Review of Calculations of Common Level Ratios (CLR s) in Pennsylvania. Study conducted by: Alan S. Dornfest, AAS

Transcription

1 Review of Calculations of Common Level Ratios (CLR s) in Pennsylvania Study conducted by: Alan S. Dornfest, AAS June 6, 2014

2 Certification and Limiting Conditions The consultant (Alan S. Dornfest) certifies that, except as otherwise noted in this report: All work performed in conjunction with this report was conducted in compliance with the International Association of Assessing Officers Code of Ethics and Standards of Professional Conduct. The findings and recommendations were benchmarked against the Standards of the International Association of Assessing Officers (IAAO). Although these standards are voluntary and never prevail over state law, and are routinely subject to revision, they provide professional consensus of preferred practices of use within the assessing profession at this time. The consultant is a member in good standing of the IAAO and has no personal interest or bias with respect to the subject matter of this evaluation report or the parties involved. Employment or compensation of the consultant is not dependent on the reporting of a predetermined conclusion or the attainment of a stipulated result from the contents or conclusions of this report. To the best of the consultant s knowledge and belief, the statements of fact contained in this report, upon which the analysis, conclusions, and opinions expressed herein are based, are true and correct. The consultant assumes no responsibility for invalid data or misinformation provided. The consultant is not an attorney, and any references made to legal issues are for general guidance and are not intended as legal research of the issues or existing case law or code requirements. The consultant conducted discussions, obtained information, and prepared the analysis, conclusions, and opinions concerning the ratio study practices and policies of the Pennsylvania State Tax Equalization Board (STEB), with respect to the determination of the Common Level Ratio (CLR) that are set forth in this report. All conclusions expressed in this report are the sole responsibility of the consultant. Although data analysis was based on sales information provided to the consultant by the client, counties from which the data was obtained are not identified in this report, but are labeled as County 1, County 2, and County 3. This work was undertaken to assist the Pennsylvania Department of Community and Economic Development ( Department ) in its provision of providing administrative support to the STEB. The Department is in the process of considering revisions to STEB s information technology system and sought advice concerning the calculations and reports the system should be able to perform based on current IAAO standards and state law. All comments and recommendations, therefore, have been developed based on the Department s intended use of this work product, and are not appropriate for use in other contexts.

3 The State Tax Equalization Board completes several functions unrelated to the annual publication of each county s Common Level Ratio. Such functions are outside of the scope of this work and are not addressed within this report.

4 Table of Contents Executive Summary Introduction Scope of Work Analysis and Findings - Background Common Level Ratio (CLR) Methodology - Background Stratification for the CLR Use of the Common Level Ratio (CLR) Underlying Assessment System Acquisition of Sales Data Review of Sales Data by STEB Recommendations Regarding Sales Acquisition and Verification Adjustments to Sales Prices Use of Adjustments in Pennsylvania Recommendation Regarding Sale Price Adjustments Representativeness Representativeness Issues Affecting the CLR Commentary on Weighting Representativeness, Precision, and Sample Size Remedies for Inadequate Samples National Perspective Applicability in Pennsylvania Comments Outliers Outlier Effects and Trimming Procedures in Pennsylvania Comments on Outlier Trimming Procedures Limits on Trimmed Ratios Recommendations Regarding Representativeness and Outliers Statistics Use of Mean v. Median Statistics Distortions in the CLR Uniformity Statistics Reliability Statistics Recommendations Regarding Use of Statistics in Computing the CLR Training and Documentation Overall Conclusions Appendices Appendix A: References Appendix B: Transfer Tax Statement of Value Appendix C: Alternate Sales Verification Form (draft under development by PA assessors). 42 Appendix D: Examples Illustrating Level and Uniformity Statistics Based on Ratio Studies. 43 Appendix E: A Distribution-Free Method for Locating Outliers and Extreme Outliers Appendix F: Sample Ratio Studies Appendix G: Representativeness Studies Using County 2 and County 3 Data Appendix H: Statistical Analysis of Categories in Each County Appendix I: Contacts, Materials Reviewed, Assumptions, and Communications Appendix J: Ratio Study Precision Appendix K: Glossary Definitions of Common Ratio Study Related Terms... 78

5 List of Tables Table E1: STEB CLR Procedures Contrasted With IAAO Recommendations... 3 Table E2: Advantages and Disadvantages of Common Statistical Measures of Assessment Level... 4 Table E3: Priorities for Implementation of Recommendations... 8 Table 1: Use of Sale Price Adjustments by U.S. States in Table 2: Representative Samples Table 3: Representativeness of County 1 Sample Table 4: County 1 Ratio Study Statistics Given Various Outlier Trim Methodologies Table 5: Hypothetical Ratio Distribution Possible Excessive Trimming Table 6: Illustration of the Upward Bias in the Mean Table 7: State Use of Mean v. Median in

6 List of Figures Figure 1: Distribution of Untrimmed Ratios in County Figure 2: Distribution of Trimmed Ratios in County 1 Using STEB Methodology Figure 3: Distribution of Trimmed Ratios in County 1 Using IAAO IQR Methodology Figure 4: IAAO Recommended Measures of Assessment Level Figure 5: Sample Size Needed Given Varying Uniformity and Precision Figure 6: Degree of Precision (Confidence Interval Width) Given Varying Sample Sizes and Uniformity... 76

7 Executive Summary This report is a review of the system used by the Pennsylvania State Tax Equalization Board (STEB) to develop and calculate the Common Level Ratio (CLR) in Pennsylvania. The purpose of this review is to determine whether the CLR is calculated in a manner consistent with International Association of Assessing Officers (IAAO) standards and is the most representative indicator of assessment level in each county. Advantages and disadvantages of IAAO and STEB approaches are discussed, and recommendations are made when analysis indicates a pathway for achieving a more representative indicator. Specific recommendations, which are discussed in detail in the body of the report, are listed below. The recommendations are intended to provide a road map for changes that ultimately will facilitate the overriding goal of increasing representativeness of the CLR and assist the Pennsylvania Department of Community and Economic Development (DCED or the Department) in providing administrative support to STEB, in particular in the design of a new automated system to assist STEB with its work related to the annual publication of a Common Level Ratio. With this in mind, these recommendations should be considered, addressed deliberatively, and, to the extent possible and practical, implemented over time, involving stakeholders in the entire process. General Observations STEB is now housed within the DCED and has the responsibility of calculating a CLR for each county using statistically acceptable techniques. 1 STEB s current method of developing the CLR is described in a draft systems operation Policy and Procedures Manual (Manual) developed by an independent contractor with limited input from system users. 2 However, that Manual was never formally adopted, leaving in question many of the procedures actually used. STEB s development of the CLR is heavily reliant on sales data provided by counties. Decisions counties make regarding sales validation may be reviewed and challenged by STEB. There is not, however, any documented, standardized process for initiating such reviews. In addition, while there is some statutory guidance regarding what constitutes a valid, bona fide sale, usable in the ratio studies that are the statistical studies performed to determine the CLR, STEB has not expanded on or clarified this guidance, 3 except for one table in the Manual. In addition, the 1 Community and Economic Development Enhancement Act, Act of June 27, 1996, P.L. 403, No. 58, as amended by the Act of April 18, 2013, P.L. 4, No. 2, (b). It should be noted that the term statistically acceptable techniques is not defined in the statute. 2 CLR Policy and Procedures Manual. July 5, The Manual does not include many items often found in administrative policy manuals relied on by state agencies. It does not, for example, provide detailed timelines for routine submission of sales by counties, provide standard guidance on sales verification and sales validation, identify interactions to occur between STEB and county staff when questions arise about particular sales that have been submitted, identify criteria used by STEB to review submitted sales data, or address STEB procedures related to certain reviews required by statute Pa.Code

8 STEB is required to publish information annually that describes the procedures for excluding sales data and how these procedures compare with practices recommended by the IAAO. 4 While this is a recent statutory mandate, such a publication has not yet been provided. Ultimately, the counties are responsible for sales validation and submittal. The scope of this report is too limited to allow for the determination of the effectiveness of county processes along these lines. Although STEB purports to follow standard statistical procedures in doing the ratio study and determining the CLR, its procedures date largely from the early 1980s, and may have been closer to IAAO recommendations at that time, but have failed to recognize changes in more recent IAAO standards. 5 Current processes, described in the Manual, deviate from those recommended by the IAAO in several important areas as listed below and in Table E1 on the following page. There is no requirement that counties have written procedures for how they review and validate sales submitted to STEB or written internal quality assurance procedures regarding sales data. There is, moreover, no formal audit of sales information submitted and validity decisions made by counties. Further, the realty transfer tax statement of value document currently used provides limited information regarding whether a sale is an arm s length transaction generally usable in a ratio study. Outlier trimming procedures used do not comport with those suggested by IAAO standards and, most likely, result in insufficient trimming to produce representative samples. The CLR is based on a trimmed arithmetic (unweighted) mean, rather than the IAAO preferred median ratio. The CLR is based entirely on the sales sample, without regard to whether the frequency of sales in different property categories creates a sample that is representative of the frequency of properties in those categories countywide. This calls into question the representativeness of the CLR, the goal of which is to establish the overall typical level of assessment in the population of all properties in a county, not just in the sample of available sales. 6 Complete statistical ratio study information is not regularly published, and there is currently no state coordinated training program to promote ratio study understanding in general and, in particular, the elements that determine the CLR. Documentation of STEB procedures is limited, missing, and sometimes contradictory. Comparison Between CLR Procedure and IAAO Guidance Table E1 on the following page reflects a general comparison of key points regarding STEB s determination of the Common Level Ratio in comparison to IAAO standards for developing similar statistical measures. 4 Community and Economic Enhancement Act, Copy of note ascribed to STEB dated 1983, formally approved, and provided to author. 6 Note: Correction of this deficiency would require counties to submit and STEB to maintain, information on numbers of parcels in each category of property, irrespective of the number of sales submitted. 2

