resulting challenge? What would have been the resulting final product, and what would have been its accuracy and value?

Revision: Original

5-12

3/3/2011

Enrollment Forecast Planning Book

Private and Homeschools Research and Analysis

6. Private and Homeschools Research and Analysis


During both birth rate and mobility factor research phases, it became apparent that SD2 students were coming and going in and out of our district. Excluding students departing from the area altogether, other avenues of education include private and homeschools. Harper obtained a cursory list of annual enrollments for some of the area private schools. Dews contacted Sherry Long, at the County Superintendents Office, and obtained as comprehensive a list of these enrollments as reasonable. A significant amount of time was spent entering the photocopied tallies, and devising a roll-up spreadsheet to facilitate proper analysis and trend research. The largest student bodies in private school are found within the Catholic system, at Billings Central High School, and St. Francis Primary/Intermediate/Upper. Second largest counts are from Trinity Lutheran. Decreasing from there are Billings Christian School, Yellowstone Academy, Central Acres, and more. Private school enrollment data was not readily or consistently available in a useful way until 1992, but more comprehensive and desirable comprehensive lists are available from 2003 forward. Charting this information, we get the following line graph.

Annual homeschooled student counts were available with limited detail as early as 1988. However, several different methods were used by the County Superintendent to obtain and catalog such counts over the next 22 years. As a result, minor variances should be expected in the direct trend accuracies, but the overall data is relatively sound. See the homeschooled student counts, by year, charted on the following page.

Revision: Original

6-1

3/3/2011

Enrollment Forecast Planning Book

Private and Homeschools Research and Analysis

And below we have the summary chart of private and homeschooled student counts, by year. Again, use caution when depending upon the private school counts prior to 2003.

Revision: Original

6-2

3/3/2011

Enrollment Forecast Planning Book

Facilities Research

7. Facilities Research
In early January, 2011, Harper loaned Dews a notebook with historical elementary classroom counts under different timeframes and configurations. One set depicted school capacities and classrooms as of 2007, with a 25:1 student-to-classroom ratio. The second set depicted classroom counts as surveyed by respective school administrator, but still with a 25:1 student-to-classroom ratio. The final set of numbers dropped the ratio to 23:1. See below these counts charted, colorized with gradients of green (lowest total school enrollment count) to red (highest total school enrollment count). Note the increasing appearance of red as each iteration of restrictions is incorporated from chart to chart. Decrease in Principal-qualified quantity of classrooms restricts capacity, followed by ratio reduction from 25 to 23 to 1. The results indicate a self-imposed appearance of crowding. After these worksheets were generated, a comprehensive all-schools Excel workbook developed by Hulteng Engineering, Inc. was forwarded to Dews from Rich Whitney, Director of Facilities. Dews copied all content and further standardized formats, allowing him to write a parsing program to support importing all school square footage measurements and room counts (by type) into a new Microsoft Access database. Dews then posed the following inquiry to Whitney, in hopes of continuing the analysis with more data. No response has yet been received. We are curious as to the utilization of specific rooms by category, such as Music, Science Labs, Computer Labs, etc. Are there other usage notes or more extensive descriptions for the individual area items beyond that which is summarized within the sheets in this workbook? We notice that the listing of Work and Classroom spaces within Alkali, Arrowhead, and Beartooth do not break down the large open classroom spaces into smaller actual current/optimum/maximum classroom utilization. Is that information available elsewhere? We are interested in up-to-date floor plans, per facility and per addition and/or floor, which Hulteng or others produced in a similar vintage. We would also hope that they were drawn to a consistent scale. Could you provide any of the above information, or direct us to a suitable source(s)? Thank you. A proper, extensive facilities study is needed to supplement the present enrollment forecast projections. As DDPs Kaitz requested, maximum and optimum capacities per school must be established. Hultengs workbook does not break down the large work spaces within Beartooth, Alkali Creek, and Arrowhead elementary schools. Moreover, consistent-scaled floor plans for all facilities, including additional floors, annexes, additions and relocatable/temporary spaces, should be created. This information is crucial to the establishment of district (and localized facility) capacity for load management and use/repair/expansion prioritizations.