9 Table E1: STEB CLR Procedures Contrasted With IAAO Recommendations Issue STEB Procedure IAAO Recommendation (in contrast) Obtains from Acceptable to obtain from counties, but some sales counties; standard should be selected for audit, review, and additional review protocols verification. There should also be tracking mechanisms are not currently to ensure submittal of all sales, with clear iden- in place. tification of valid and invalid sales. Obtaining valid sales data for analysis Sales verification forms and process Identification and trimming of outlier ratios Statistics used to measure level of assessment Representativeness of sales sample Sample size State law does not currently require sales verification forms, and the transfer tax document that is available provides limited information. Ratios below 1% and above 500% eliminated; Mean of remaining ratios multiplied and divided by 4 to define range outside of which ratios are trimmed; Additional trimming provided for ratios above 200% Trimmed mean Not specifically determined; May be an issue when frequency of sales does not match proportions in underlying population of all properties One year of sales is used. Transfer tax document does not contain sufficient information to prove validity of sales as arm s length transactions, nor does it provide sufficient information to determine if an arm s length transaction meets other criteria for inclusion as a valid sale for purposes of a ratio study. For example, it does not provide information as to whether the property or its use has changed since the assessment and prior to the arm s length sale. Adoption of a more complete form is recommended. Advises against trimming based on high and low ratios, per se; Offers examples of procedures based on interquartile range (IQR) Median Weighting by proportion in population to avoid overrepresenting property types that have more sales. Proportions should be determined based on categories of property with sufficient sales for ratio study analysis. Provided these criteria are met, types can include residential, commercial, industrial, agricultural, and other property categories. Longer periods or added independent appraisals as proxies for sales are possible approaches that may be used if and when it is determined that one year of sales is an insufficient sample. 3

10 Documentation of procedures/ Presentation of findings The Manual was last updated in The Manual is a draft, not formally adopted. CLR is determined and reported; some uniformity measures are computed. Ratio study procedures should be thoroughly documented and up-to-date. If uniformity measures are to be computed, this should be done in accordance with procedures found in IAAO standards. Software applications should be documented so program logic can be reviewed and modified as needed. Section (d) of the Community and Economic Enhancement Act also requires STEB to review the CLR for a given county if it increases or decreases more than 10% from the previous year. Advantages and Disadvantages of Using Different Statistics to Determine the Level of Assessment In using a ratio study to analyze the level of assessment, the goal is to: measure how close appraisals are to market value on an overall basis. 7 In the context of the CLR, this measurement is with respect to assessed values (as opposed to appraisals). The most common statistics used to accomplish this determination are the: Mean, Weighted Median, Median, and Weighted Mean. Given the intended use of the ratio study in developing a CLR for each county, the weighted median is considered particularly appropriate. With regard to similar uses, the IAAO Standard on Ratio Studies comments: the generally preferred approach is to weight the median ratio of each stratum on the basis of the relative number of properties in the stratum. 8 The term stratum, as used in the IAAO Standard is analogous to category with respect to the CLR. Table E2 shows advantages and disadvantages of each of these statistics. Table E2: Advantages and Disadvantages of Common Statistical Measures of Assessment Level Statistic Advantages Disadvantages Mean Uses all data; basis for uniformity and reliability statistics; commonly understood Biased high; affected and distorted by outliers Weighted Median Incorporates the unbiased aspects of the sales sample median with the appropriate relative proportions of property in each category in the population As with the median, not as precise as mean 7 IAAO. Standard on Ratio Studies. Part 1, Section 9.1. April, IAAO. Standard on Ratio Studies. Part 2, Section 6.2. April,

11 Median Unbiased; not influenced by outliers Limited predictive ability; not as precise as mean Weighted Mean (price weighted) Reflects distribution of values in sample; ratios on high priced property have greatest influence Does not reflect distribution of ratios by parcel; may be distorted by extreme ratios on high priced property There is an inherent high bias in the mean and a tendency for this statistic to be distorted on the high side, largely due to the nature of ratio distributions, which tend to be skewed to the right (i.e. the high side). Although trimming can ameliorate this effect, the median, and more specifically, the weighted median, is the primary statistic recommended by the IAAO for the type of parcel-based (as opposed to value-based) assessment level determination intended for the CLR. Although the weighted mean is described in Table E2, it is not considered an appropriate measure because it does not reflect the level of assessment of typical parcels, the goal of the CLR 9. On the other hand, the weighted median captures the lack of bias in the median and provides a statistical measure that is more representative of the level of assessment in the underlying population of properties, not just the sales sample. List of Recommendations The following is a listing of recommendations found throughout the report. Explanatory detail about each recommendation is found in the section of the report where the recommendation is made. Recommendations are found in the order in which they occur in the body of this report and not the order in which they should be pursued or their significance. Table E3 suggests priorities for implementation of the recommendations and is in order of priority of implementation. For implementation, many of the recommendations will require DCED to develop a new automated IT system. The author understands that this is under consideration, but it is beyond the scope of this report to make specific IT technical system related recommendations other than to identify key data and statistics to be incorporated into the design of the system. Recommendation 1: The STEB should consider formally developing and adopting an administrative policy and procedures manual. Such a manual should include more concise guidance regarding sales verification and provide for review or audit of information submitted for accuracy and validity. The manual should also detail procedures for county and STEB staff interaction to address issues or questions identified as part of such reviews. Recommendation 2: To the extent resources permit, STEB should consider auditing sales records to ensure that all valid data have been submitted. Such an audit would require the ability for STEB to know when sales have occurred and review and, possibly, re-verify the validity of a subset of the submitted sales. The new IT system should be programmed to select a small random sample of sales for audit; unique sales identifiers would be required for the new system. Recommendation 3: STEB should cease using high and low ratios for sales rejection (except for obviously unusable sales, such as $1 sales although even these should be identified with a valida- 9 IAAO. Standard on Ratio Studies. Part 2, Sections 6.1 and 6.2. April,

12 tion code). STEB should adopt the IAAO trimming method. This would correct several problems, some of which have been identified in the report. STEB s current trimming method differs from the one initially adopted by the STEB Board, it is inconsistent with the IAAO s recommended method, effectively treats counties differently, and is not based on recognized applied statistics. Recommendation 4: All sales reports should include parcel identifiers to enable reconciliation between sales information obtained by the state from Recorders and sales submitted for the CLR. If there are additional valid sales for which no transfer tax document is required, these should be identified as well. Distinct property use and category codes should also be included to enable samples that better reflect the distribution of properties in the population to be developed. Provided there are sufficient sales, property use and category types can include residential, commercial, industrial, agricultural, and other property categories. Recommendation 5: The STEB should consider encouraging assessors to verify sales using a more complete document such as the form developed by AAP and modeled after the one recommended in the IAAO Standard on Ratio Studies. A sample document should be included in the Administrative Policy and Procedures Manual. The Assessors Association of Pennsylvania (AAP) created a sales verification form for use in Pennsylvania for all recorded sales and has provided a draft form to STEB. A copy of this form is attached to this report in Appendix C. This form was adapted from IAAO s sample sales verification form and tailored to Pennsylvania s needs. Recommendation 6: The Procedures Manual should include instructions advising assessors to adjust sale prices for personal property included with the sale, so the resulting ratio reflects the real property that is subject to assessment. Recommendation 7: Consider weighting measures of level for different property categories so that the overall CLR better reflects the distribution of properties in the population rather than just the sample. Recommendation 8: Review outlier identification and trimming procedures. Consider use of procedures based on the Interquartile Range (IQR) that tend to mitigate the effect of skewed ratio study distributions. Regardless of method chosen, report trimmed outliers and develop rules to limit trimming using IAAO guidelines that suggest no more than 10% - 20%. All outlier review procedures, including elimination or correction of $1 (and similar) sales, should be documented in the Policies and Procedures Manual. Recommendation 9: To move toward a system most consistent with IAAO principles, replace the unweighted, arithmetic mean, with the weighted median for determining the CLR. Recommendation 10: STEB should provide reports that include statistical information described in Section 6.0 of this report to counties for their review and use and to verify STEB s calculation of the CLR. 6

13 Recommendation 11: STEB should consider working with the Pennsylvania Assessors Association to provide general ratio study training and training specific to the development of the CLR. Recommendation 12: STEB should develop, formally adopt, and publish electronically a ratio study Policy and Procedures Manual that describes the entire CLR development process. This Manual should be developed collaboratively with input from the AAP and DCED. Prioritization of Recommendations Prioritizing recommendations is, to an extent, subjective. The hierarchy offered in the table on the following page combines two ranking mechanisms, need (i.e. high or medium) and timeline (i.e. long, medium, or short term). The recommendations found in this table are in order from highest to lowest priority. The highest priority has been assigned to areas that deviate significantly from IAAO standards and are likely to have a significant impact on the CLR. The timeline is based on the author s understanding of general process change implementation, not specific procedures or process requirements in Pennsylvania. In general, however, implementation of changes to calculation procedures should be more immediate than processes, such as training or sales validation audits, that might require additional resources as well as more major IT adjustments. A common theme throughout the recommendations and in establishing the priority for implementing recommendations is to cure, as expediently as practical, the distortions in the CLR that result from the use of non-standard methodologies in the calculation of this measurement. Hence, the importance assigned to the first five (listed) recommendations, which deal with: Getting good, valid sales data Trimming when warranted and using standard procedures to produce more representative samples and ratio study results Using statistical measures more appropriate for determination of the overall level of assessment (i.e. the median, instead of the mean, and a weighted median that reflects the relative proportions of properties in each category in the underlying population of properties) 7

14 Table E3: Priorities for Implementation of Recommendations Priority (timeline) Recommendation Number Description High (short term) 5 Adopt IAAO based sales verification form and procedures High (short term) 8 Adopt IAAO outlier trimming procedure High (short term) 3 Eliminate using high and low ratio based rejection procedure (see recommendation 8) High (short term) 9 Replace mean with median High (short term) 7 Weight measures of level (medians) by category of property High (medium term) High (medium term) Medium (medium term) Medium (medium term) Medium (long term) Medium (long term) Medium (long term) 1 Provide consistent guidance on sales verification 12 Develop, adopt, and publish procedures manual 6 Make adjustments to sales price for personal property included in the sale when appropriate 4 Include parcel identifiers, property use and category codes in sales reports 2 Provide procedures for audit of sales 10 Provide statistical reports 11 Provide or work with Assessors Association to provide training on ratio studies 8

15 General Conclusions The current procedures used by the STEB to develop a Common Level Ratio (CLR) for each county attempt to reach that goal and may have provided a close approximation of a statistically sound measure when developed in the 1980s. However, particularly given the evolution of the field of ratio studies, as evidenced by significant changes in the IAAO standards and other publications since that time, STEB procedures fail to reflect the current state of the art in this area. 9

16 1.0 Introduction Scope of Work The primary purpose of this project is to review the method used by the Pennsylvania State Tax Equalization Board (STEB) to calculate the Common Level Ratio (CLR) for each county in Pennsylvania and to identify possible revisions and denote their advantages and disadvantages. Any such revision suggestions are to be made consistent with the principle that one overall measure of assessment of assessment performance must be developed and must reflect all property types in a jurisdiction 10. Specifically, this report will address or provide the following: A comparison of the current CLR methodology with methods recommended by the International Association of Assessing Officers (IAAO) for identifying a county s overall level of assessment; Possible alternatives to the current CLR methodology which comply with the requirement to have a single measure that is representative of the bulk of individual assessments and the distribution of property types within a county; Procedures and recommendations concerning: o Sales validation o Outlier trimming o Confidence intervals o Sales sample size o Remedies for inadequate sample size o Stratification by value, property characteristics, classes, subtypes, use etc. Analysis of application of alternate suggested CLR procedures using actual county data. DCED contracted for this work as part of its role in providing administrative support to STEB. As DCED has recognized the need to revise STEB s 2011 IT system, it anticipates the report will provide information important to consider in the design of the new system. An important limiting consideration is related to available documentation of procedures used by STEB in conducting the ratio study used to determine the CLR. My analysis of these procedures and their comparison to IAAO standards is based on discussion with staff and excerpts from the Manual. I have not independently determined whether the software and operations in place actually follow these procedures. The author would like to thank all of those who contributed to my understanding of the use and development of the Pennsylvania ratio study and CLR methodology. Specific contacts, and materials received and reviewed are delineated in Appendix I. 10 See generally, Keebler v. the Board of Revision of Taxes of Philadelphia, 436 A.2d 583 (Pa. 1981). 10

17 2.0 Analysis and Findings - Background 2.1 Common Level Ratio (CLR) Methodology - Background The CLR is intended to enable users to consider whether their assessments are proportionally comparable to those of other property owners in each county. 11 The statistic is a measure of assessment level 12 calculated based on a trimmed mean from a ratio study using sales prices and assessed value information submitted for annual periods by each county to the STEB. The author s understanding of procedures used by STEB to calculate the CLR is based primarily on information contained in the Manual that has never formally been adopted. For comparing procedures in use with those recommended in IAAO standards, the author assumes that procedures found in the July, 2011 version of the Manual are being used. The validity of conclusions regarding the CLR methodology is incumbent on this assumption. Since this assumption is not testable, it is not possible to determine whether the determination of valid v. invalid sales is consistent from county to county. Regardless, the CLR is calculated as the arithmetic mean of the individual sales ratios for every valid sale received from the county. 14 The STEB modifies this result by defining and eliminating outliers based on ratios that exceed the original mean by more than 4 times the mean and ratios that are lower than the mean divided by The intent of this procedure is to include every valid sale with a ratio from 1% to 100%... In counties with a predetermined assessment ratio (EPR) 16 of 100%, valid sales with ratios up to 200% are to be used. 17 The Manual also indicates that ratios less than 1% or greater than 500% do not enter into the initial calculation of the pre-trim mean Stratification for the CLR CLR s are not developed for separate classes of sales (e.g., residential, commercial, etc.). Pennsylvania courts have interpreted the Uniformity Clause as requiring all property to be treated as a single class. Historically, this has led to the conclusion that there is one overall measure of assessment level in each county and not multiple measures that might otherwise reflect different levels of assessment in different property categories. 19 Although methodology will be discussed more under representativeness, it may be worth considering stratifying the sample in such a way as to better reflect the distribution (frequency) of property types in the underlying population. In other words, if 10% of the population of all properties in a county is commercial, in the- 11 Legislative Budget and Finance Committee. Pennsylvania s System for Property Valuation and Reassessment. P. S-11. July, Level The overall or typical ratio at which properties are appraised, as defined in IAAO. Property Appraisal and Assessment Administration. P Chicago, IL STEB letter of June 16, Ibid. 16 Established Predetermined Ratio, a percentage of value chosen by each county and applied to base year values to determine assessments in the county. See 53 Pa.C.S 8842(a). 17 STEB CLR-COD Explanation. April 7, CLR Policy and Procedures Manual. July 5, Deitch v. Board of Property Assessment, Appeals and Review of Allegheny County, 209 A.2d 397 (Pa. 1965). 11

18 ory 10% of the sample (i.e. ratios) should be from commercial sales. It appears that Pennsylvania case law would permit such an approach Use of the Common Level Ratio (CLR) CLR s are used only in appeals of assessed value where the CLR is applied to current market value to determine the assessment. In a recent case, Pennsylvania s Commonwealth Court stated that despite any inherent weaknesses, the CLR is acceptable calculation of the common level existing in the district and the standard against which the taxpayer s assessment ratio should be measured. 22 The purpose of this report is to compare and contrast the CLR with IAAO s recommended methods for identifying a county s level of assessment and identify the benefits and disadvantages of key differences. The goal is to make recommendations to improve Pennsylvania s calculation of the CLR within the parameters set forth in statute and case law by encouraging the use of statistically acceptable techniques. 2.4 Underlying Assessment System Pennsylvania is one of only eight states without either a current market value standard (as exists for example in Alaska) or a specified reappraisal cycle. 25 Instead there is a base year assessment system in place under which values are set as of a particular year, with the base year established independently by each county. 21 Keebler v. the Board of Revision of Taxes of Philadelphia, 436 A.2d 583 (Pa. 1981). 22 Appeal of Sullivan, 37 A.3d 1250 (Pa. Commw. Ct. 2012) citing Clifton v. Allegheny County, 600 Pa. 662, 693 (2009). 25 Dornfest, Alan S., Steve Van Sant, Rick Anderson, and Ronald Brown. State and Provincial Property Tax Policies and Administrative Practices (PTAPP): Compilation and Report. Journal of Property Tax Assessment & Administration. Volume 7, Number

19 3.0 Sales Data Sample Development Based on statute, a calendar year of valid sales constitutes the sample from which the CLR is calculated. For this measure to be meaningful, the sales must be representative of market value and of the properties in the underlying population of all parcels within any county. Several steps are involved to ensure such representativeness. These include: Acquisition of sales data, including establishment of validity of sales as arm s length transactions for which sales prices are indicative of market value and that the sales meet other requirements for inclusion in a ratio study (e.g., the property s characteristics and use have not changed from the time of the assessment to the arm s length sale, etc.). Review of sales data Adjustments to sales price (e.g., the sale price does not include personal property) to enable prices to reflect the market value of the real property being assessed Identification and review or elimination of outliers These steps are further elaborated in sections 3.1, 3.2, 3.4, and Acquisition of Sales Data The STEB acquires sales data from the counties. STEB provides the counties with a list of rejection codes. With respect to the rejection codes, those currently in use in Pennsylvania are substantively similar to those recommended by IAAO. 27 However, although foreclosure related sales are recognized and discussed in some of the sales related documentation, 28 there does not appear to be STEB guidance on the use of such sales, particularly and importantly use when they are a significant portion of the overall sales volume. IAAO recommends use when such sales constitute more than 20% of sales in a market area. 29 At least one county reported knowledge of this guidance and use of the sales. However, it is not clear to what extent this guidance is followed by all the counties in Pennsylvania when they submit sales to the STEB. Pennsylvania has statutory transfer tax provisions, and Pennsylvania assessors have access to the state realty transfer tax affidavit to assist in determining if a sale is valid. There is, however, no mandatory disclosure. While Pennsylvania assessors have access to the state realty transfer tax affidavit, the affidavit is not required if the sales consideration is stated as long as neither party claims an exemption from the realty transfer tax. Moreover, the state realty transfer tax affidavit (see Appendix B) does not have all of the information suggested by the IAAO to consider if a sale is an arm s length transaction and meets other IAAO requirements for inclusion in a ratio study 30. It should be noted that the transfer tax form is not designed under the auspices of STEB. 27 IAAO. Standard on Ratio Studies. Appendix A.4. April, See, for example, Assessors Association of Pennsylvania, Rejection Codes, 2012, approved by STEB. 29 IAAO. Standard on Ratio Studies. Appendix A.4. April, IAAO. Standard on Ratio Studies. Appendix H. April, IAAO Standard on Verification and Adjustment of Sales, Appendix A. November,