Revision: Original

7-1

3/3/2011

Enrollment Forecast Planning Book

Facilities Research

Revision: Original

7-2

3/3/2011

Enrollment Forecast Planning Book

Facilities Research

Revision: Original

7-3

3/3/2011

Enrollment Forecast Planning Book

Facilities Research

Revision: Original

7-4

3/3/2011

Enrollment Forecast Planning Book

Redistricting Modeling

8. Redistricting Modeling
Through extensive work with previous and additional SD2 GIS information, Dews was focused repeatedly on the established geographical area boundaries that define study areas (SAs). SA definitions dictate which students shall attend which schools. Districts often adopt historical boundary definitions when developing their first GIS maps. SD2 had already broken up its overall boundary into 200+ SAs. Early into the DDP study, Kaitz suggested and implemented splits of several large SAs into smaller units to facilitate better analysis and planning. DDPs User Guide for enrollment forecasting states on page 29: Study Areas are the building blocks of a school district. Study areas are geographically defined following logical boundaries of the neighborhood and used for gathering data. Each of these areas are then coded to the existing school boundaries. By using these building blocks or study areas, various groupings of these areas can be used to simulate attendance boundary changes and can be useful for deciding where to open or close schools. Study Area Boundary Line Criteria: o o o o o o o o o o 1) All major streets (existing or future) 2) All railroads 3) All drainage channels or waterways 4) Lines between developed and undeveloped portions of the district 5) Lines between different types of zoning 6) Unusual terrain features and impediments to pedestrian traffic 7) Present school attendance zones 8) City limit lines 9) Different types of housing such as old versus new, apartment versus single family 10) Census geography

Study areas should be small enough to allow flexibility in modifying boundaries. DDP suggests no more than 50-100 students per study area. Although the number of study areas per district varies, the general rule is, more study areas equals more flexibility. Urban districts will require more study areas due to population density. The opposite is the case for rural areas. Study area boundaries in established neighborhoods, however, should never change once established. Study areas in undeveloped areas can be split as development becomes evident and prior to students moving in. Study areas can be broadly considered in groups, feeding specific high schools, middle schools, or elementary schools. They can also be identified individually by number and location. Thanks to the GIS, data is easily referenced in related groups, highlighted with color, etc. For the sophisticated user, datasets may be joined, related, queried, and exported for additional programmatic analysis within Access or Excel.

Revision: Original

8-1

3/3/2011

Enrollment Forecast Planning Book

Redistricting Modeling

County map extent:

Revision: Original

8-2

3/3/2011

Enrollment Forecast Planning Book

Redistricting Modeling

County map with SD2 and outlying school study areas:

Revision: Original

8-3

3/3/2011

Enrollment Forecast Planning Book

Redistricting Modeling

County map cropped to SD2 and outlying study areas:

Revision: Original

8-4

3/3/2011

Enrollment Forecast Planning Book

Redistricting Modeling

High school-oriented conglomerate study areas:

Revision: Original

8-5

3/3/2011

Enrollment Forecast Planning Book

Redistricting Modeling

Middle school-oriented conglomerate study areas:

Revision: Original

8-6

3/3/2011

Enrollment Forecast Planning Book

Redistricting Modeling

Middle school-oriented conglomerate study areas zoomed:

Revision: Original

8-7

3/3/2011

Enrollment Forecast Planning Book

Redistricting Modeling

Elementary-oriented conglomerate study areas:

Revision: Original

8-8

3/3/2011

Enrollment Forecast Planning Book

Redistricting Modeling

Elementary-oriented conglomerate study areas with overlaid individual study area boundaries:

Revision: Original

8-9

3/3/2011

Enrollment Forecast Planning Book

Redistricting Modeling

Elementary-oriented conglomerate study areas with overlaid individual study area boundaries, elementary schools, and 2010 students:

Revision: Original

8-10

3/3/2011

Enrollment Forecast Planning Book

Redistricting Modeling

Even a cursory review of the middle school and elementary (individual) study areas easily exposes contiguity anomalies. Especially within the elementary study area definitions, numerous discontiguous areas exist even to the point of spanning across another schools area. Specific examples include: Arrowhead Beartooth Boulder Burlington Central Heights Lewis and Clark Miles Avenue Ponderosa Riverside

Once facilities have been properly analyzed and cataloged, with square footages, classroom counts and uses, deferred maintenance status and priorities, countless redistricting models can be generated to simulate facility loads and average student-facility distances. Study areas can be temporarily redefined to point students (of any level) to specific facilities, and they can further be redefined in their shape and size, too.