20 Regardless, among missing items are the following that are found in the IAAO demonstration forms, which ask for affirmation of or information about the following topics: Deed transfer in lieu of foreclosure or repossession Sale by judicial order Personal property included in the sale price How the property was marketed Length of time on the market Method of financing Explanation of unusual circumstances While not all of these items are crucial, especially in counties with large sample sizes, many of them are designed to help assessing officers understand whether the sale is, in all likelihood, an arm s length transaction in which the sale price is likely to reflect the market value of the underlying real estate. Some of this information could lead the assessor or the state to contact the buyer or seller for clarification and, ultimately, could result in sales data that is more likely to be representative of market value in a county. A draft form that would fill many of these gaps is being developed by the AAP and is found in Appendix C. Except for some of the foreclosure and judicial order information, this form appears to be modeled much more closely after the IAAO archetypes and includes the most critical information. The author has been informed that a more current modification includes information on short sales and would further approach the IAAO recommendation. STEB, as noted above, provides certain guidance on what constitutes a valid sale. There is, however, no consistent STEB-provided guidance regarding the sales verification process. As an example, sales data submitted as valid and reviewed as part of this report included sales with $1 prices and duplicate sales. It is not clear whether STEB review practices would have identified and eliminated such sales. No matter what statistical measures are used to calculate the CLR, this lack of guidance and STEB review of verification processes may be expected to diminish the validity of the CLR. 3.2 Review of Sales Data by STEB STEB staff indicate that they do review sales if there appears to be an issue. 33 The extent and exact reason for any such review or modifications to submitted sales has not been established and should be clarified by STEB. STEB staff also attempt to track the number of deeds recorded and utilizes this information as a reference point to ensure reporting of sales. Regardless, however, counties have the first opportunity to determine validity of sales. Although the lack of a formally adopted procedures manual lessens consistency, the draft Manual notes rejection codes that often reflect similar recommendations found in the IAAO literature. 34 It is notable that two of the rejection codes provided in the draft Manual are high ratio and low ratio. This appears to refer to ratios above 500% and below 1%. 35 With regard to outlier identification 33 From comments ed to the author by Don Welker on Feb. 28, STEB CLR Policy and Procedures Manual. Last Updated 7/5/ Ibid. p

21 and trimming procedures, an upper limit of 200% is applied by STEB 36. Additional information received from participants in this project indicates that the high and low ratio rejection codes are no longer valid. The exact review procedure undertaken by the STEB is unclear and needs to be documented and segregated from their outlier review process. Ultimately, all sales, should be reported to STEB along with codes indicating if a sale is valid or the reason it is not. Regarding outlier ratios, the IAAO Standard on Ratio Studies suggests using mathematical procedures for identification (discussed more thoroughly in Section 4.5 on outliers). In addition to the issues raised about sales verification, it is important for states that derive most of their ratio study sales data from local jurisdictions (as is the case in Pennsylvania), to independently verify or at least audit data that is submitted to ensure that sales have been adequately validated. The IAAO Standard on Ratio Studies contains the following statements on this issue: For sales-based studies, oversight agencies 37 should conduct an independent sales verification and screening program if resources permit. Alternatively, oversight agencies should develop audit criteria to review data submitted to qualify sales, corroborate representativeness and confirm adequate sample size. 38 The highest level of independence and objectivity requires independent sales validation. If resources are not available to achieve this level of sophistication, then a comprehensive audit program should be developed to review the validation and screening work of the local jurisdiction. 39 Fourteen states that derive their sales samples from local jurisdictions indicate that they also perform an audit of the sales validation process. 40 Although STEB does not function as an oversight agency, in that the purpose of the CLR study is not to alter or overrule county assessment decisions, the information required to perform the ratio study that is the basis for the CLR is identical to what would be required if the agency were acting in an oversight role. Hence, there is a similar need for independent sales verification and screening to ensure the validity of the CLR. 3.3 Recommendations Regarding Sales Acquisition and Verification Recommendation 1: The STEB should consider formally developing and adopting an administrative Policy and Procedures Manual. Such a manual should include more concise guidance regarding sales verification and provide for review or audit of information submitted for accuracy and validity. The manual should also detail procedures for county and STEB staff interaction to address issues or questions identified as part of such reviews. 36 Ibid. 37 STEB is not an oversight agency. 38 IAAO. Standard on Ratio Studies. Part 2, Section 3.3. April, Ibid. Section IAAO Technical Standards Committee. Ratio Study Practices in the United States and Canada: Results of 2011 Survey. Journal of Property Tax Assessment & Administration. Volume 9, Issue 1, p

22 Recommendation 2: To the extent resources permit, STEB should consider auditing sales records to ensure that all valid data have been submitted. Such an audit would require the ability for STEB to know when sales have occurred and review and, possibly, reverify the validity of a subset of the submitted sales. The new IT system should be programmed to select a small random sample of sales for audit; unique sales identifiers would be required for the new system. Recommendation 3: STEB should cease using high and low ratios for sales rejection (except for obviously unusable sales, such as $1 sales, although even these should be identified with a validation code). STEB should adopt the IAAO trimming method. This would correct several problems, some of which have been identified in the report. STEB s current trimming method differs from the one initially adopted by the STEB Board, it is inconsistent with the IAAO s recommended method, effectively treats counties differently, and is not based on recognized applied statistics. Recommendation 4: All sales reports should include parcel identifiers to enable reconciliation between sales information obtained by the state from Recorders and sales submitted for the CLR. If there are additional valid sales for which no transfer tax document is required, these should be identified as well. Distinct property use and category codes should also be included to enable samples that better reflect the distribution of properties in the population to be developed. Provided there are sufficient sales, property use and category types can include residential, commercial, industrial, agricultural, and other property categories. Recommendation 5: The STEB should consider encouraging assessors to verify sales using a more complete document such as the form developed by AAP and modeled after the one recommended in the IAAO Standard on Ratio Studies. A sample document should be included in the Administrative Policy and Procedures Manual. The Assessors Association of Pennsylvania (AAP) created a sales verification form for use in Pennsylvania for all recorded sales and has provided a draft form to STEB. A copy of this form is attached to this report in Appendix C. This form was adapted from IAAO s sample sales verification form and tailored to Pennsylvania s needs. 3.4 Adjustments to Sales Prices Regardless of the use of base year assessments or current market value assessments, the goal of the assessment process is to develop a value that reflects the value of the underlying property in most cases (and clearly in Pennsylvania where personal property is not taxable), the underlying real property value as of a specific date. 41 The IAAO Standard on Ratio Studies recommends consideration of four potential types of adjustments to sale prices to ensure that ratio studies properly measure assessment conditions (i.e. level and uniformity): Adjustments for financing, typically when financing is not at prevailing market rates, 41 Pennsylvania law permits counties to adopt a base year value or current market value. 53 Pa.C.S Base year is defined as The year upon which real property market values are based for the most recent countywide revision of assessment of real property or other prior year upon which the market value of all real property of the county is based for assessment purposes. Real property market values shall be equalized within the county and any changes by the board shall be expressed in terms of base-year values. 53 Pa.C.S

23 Adjustments for assumed leases when contract rent differs significantly from market rent, Adjustments for personal property included with a sale and reflected by the sale price, and, Adjustments for time. 42 Table 1 shows that adjustments are widely used by U.S. states. Table 1: Use of Sale Price Adjustments by U.S. States in Criteria/Description of Adjustment Number of states States doing annual ratio studies 43 Time adjustments 23 Financing adjustments 12 Personal property adjustments 27 Closing costs 8 Brokerage fees 4 Intangibles 14 Other Use of Adjustments in Pennsylvania Assessing Officers report making adjustments for personal property included with a sale if it is reported and its influence discernable. STEB reports no authority to make sale price adjustments as all STEB staff are not certified assessors. 44 Instructions provided in the state s Realty Transfer Tax Statement of Value explain that the amount to be reported is the amount: that the grantor received for the transfer of the real estate. 45 This implies that personal property transferred should not be included, but there is no specific place for reporting such personal property on the document itself (see Appendix B). This reinforces the need for the sales validation form Recommendation Regarding Sale Price Adjustments Recommendation 6: The Procedures Manual should include instructions advising assessors to adjust sale prices for personal property included with the sale, so the resulting ratio reflects the real property that is subject to assessment. 42 IAAO. Standard on Ratio Studies. Appendix A.5. April, IAAO Technical Standards Committee. Ratio Study Practices in the United States and Canada: Results of 2011 Survey. Journal of Property Tax Assessment & Administration. Volume 9, Issue 1, p IAAO Technical Standards Committee. Ratio Study Practices in the United States and Canada: Results of 2011 Survey. Journal of Property Tax Assessment & Administration. Volume 9, Issue 1, p Instructions for Completing Realty Transfer Tax Statement of Value. Provided to author. 17

24 4.0 Representativeness Representativeness of the sample used to determine the CLR is critical if the statistic generated is to be a valid indicator of the overall level of assessment, which is the goal. To achieve representativeness, it is necessary for types of property to appear with approximately the same relative frequency in both the sample and the population. 49 This principle is further elaborated in the IAAO Standard on Ratio Studies, which comments: A ratio study sample is considered sufficiently representative for direct equalization and mass appraisal performance evaluation when the distribution of ratios of properties in the sample reflects the distribution of ratios of properties in the population. A ratio study is considered sufficiently representative for indirect equalization when the distribution of ratios of dollars of property value in the samples reflects the distribution of ratios of dollars of property value in the population. 50 A ratio study that measures the level of assessment of the underlying parcels of property (whether in a class, category, or jurisdiction) provides a statistic 51 that may be used by some states to adjust assessments in the class, category, or jurisdiction to market value or another common base. Application of such an adjustment is deemed direct equalization. Although application is on a more limited basis (appealed properties) in Pennsylvania, the statistical principles are the same. This is not the case, however, with respect to indirect equalization, the goal of which is to establish a value weighted measure of level, which places greater weight on the assessment level (ratio) of high value properties. Such a process is appropriate when, for example, states seek to distribute monetary aid to schools on the basis of underlying property wealth (value) in each school district. More formal definitions are found in the Glossary in Appendix K. Twenty-six states indicate that they attempt to determine the representativeness of sales used in their ratio studies Representativeness Issues Affecting the CLR As has been noted in Section 4, the intended use of the Pennsylvania CLR is more akin to a measure of assessment level for direct equalization or mass appraisal performance evaluation than to indirect equalization, in which, usually, state funds are distributed on the basis of underlying market, rather than assessed, value. So, the goal should be to reflect the distribution of ratios in the population regardless of the value of a particular property. This principle is illustrated hypothetically in Table 2 on the following page. 49 IAAO. Property Appraisal and Assessment Administration. P Chicago, IL IAAO. Standard on Ratio Studies. Part 2, Section 4.2. April, Generally, the median is preferred see Figure 4, Section 5.1 of this report. 52 IAAO Technical Standards Committee. Ratio Study Practices in the United States and Canada: Results of 2011 Survey. Journal of Property Tax Assessment & Administration. Volume 9, Issue 1, p

25 Table 2: Representative Samples Property Type # of Parcels Relative Proportion Assessed Value Residential 89,000 89% $8.9 billion Commercial 9,000 9% $4.5 billion Agricultural 2,000 2% $1.0 billion Total 100, % $14.4 billion Proportion by value Level of Assesssessment CLR Using Parcel Weighting 61.8% 50% NA NA 31.3% 80% NA NA 6.9% 20% NA NA CLR Using Value Weighting 100% NA 52.1% 57.3% The analysis in Table 2 is designed to look solely at the effect of relative proportions of properties, using two methods of determining these proportions: numbers of properties and value of properties. The statistic chosen to measure the level of assessment is not in question in this demonstration. However, the goal of the CLR is to reflect the: ratio of assessed value to current market value used generally in the county. 53 The term, used generally appears to reflect proportional counts of properties rather than proportions of value, which might be skewed because of low or high assessments on a few high value properties. It is not contended that the current CLR uses value weighting. It properly does not. However, the relative proportions of properties in the population are not taken into account either. This can lead to distortions of the overall CLR, which, effectively, is weighted by proportions of different types of property in the sample, but not by those proportions in the population. The prospects for this distortion have been noted in previous studies of CLR methodology. In one such study, conducted in 2010 by the Legislative Budget and Finance Committee of the Pennsylvania General Assembly, the following note is provided in relation to this issue: In a representative sample, the types of properties included in STEB s calculations for the ratio studies would be in the same proportion as those property types occur in the total county property inventory. However, because some types of property are easier to sell in certain real estate markets, the sold properties are not necessarily representative of all of the actual property types in a county. 54 The effect can be demonstrated further when reviewing actual sample and population data, with which the author was supplied for several Pennsylvania counties. Table 3 (on the following page) compares population and sample proportions for County 1 and includes ratio study analysis for each stratum and for the county as a whole. Overall ratios correspond methodologically Pa.C.S Legislative Budget and Finance Committee. Pennsylvania s System for Property Valuation and Reassessment. P. 83. July,

26 to the CLR and employ the STEB trim for the entire sample, prior to stratification. Similarly, calculations for each category used the STEB trim method for comparability purposes. Table 3: Representativeness of County 1 Sample County 1 - Sample Representativeness Comparison to Population CLR - if adjusted to reflect pop. Proportion* CLR - if adjusted to reflect pop. Prop. based on median Number in Percent of Number in Percent of Mean Median Type of Property Population Population Sample Sample Ratio Ratio A - Residential w 10 acres % % C - Commercial % % L - Vacant land % % R - Residential % % T - Mobile Homes % 0 0.0% na V - vacant land 10+ acres % 1 0.3% na Subtotal of types with suff. sales % % na Total % % CLR CLR CLR CLR Note: "Clean and Green" parcels and tax exempt parcels excluded when identifiable. * adjustment based on application of each stratum mean ratio to population - excluding T and V types which had no or insuffient sales for anlaysis. Where the percent of sample is lower than the percent of population, the sample size needs to be higher. This is the case for example in the residential properties with 10 or more acres. The opposite is true for residential properties as a whole. Since the residential mean ratio is higher than most of the other strata mean ratios, it unduly influences the CLR, which is based on the overall analysis. If the stratum ratios were applied to the population proportions, rather than allowing the sample weighting to dictate the final result, the overall CLR, in this case, would be lower, but would be more reflective of the population mean. A similar result would be noted if the median, rather than the mean were to be used for the CLR. It should be noted that the CLR in Table 3 was calculated from data trimmed using the STEB method. Changes in trim procedures recommended later in this report would alter these results, but that is not relevant to the principle of representativeness being demonstrated in this table. Similar tables have been computed for County 2 and County 3, and are found in Appendix G, which also expands the calculations shown for County 2 to demonstrate the effect of combining the population based weighting system with trimming based on the interquartile range (IQR X 1.5). Detailed statistical analysis of each category for which sufficient sales were available is provided in Appendix H. In all cases, statistics have been computed from lists of valid sales after first excluding ratios less than 1% and greater than 500%. This again is for consistency with STEB procedures. In reality, most trim methods would identify such ratios as candidates for trimming and such extreme values probably mean that there are other issues, such as physical changes, improvements added after sales, etc. that essentially invalidate the sales. However, it is beyond the scope of this analysis to investigate sales validity coding to that extent. 20

27 4.2 Commentary on Weighting In developing the CLR, Pennsylvania courts have held that uniformity requirements do not permit determination of separate measures (i.e. CLR s or statistical measures of levels of assessment) for different property types. 55 If such stratification were allowed, the need for an overall measure and the incumbent weighting and representativeness problems would be obviated. However, the courts appear to have left an opening for stratification designed to produce one overall CLR, but avoid the computation of a common-level ratio which would unduly reflect the ratios of residential properties. 56 The IAAO Standard on Ratio Studies further comments on this problem as follows: Sales from areas or substrata in which the number of sales is disproportionately large can distort ratio study results by weighting level and uniformity indicators toward whatever conditions exist in the overrepresented area. To alleviate this problem and create better representativeness, large samples can be further stratified by Randomly selecting sales to be removed Isolating the overrepresented groups into substrata Redefining the time period for the overrepresented groups Weighting the data 57 In further commentary about this principle, specifically aimed at state agencies determining overall appraisal level, the IAAO Standard on Ratio Studies observes that: the generally preferred approach is to weight the median ratio of each stratum on the basis of the relative number of properties in the stratum. 58 Table 3 demonstrates a weighting regime that would be consistent with the principles espoused in the IAAO Standard on Ratio Studies and that would more closely approach the goal of having a CLR that is representative of the underlying population (i.e. inventory) of properties, without necessitating the creation of legally proscribed multiple CLR s within one county. Table 3 also shows the difference between the mean and median ratios as measure of assessment level. To modify the CLR to reflect relative proportions of properties in the population, the STEB would need to obtain information on county property inventories (i.e. populations) that it does not currently receive. This information would have to be annually updated and incorporated into the computer programs that generate ratio study statistics and the CLR. Weighting may also need to take into account the varying reliability of subcategory results caused by widely differing sample sizes. 55 Keebler Co. v. Board of Revision of Taxes of Philadelphia, 496 Pa. 140 (1981), citing Deitch Co. v. Board of Property Assessment, etc., 417 Pa 213 (1965). 56 Ibid. 57 IAAO. Standard on Ratio Studies. Part 1, Section 6.5. April, Ibid. Part 2, Section

28 4.3 Representativeness, Precision, and Sample Size It is extremely important for ratio studies to be based on adequate samples of valid indicators of market value. An adequately sized sample will have greater precision (i.e. reliability) and will be more likely to be representative of the population of properties for which the level of assessment is to be ascertained. In principle: the representativeness of a sample will increase with its size. As size increases, the characteristics of the sample converge with those of the population, so that statistics computed from the sample are more likely to be valid measures of corresponding population parameters. 59 Assuming, as in Pennsylvania, sales constitute the sample, this principle would be true provided the sales are valid, arm s length transactions for which the price of the real property has been correctly ascertained and provided that the sales are not unduly concentrated in certain areas or types of property whose appraisal levels differ from the general level of appraisal in the population. 60 The excessive concentration issue would occur, for example, if the proportional share of residential property in the sales sample exceeded the proportion of residential property in the population, provided the assessment level for residential property differed from that of other property types. This is demonstrated with respect to County 1 in Table 3. Currently, proportions of different property categories in the underlying population do not enter into the calculation of the CLR; only proportions in the sample affect the outcome of this calculation. Although this maximizes overall sample size, it fails to account for undue concentrations in the sales sample. The principle of proportional representation of the population by the sample is more critical than mere sample size. Many technical arguments can be made regarding sample size. However, although precision appears to grow given larger samples, this is not meaningful if the larger samples do not adequately provide a representative sample. There is also a common misconception that an adequate sample must include some definable percentage of the population. This implies larger samples given larger underlying populations. Although there is a slight relationship between sample size and population size the more important determinants of precision are the underlying variation in the population (in this case, the population of ratios predicted in the population) and desired degree of confidence. For example, it requires a larger sample to achieve a given degree of precision if the underlying population has a high standard deviation or coefficient of variation, than if there is good uniformity in the population. Conversely, small sample sizes produce less precision, regardless of the uniformity in the population. Regardless, except in extreme cases, where categories may need to be combined, smaller sample size is not justification for failing to develop samples representative of the population. Additional illustration of these principles is found in Appendix J. 59 IAAO. Property Appraisal and Assessment Administration. P Chicago, IL IAAO. Standard on Ratio Studies. Part 1, Section 4.5. April,