Revision: Original

8-11

3/3/2011

Enrollment Forecast Planning Book

Redistricting Modeling

This Page Intentionally Left Blank

Revision: Original

8-12

3/3/2011

Enrollment Forecast Planning Book

Appendix A: Enrollment Actuals Screen Shots

Appendix A: Enrollment Actuals Screen Shots


Early in the fourth quarter of 2009, SD2 CFO Harper requested that Dews repair an aging enrollment forecasting model previously built within a Microsoft Excel workbook. Dews analyzed the model and advised Harper that it utilized generalized linear factors and standard matriculation based on undocumented assumptions. In order to better understand the base numbers, and their trends, of historical SD2 enrollments upon which statistical forecasting models might be built and tuned Dews designed and crafted an Access database analysis/charting system (see the system primary screen below).

The left-to-right listboxes, at the top of the main form, are initially populated with SD2 subdistricts. One or more of these may be mouse or keyboard selected, populating the Types listbox with abbreviations of the types of schools existing within that/those subdistrict(s).

Revision: Original

A-1

3/3/2011

Enrollment Forecast Planning Book

Appendix A: Enrollment Actuals Screen Shots

Upon selecting (E)lementary, (M)iddle and/or (H)igh type(s), the Elementary, Middle and High Schools listboxes to the right will be appropriately populated in ascending sort order by school name. One or more of these may be selected, and the respective grades offered within that school(s) list will be offered for selection in the Grades listbox.

Subsequent selection of one or more grades will scope the school(s) and grade(s) range of data for display. A final scope choice is made by selecting one or more years of OPI audited actual enrollment numbers for said school(s) / grade(s) (either Fall October numbers, and/or Spring February numbers). The resulting final dataset is displayed in the lower half of the screen.

Revision: Original

A-2

3/3/2011

Enrollment Forecast Planning Book

Appendix A: Enrollment Actuals Screen Shots

Counts of the selected grade(s) are listed by year, by school. Data may then be charted automatically for trend analysis and presentation.

Selection across all years, for a given grade, shows such as Kindergarten (0), looks like this.

Trends for single or grouped grades for individual schools could easily be seen, documented, and initiate questions and discussion of cause, broader district effect, policy, etc. If Beartooth Elementary had a stop gap in Kindergarten enrollments from 2001 through 2006, we may select a broader range of grades to see if the whole school was closed.

Revision: Original

A-3

3/3/2011

Enrollment Forecast Planning Book

Appendix A: Enrollment Actuals Screen Shots

With comprehensive inclusion, we find that Beartooth Elementary was completely closed from 2001 through 2006. We may then look to surrounding schools, to see where the population from Beartooth uptrended their respective enrollments unnaturally.

Here we see exceptional inflation of Bitterroot and Eagle Cliffs counts for those exact years, but we cannot be sure of the extent of impact upon Alkali Creek, Bench or Sandstone populations.

Revision: Original

A-4

3/3/2011

Enrollment Forecast Planning Book

Appendix A: Enrollment Actuals Screen Shots

At this point, Dews and Walsh attempted to query and collect policy and/or decisional data which were employed by SD2 executives in closing Beartooth and routing which populations to which nearby schools. Such queries did not result in any useful responses. Other strategic changes could easily be detected by trend and anomaly identification, resulting in queries to SD2 executives for policy and decisional data.

Revision: Original

A-5

3/3/2011

Enrollment Forecast Planning Book

Appendix A: Enrollment Actuals Screen Shots

Revision: Original

A-6

3/3/2011

You might also like