29 4.4 Remedies for Inadequate Samples National Perspective It is beyond the scope of this report to analyze the adequacy of samples used in Pennsylvania, in terms of sample size, precision, and underlying population variation. If samples are deemed inadequate in size, there are several methods by which they may be enlarged. The main ones recommended by IAAO and in use nationally are: Extending the period from which sales are drawn. The IAAO Standard on Ratio Studies comments that to develop an adequate sample size, the sales used in ratio studies can span a period of as long as five years provided there have been no significant economic shifts or changes to property characteristics and sales prices have been adjusted for time as necessary. 61 Validating previously rejected sales. Performing supplemental independent appraisals Applicability in Pennsylvania Re-validation of previously rejected sales could be suggested as part of guidance provided to counties in the Policy and Procedures Manual that needs to be developed (see Section 6) and should be considered when sample sizes produce unacceptable measures of precision or samples are deemed otherwise non-representative. Chapter 603 of the Pennsylvania Code appears to provide authority for collecting and considering independent appraisal data, subject to resource constraints. The Code makes specific reference to the use of appraisal samples of properties to supplement sales sample Comments Although such a remedy may not be applicable in Pennsylvania because of resource limitations and agency structure, it is notable that nineteen states indicate that they include both sales and appraisals in their ratio studies. 64 In addition to improving sample sizes, use of appraisals improves representativeness, since properties comprising segments of the population that rarely sell can still be appraised, and jurisdictions can be treated more equally in terms of controlling the degree of precision in the CLR. Disadvantages include cost and resource needs, and subjective elements innate to appraisals, as opposed to market derived sales IAAO. Standard on Ratio Studies. Part 1, Section 4.4. April, Ibid. Part 1, Section 6.4 and Part 2, Section Pa.Code IAAO Technical Standards Committee. Ratio Study Practices in the United States and Canada: Results of 2011 Survey. Journal of Property Tax Assessment & Administration. Volume 9, Issue 1, p IAAO. Standard on Ratio Studies. Part 2, Section 8. April,

30 4.5 Outliers Outliers are observations that have unusual values, that is, they differ markedly from a measure of central tendency. 66 Outliers may reflect data errors or undiscovered validity issues. Inclusion of outliers often results in distortion of ratio study statistics, particularly those statistics based on the mean. The IAAO Standard on Ratio Studies comments: The validity of ratio study statistics used to make inferences about population parameters could be compromised by the presence of outliers that distort the statistics computed from the sample. One extreme outlier can have a controlling influence over some statistical measures. However, some statistical measures, such as the median ratio, are resistant to the influence of outliers and trimming would not be required. 67 The recommendation for dealing with outliers is to: Identify Scrutinize to validate and correct errors Trim if necessary to improve representativeness Outlier Effects and Trimming Procedures in Pennsylvania The STEB has established procedures to identify and trim outliers. The procedures are as follows: 1. Eliminate all sales with ratios outside a range from 1% to 500% 69 as part of the validation process. 2. Determine the average (mean) ratio from the remaining sales. 3. Divide this mean by 4 to establish a lower trim limit. 4. Multiply the mean by 4 to establish an upper trim limit. 5. In addition to trimming based on the above parameters, there is an extreme upper limit of 200% (100% for counties with predetermined ratios less than 100%) IAAO. Glossary. P. 97. Chicago IAAO. Standard on Ratio Studies. Part 1, Section 5.2. April, Ibid. 69 STEB CLR Policy and Procedures Manual. Last Updated 7/5/11. P Note however, that some documentation shows an extreme upper limit of 200% regardless of predetermined ratio. See, for example STEB CLR Policy and Procedures Manual. Last Updated 7/5/11. P

31 4.5.2 Comments on Outlier Trimming Procedures The IAAO Standard on Ratio Studies cautions against trimming using pre-selected limits: trimming of outliers using arbitrary limits, for example, eliminating all ratios less than 50% or greater than 150%, tends to distort results and should not be employed. 71 Older IAAO literature accepted such trimming limits, but has been superseded by the current Standard. While the outlier identification and trimming procedures outlined in Section do not comport with those recommended in the current IAAO literature or standards, they are an obvious attempt to ameliorate the effects of outlier ratios that may be included in sales samples submitted by counties. Because the mean that is computed in step 2 (above in Section 4.5.1) may be based on samples that include ratios as high as 500%, that mean may be distorted, probably in such a manner that it would be much higher than if high ratios had not been included. That, in turn, reduces the effectiveness of range of the mean times 4 in identifying outliers. In other words, if the initially computed mean is too high, the range is too wide and fails to trim as many outliers. Distortion related to the 500% limit could be even worse given counties with low (below 100%) predetermined ratios, since it would take a more extreme outlier to reach that limit. In a test case, using sales data from County 1 (which was a county with a predetermined ratio of 50 percent of market value), and comparing the state trim method with one based on the interquartile range (IQR) and found in the IAAO Standard on Ratio Studies 72 and further demonstrated in Appendix E, the following results were noted as shown in Table 4: Table 4: County 1 Ratio Study Statistics Given Various Outlier Trim Methodologies Sample High Low Mean Standard Size Ratios Ratios Deviation Outlier Trim COV/ COD Median Method Trimmed Trimmed Ratio None 364 NA NA 19.21% 10.68% % STEB % 7.56% % Method % 4.16% % Interquartile Range (IQR X 1.5) In all of the analyses shown in Table 4, six sales with $1 sale prices were eliminated before calculating any statistics. Therefore, these $1 sales are not included in the number of ratios trimmed. Although, as expected, the median was highly resistant to the effect of outlier ratios, the mean was distorted on the high side originally. The STEB procedure did little to ameliorate this effect 71 IAAO. Standard on Ratio Studies. Appendix B.3 April, IAAO. Standard on Ratio Studies. Appendix B. April,

32 and the difference between the mean and median remained. In addition, the COV / COD ratio often indicates the existence of outliers when it exceeds , as it does following application of the STEB trim, but not following the application of the IAAO suggested trim method. Such high COV / COD ratios are often indicative of a small number of very high ratios, which may again be non-representative. 74 The implication is that, at least in the County 1 sample data, the STEB trim method is not sufficient to eliminate all outlier influence. Although one may question the use of an asymmetrical trim method, such as the one based on multiples of the interquartile range, there is an acknowledged tendency for the distribution of ratios to be skewed to the right. This occurs because ratios are bounded on the low end (i.e. ratios cannot be below zero), but not on the high end. It also occurs because of the mathematics of ratios. The IAAO textbook, Mass Appraisal of Real Property comments: An appraisal 20 percent above the sale price, for example, results in a ratio that differs from 1.00 more than does an appraisal 20 percent below the sale price: 1.20 versus As an example, note the quite typical skewed distribution of untrimmed ratios in the County 1 sales data base as shown in Figure 1. Figure 1: Distribution of Untrimmed Ratios in County 1 Given such a distribution, a trimming method based more on normally distributed data that assumes the mean is in the center of the distribution is likely to be less effective in producing a representative sample that reflects the underlying population. 73 IAAO. Standard on Ratio Studies. Appendix F. April, Ibid. 75 Gloudemans, Robert J. Mass Appraisal of Real Property. IAAO Chicago, IL. P

33 Note, for example, Figure 2, which shows the sample distribution resulting from application of the STEB trim methodology, which removes seven sales. Figure 2: Distribution of Trimmed Ratios in County 1 Using STEB Methodology The degree of skewness is almost unaffected by the STEB trim. Contrast this result with the IQR trim in Figure 3 shown below. Figure 3: Distribution of Trimmed Ratios in County 1 Using IAAO IQR Methodology 27

34 The outlier tails are considerably reduced using the IQR X 1.5 trim, the broadest trim recommended by the IAAO Standard on Ratio Studies. Complete statistical analyses for the trimming method can be found in Appendix F for trimmed and untrimmed samples from County 1, County 2, and County 3. In addition, Appendix G includes an example showing calculation of the CLR using population based category weighting, as described previously in Section 4.1 and Table 3, and the sample median determined after applying the IQR X 1.5 trim method. As of 2011, twelve states had adopted some version of the interquartile range trim method and one additional state indicated use of a fixed, asymmetrical trim range. 76 It should also be noted that the interquartile range trim method is considered distribution-free, meaning that it is applicable in the non-normal distributions common in ratio studies. 77 As a cautionary note, however, because the distribution of ratios is expected to be skewed to the high side, the interquartile range based trim method is expected to trim more high than low ratios. As in the analyses conducted for this study, this probably will lower the sample medians. Nevertheless, the distorting influence of high ratios will have been eliminated, and this is the fundamental reason that the trimming method is supported in the IAAO Standard on Ratio Studies. 78 The inherent goal of the ratio study sample is to replicate the underlying population of properties being tested. As has been noted previously, this occurs when the distribution of ratios in the sample matches the expected distribution in the population. This assertion is less likely true if high outlier ratios are left in the sample, since, in the population, properties with such high ratios are more likely to seek corrections, whereas there is no incentive to do so for properties with low ratios Limits on Trimmed Ratios By its nature, the STEB trimming procedure will not result in large numbers of sales being trimmed in most cases. However, as is true with other outlier identification methods, this may not be true when uniformity is very good and yet there are extended data tails, especially on the low end, since ratios over 200% are automatically excluded. Assume for example the hypothetical ratio distribution in Table 5 on the following page. 76 IAAO Technical Standards Committee. Ratio Study Practices in the United States and Canada: Results of 2011 Survey. Journal of Property Tax Assessment & Administration. Volume 9, Issue 1, p IAAO. Standard on Ratio Studies. Appendix B.1 (and embedded citations). April, Ibid. 28

35 Table 5: Hypothetical Ratio Distribution Possible Excessive Trimming Sales Ratios (%) In this table, the mean is 82%. Dividing by 4, any ratio below 20.5% would be considered an outlier and trimmed. This results in a trim of 20% of the data. The IAAO Standard on Ratio Studies recommends that no more than 10% of the sales be trimmed in small samples, although the Standard does allow 20% in the most extreme cases. 79 The Standard also notes that trimming of outliers using arbitrary limits, for example, eliminating all ratios less than 50 percent of greater than 150 percent, tends to distort results and should not be employed. 80 Similarly, the IAAO Standard advises against multiple trims for any one sample. 81 Regardless of the outlier identification method employed, it is considered appropriate to set maximum trimming limits. 82 STEB does not seem to have such limits in place. This issue is further exacerbated by the process through which STEB obtains sales samples. As has been mentioned elsewhere in this report (see Section 3.2), there is a risk that some trimming has already occurred before sales are submitted, in which case multiple trims, cautioned against in the IAAO Standard on Ratio Studies, may inadvertently be applied. 83 Finally, ratio reports should include lists of outliers that have been trimmed, as well as the basis chosen IAAO. Standard on Ratio Studies. Appendix B.4. April, IAAO. Standard on Ratio Studies. Appendix B.3. April, IAAO. Standard on Ratio Studies. Part 2, Section April, Ibid. 83 Ibid. Part 2, Section Ibid. Appendix B.6 29

36 4.6 Recommendations Regarding Representativeness and Outliers Recommendation 7: Consider weighting measures of level for different property categories (e.g. residential, commercial, etc.) so that the overall CLR better reflects the distribution of properties in the population rather than just the sample. Recommendation 8: Review outlier identification and trimming procedures. Consider use of procedures based on the Interquartile Range (IQR) that tend to mitigate the effect of skewed ratio study distributions. Regardless of method chosen, report trimmed outliers and develop rules to limit trimming using IAAO guidelines that suggest no more than 10% - 20%. All outlier review procedures, including elimination or correction of $1 (and similar) sales, should be documented in the Policies and Procedures Manual. 30

37 5.0 Statistics Once the sample is acquired, validated, subjected to appropriate adjustments, and reviewed by STEB, statistics need to be calculated to develop the most appropriate measure of assessment level for the CLR. Currently, the STEB uses the mean ratio following application of their trimming procedures. Statutorily, section (b) of the Community and Economic Enhancement Act requires the use of statistically acceptable techniques, including sales ratio studies. 85 However, the Act does not specify the statistic to be used. Notes provided indicate that there was discussion of this issue in 1983, at which time the mean was selected. Reasons appear to be related to simplicity. In 1983, the only guidance available from IAAO in the form of standards was the very first ratio study standard, published in There have been a number of subsequently published standards; however, even this earliest standard advised...reliance on the mean assessment ratio as a measure of central tendency is not recommended. 87 The statistic that has been overwhelmingly recommended to provide the most unbiased measure of assessment level is the median. This is especially true when the goal is to develop an overall measure of assessment level for a jurisdiction, except when indirect equalization of funding distributions is needed. In that case, the weighted mean is recommended as capturing the value related tendencies in the data; in other words weighting the ratios to reflect individual property value and, ultimately, permitting the determination of overall jurisdiction value. 5.1 Use of Mean v. Median Statistics The goal of the Pennsylvania ratio study under review in this report is to determine the CLR. As that statistic is to reflect the overall level of assessment for all parcels in a county, the best statistic is the median. The issue at hand is the presupposed bias inherent in the mean due in part to the lack of an upper bound and, in part, to the nature of ratios. Table 6 (on the following page) demonstrates this latter principle. Table 6: Illustration of the Upward Bias in the Mean 88 Assessed Value ($) (A) Market Value ($) (M) True Ratio (%) (A/M) Sale Price ($) (S) Sales Ratio (%) (A/S) 100, , , , , , , , , The true mean, based on market value, is 100% in Table 6. However, assessments that exceed sale price by the same amount as other assessments are below sale price do not contribute the same ratio effect. Hence, the ratio study sample mean is 100.7% and the sample mean is a bi- 85 Community and Economic Enhancement Act, (b). 87 IAAO. Standard on Assessment-Ratio Studies. Section September 28, IAAO. Property Appraisal and Assessment Administration. P Chicago, IL

38 ased estimator of the population mean, the estimation of which is the goal of the CLR and, indeed, every ratio study. Further examples of the effect of different sample ratio distributions on common measures of assessment level are found in Appendix D. Comparison of all level, uniformity, and reliability statistics, given various outlier trim methods, is shown in Appendix F for the three counties for which data was provided. Although most sources note the bias in the sample mean and its tendency to be higher than the true mean in the population, the extent of the bias is more difficult to quantify. Factors that influence the degree of bias include sampling error (i.e. statistical margin of error in the sample) and the presence of outliers. The former can and should be measured using confidence intervals. Outlier trimming is part of the approach taken by the STEB and has already been discussed. At least one study has found that: the unweighted mean appears to become less biased as the level of trim increases. 89 As a cautionary note, the trims used in the referenced study were very large, varying between 15% and 25% and, as is noted in the article, substantially exceed the level recommended in the IAAO Standard on Ratio Studies (IAAO 1990). 90 Given careful attention to the symmetry of the data distribution (and its normality), and application of recommended trimming procedures, the trimmed mean could be used in developing the CLR. If so, supportive documentation should demonstrate pictorially and statistically the absence of significant skewness, kurtosis, and non-normality in the trimmed data. In other words, a mean developed after application of acceptable trimming procedures (as discussed throughout Section 4.5 of this report), could be used provided the underlying data distribution is symmetrical and normal after trimming. Because of the additional analysis involved and remaining statistical questions, the use of the median (and, in applied form, the weighted median) is preferred (see discussion below). Examples of distributions and these statistics, using county data provided to the author, are shown in Appendix F. Alternatively, the STEB could adopt the median, the predominant statistic recommended in the literature, which is replete with commentary similar to the following: The median is less affected by extreme ratios than the other measures of central tendency. Because of these properties, the median is the generally preferred measure of central tendency for evaluating overall appraisal level. 91 The goal of the Pennsylvania ratio study is to determine an overall ratio applicable to all property in a county. To obtain representative results that provide one measure for combined strata, weighting, as discussed in Sections 4.1 and 4.2 of this report, should be considered. Nationally, frequency of use of the median v. the mean by states doing ratio studies is shown in Table 7 on the following page. 89 Denne, Robert C. The Mean, Weighted Mean, Median, and Other Robust Estimators of the Assessment Ratio: An Independent Exploration of Issues Raised by Nielsen. Property Tax Journal. Volume 12, No. 3. IAAO. September, Pp Ibid. p IAAO. Standard on Ratio Studies. Part 1, Section April,

39 Table 7: State Use of Mean v. Median in Survey Question # of States Using Mean # of States Using Median Do you calculate the measure? Do you use the measure for both indirect and direct equalization? Do you use the measure for direct equalization? 3 18 Although the questions asked do not comport exactly with the use of the ratio study in Pennsylvania, the statistic needed for the CLR is akin to the statistic needed for direct equalization, in which an adjustment to locally determined assessments is to be made based on the one number that best represents the level of assessment in either a category, class or property, or entire jurisdiction. As can be seen from this table, few states use the mean for this purpose. The IAAO Standard on Ratio Studies comments specifically on the applicability of the different measures of central tendency, specifically recommending use of category medians in developing an overall weighted median for the CLR: For purposes of oversight monitoring of overall appraisal performance and direct equalization, the generally preferred approach is to weight the median ratio of each stratum on the basis of the relative number of properties in the stratum. 93 As has been noted elsewhere in this report, even though the STEB does not function as an oversight agency, the principle of weighting results from each strata to achieve a more representative measure of the overall level of assessment is no different. Finally, it is notable that, as shown below in Figure 4, the IAAO Standard on Ratio Studies does not recommend the use of the mean, despite acknowledging that the trimmed mean may be acceptable. Figure 4: IAAO Recommended Measures of Assessment Level 94 Table 2-2 Preferred Estimators Indirect Direct Monitoring Equalization Equalization Performance Median X X Mean Weighted Mean X* * Caution should be exercised when the sample contains value outliers or indicates value bias based on the PRD 92 IAAO Technical Standards Committee. Ratio Study Practices in the United States and Canada: Results of 2011 Survey. Journal of Property Tax Assessment & Administration. Volume 9, Issue 1, p IAAO. Standard on Ratio Studies. Part 2, Section 6.2 April, Ibid. P

40 5.2 Distortions in the CLR Regardless of the trimming procedure, especially in small samples, the CLR is more likely to suffer distortion since it is based on the mean, than would be the case if it were based on the median (and associated weighted median to properly reflect the underlying population). Appendix D provides hypothetical examples showing ratio study level and uniformity calculations based on small samples. Such samples illustrate more obviously the distortion that can affect certain statistics, particularly the mean. 5.3 Uniformity Statistics The IAAO Standard on Ratio Studies identifies several other statistics that are useful to consider in ratio studies, including measures of variability such as the coefficient of dispersion (COD) and the price-related differential (PRD). The statute does not authorize STEB to consider and publish statistical measures like the COD that characterize the amount of dispersion around a measure of central tendency such as a median. IAAO standards related to such measures of dispersion are based on separate property strata. For example, there is one standard for residential improved property in a large, densely populated active market and another for residential property in a small rural area with a depressed market. Similarly, the IAAO has one standard for income-producing properties in densely populated active markets, and another for rural areas with older properties in a depressed market. As the courts do not allow for separate CLR s for different property strata, such statistical measures are not meaningful for considering the CLR as currently calculated. However, uniformity measures would be useful in reviewing strata results if measures of central tendency from strata were to be used to develop a population weighted CLR. Regardless of current use, both the COD, which measures general or horizontal equity, and the PRD, which measures value related or vertical equity, should be calculated as part of the ratio study, especially since the amount of dispersion affects the reliability of the measures of central tendency. It should be noted that draft STEB procedures for the COD base the relative amount of dispersion on the mean, rather than the IAAO preferred median. This may have a historic basis since early (1978) guidance in an IAAO reference book indicated that the COD may be calculated with respect to any of the three measures of central tendency. 95 By 1990, this guidance had changed to the admonition that although the COD could be calculated about the mean, this is not recommended because the mean is more affected by extreme ratios than the median and because of the inherent bias of the mean. 96 Standards and textbooks published by IAAO following the 1990 Standard on Ratio Studies have universally stated that the COD is to be calculated with respect to the median, not the mean. 5.4 Reliability Statistics Regardless of which measure of central tendency (assessment level) ultimately is chosen, it is necessary to compute the statistical sampling error associated with any such measure derived 95 IAAO. Improving Real Property Assessment; A Reference Manual P IAAO. Standard on Ratio Studies. Section July,

41 from a sample constituting a subset of the population of properties intended to be represented. By understanding the amount of sampling error, conclusions can be drawn about the degree of precision in statistics and adequacy of samples. This will enable better identification of sampling weaknesses due to high variability and small sample sizes. Understanding the precision of the statistics will become more important if the CLR is to be calculated based on the weighted median, since the development of a median ratio for each property category necessitates use of somewhat smaller sample sizes. The IAAO Standard on Ratio Studies suggests combining strata (categories) when there are unacceptably large margins of error. 98 Confidence intervals can be calculated around the median for each category to enable such conclusions to be drawn. 5.5 Recommendations Regarding Use of Statistics in Computing the CLR Recommendation 9: To move toward a system most consistent with IAAO principles, replace the unweighted, arithmetic mean, with the weighted median for determining the CLR. 98 IAAO. Standard on Ratio Studies. Part 2, Section April,

42 6.0 Training and Documentation A crucial part of any ratio study based process is providing documentation complete enough for the user to understand the steps undertaken to produce the results. Many states, including Kansas, Idaho, and Texas, produce lengthy manuals to assist assessors by enhancing their understanding of the process. Such explanative documentation leads to more consistent ratio study development, especially with regard to areas requiring cooperation and participation. The main documentation provided (or discoverable online) for this study with regard to the CLR is the Manual. This document was never formally adopted, but the version referred to in this report is dated 7/5/2011 and contains much descriptive information about the purpose and calculation of the CLR. The author notes that the Manual is considered to be a draft developed without direct input from county assessors or direct input from key STEB staff. IAAO supports a more collaborative approach so that ratio studies are produced in a manner that is transparent in all stages to all stakeholders. 104 Finally, two separate documents, one entitled CLR-COD Explanation and the other entitled CLR, refer respectively to use of every valid sale with a ratio from 1% to 100%... or every valid sale with a ratio from 1% to 500%... to compute a mean used as the basis for trimming limits The existence of two documents with conflicting procedures lends doubt on the actual procedures being employed. Complete documentation should be available and should include discussion of procedures for: Quality assurance Sales verification Sales tracking Communication with STEB Data modification Reviews and audits CLR s can be found for each county for several years through 2012 on the STEB/Tax Equalization Division (TED) webpage located on DCED s website: April 24, 2014 (accessed May 2, 2014). Regardless of the more limited use of the STEB ratio study, the IAAO Standard on Ratio Studies suggests that all reports include the following information: Date and tax year of the appraisals being evaluated Number of parcels in each stratum Number of sales Number of sales trimmed from the study Measures of central tendency (appraisal level) Measures of uniformity (variability) and price-related biases Confidence interval (measures of reliability) about the measures of central tendency Summary of adjustments made to sales prices IAAO. Standard on Ratio Studies. Part 2, Section 10. April, STEB. CLR-COD Explanation. Last updated 4/7/2011 and CLR Policy and Procedures Manual. Last updated 7/5/ IAAO. Standard on Ratio Studies. Part 1, Section 8.2. April,

43 Similarly, IAAO promotes training and education on ratio studies. The organization offers, for example, a workshop entitled Fundamentals of Assessment Ratio Studies (FARS) described as follows: Workshop Fundamentals of Assessment Ratio Studies This workshop provides a very basic introduction to the development and uses of assessment sales ratio studies based upon the IAAO Standards on Ratio Studies: January, The workshop covers the topics of sales analysis, sampling, and the development of assessment ratio studies. The workshop also covers managerial and technical issues, which include staffing, timing of studies, use of study results, statistical testing and assessment regressivity and progressivity. This workshop includes an exam. Recommended: Course 101. IAAO recertification credit: 18.5 hours-2 ½ days 107 While a general course, such as the IAAO workshop, can provide a good basis for enabling assessors to understand ratio study fundamentals, it will not teach specifics with respect to the Pennsylvania ratio study and its use in developing the CLR. While general ratio study training may be beyond the scope of STEB functions, STEB could consider explanatory training to improve understanding of the procedures used to develop the CLR. Such training could emphasize the need for consistent sales validation processes, for example. Recommendation 10: STEB should provide reports that include statistical information described in Section 6.0 of this report to counties for their review and use and to verify STEB s calculation of the CLR. Recommendation 11: STEB should consider working with the Pennsylvania Assessors Association to provide general ratio study training and training specific to the development of the CLR. Recommendation 12: STEB should develop, formally adopt, and publish electronically a ratio study Policy and Procedures Manual that describes the entire CLR development process. This Manual should be developed collaboratively with input from the AAP and DCED last accessed March 22,

44 7.0 Overall Conclusions The current procedures used by the STEB to develop a Common Level Ratio (CLR) for each county attempt to reach that goal and may have provided a close approximation of a statistically sound measure when developed in the 1980s. However, particularly given the evolution of the field of ratio studies, as evidenced by significant changes in the IAAO standards and other publications since that time, STEB procedures fail to reflect the current state of the art in this area. Many of the errors in the CLR may be on the high side, particularly due to use of a mean trimmed in accordance with non-standard procedures. Such a high bias in the CLR would promote less equity resulting from appeals, since the appellant would have a higher hurdle to overcome. The recommendations in this report are intended to provide a road map by which the STEB can alter its procedures so they are more in line with IAAO standards. These recommendations are not intended to redefine the outcome, only to provide a more supportable underlying process. Broadly, recommendations fall into the following areas: Sales Validation adopting documentation to provide more comprehensive guidance and training to counties; adopting a more comprehensive sales validation form; instituting more of a review process Representativeness taking steps to ensure that at least broad types of properties in samples occur with similar frequency to that in the population or weighting the results according to frequencies in the population Trimming adopting an outlier trim mechanism more based on IAAO standards Measures of Central Tendency replacing the mean with the median Measures of Reliability computing confidence intervals and related statistics and considering ways to use these statistics to better identify and improve sampling weaknesses due to high variability and small sample sizes Documentation and Training officially adopting and expanding current documentation to include published results of complete ratio studies and to more completely explain all aspects of the ratio study; working with county assessors to develop a training program to enhance general ratio study knowledge and understanding of the CLR 38

45 Appendices 39

46 Appendix A: References Almy, R.R., R.J. Gloudemans, R.C. Denne, and S.W. Miller Improving real property assessment: A reference manual. Chicago: International Association of Assessing Officers. Denne, Robert C. The Mean, Weighted Mean, Median, and Other Robust Estimators of the Assessment Ratio: An Independent Exploration of Issues Raised by Nielsen. Property Tax Journal. Volume 12, No. 3. IAAO. September, Dornfest, Alan S., Steve Van Sant, Rick Anderson, and Ronald Brown. State and Provincial Property Tax Policies and Administrative Practices (PTAPP): Compilation and Report. Journal of Property Tax Assessment & Administration. Volume 7, Number Eckert, J.K., R.J. Gloudemans, and R.R. Almy, eds Property appraisal and assessment administration. Chicago: International Association of Assessing Officers. Gloudemans, R.J Mass appraisal of real property. Chicago: International Association of Assessing Officers. International Association of Assessing Officers (IAAO) Standard on ratio studies. Kansas City, MO: IAAO d. Standard on verification and adjustment of sales. Kansas City, MO: IAAO., Technical Standards Committee Ratio study practices in the United States and Canada: Results of the 2011 survey. Journal of Property Tax Assessment & Administration 9 (1): Glossary for property appraisal and assessment Kansas City, MO: IAAO. Improving Real Property Assessment; A Reference Manual

47 Appendix B: Transfer Tax Statement of Value 41