1 N ational A gricultural S tatistics S ervice Washington, D.C. Practical Considerations in Selecting Statistical Disclosure Methodology for Tabular Data BTS Confidentiality Seminar Series, March 2003

2 Examples in the Presentation are Based on Actual Occurrences N ational A gricultural S tatistics S ervice Washington, D.C.

3 Data have been changed to protect the innocent N ational A gricultural S tatistics S ervice Washington, D.C.

4 Goals: Discuss Decision Factors Discuss Alternatives Present Illustrations Provide a Perspective for Future Decisions N ational A gricultural S tatistics S ervice Washington, D.C.

5 DECISION FACTORS - 1 What is your data structure? Household, establishment, opinion, census data, sample results, etc… N ational A gricultural S tatistics S ervice Washington, D.C.

6 DECISION FACTORS - 2 N ational A gricultural S tatistics S ervice Washington, D.C. What are your population characteristics? Limited variation, extremely clustered, highly skewed, etc…

7 DECISION FACTORS - 3 N ational A gricultural S tatistics S ervice Washington, D.C. Who are your customers? Casual readers In-depth analysts People (companies) from the data population Etc…

8 DECISION FACTORS - 4 N ational A gricultural S tatistics S ervice Washington, D.C. What decisions will be made by your customers? Social Research Studies Economic Investment Decisions Implementation of Government Programs Etc…

9 BASIC SUPPRESSION APPROACHES (n, p) Rules (p, k) Rules (n, p, t) Rules N ational A gricultural S tatistics S ervice Washington, D.C.

10 Should You Publish Your Suppression Rule? Conventional wisdom says NO! It might be essential for certain situations N ational A gricultural S tatistics S ervice Washington, D.C.

11 BASIC NASS APPROACHES (p, k) Rule for Ag Census (n, p) Rule for Periodic Surveys Special Techniques N ational A gricultural S tatistics S ervice Washington, D.C.

12 SPECIAL TECHNIQUES Signed Release Agreements (Waivers) Geographic Combinations Size Group Combinations Reclassification of Categories N ational A gricultural S tatistics S ervice Washington, D.C.

13 WAIVERS Large Operations Permission to Publish is Requested Must Agree in Writing Updated Annually N ational A gricultural S tatistics S ervice Washington, D.C.

14 Waiver Example Hypothetical Pecan Production Example 1/ StateProduction – 1,000lbs AL15,000 GA80,000 LA18,000 NM35,000 TX30,000 All Other 2/31,850 US209,850 1/ Assumes One person controls all Arizona Production 2/ Includes AZ, AR, CA, FL, KS, MS, NC, OK, SC

15 Waiver Example Pecan Production Example, Actual Results State2000 Production/1,000lbs AL15,000 AZ14,500 AR900 CA3,400 FL3,300 GA80,000 KS550 LA18,000 MS3,500 NM35,000 NC1,600 OK2,500 SC1,600 TX30,000 US209,850

16 Geographic Combinations N ational A gricultural S tatistics S ervice Washington, D.C. Hypothetical Olive Acreage StatesGrowersAcres CA1,36337,714 FL22 U.S. Total1,36537,716 Traditional Approach CA1,363(D) FL2(D) U.S. Total1,365(D) (D) Withheld to avoid disclosing data for individual farms.

17 Geographic Combinations N ational A gricultural S tatistics S ervice Washington, D.C. Handling hypothetical Olive data Situation by moving Florida Olives to “Other” Olive Acreage StatesGrowersAcres CA136337,714 U.S. Total136337,714

18 Size Group Example Percent of Farms, by Economic Sales Class, U.S Economic Sales ClassPercent of Total Farms $1,000 - $2, $2,500 - $4, $5,000 - $9, $10,000 - $24, $25,000 - $49, $50,000 - $99, $100,000 - $249, $250,000 - $499, $500,000 - $999, $1,000, Total100.0

19 Size Group Example Number of Farms: Economic Sales Class by State and Region, 2001 State and Region Economic Sales Class $1,000 - $9,999$10,000 - $99,999$100,000 & Over Number Northeast CT2,450 ME4,200 MA3,200 NH2,100 NJ6,400 NY17,80011,0008,700 PA32,80016,5009,700 RI350 VT3,500 Other States10,6003,800 Total72,80038,10022,200

20 Reclassification Example Hypothetical Potato Variety Data, 1 State Fall Potatoes: Percent of Major Varieties Planted, State and VarietiesPercent of Planted Acres R Burbank71.0 Ranger R12.0 R Norkotah7.5 Shepody3.4 Silverton R2.4 De Wolf1.0 Other2.7 TOTAL100.0 However, assume that deWolf is a new variety planted only on one large farms

21 Reclassification Example Actual Potato Variety Data, 1 State Fall Potatoes: Percent of Major Varieties Planted, State and VarietiesPercent of Planted Acres R Burbank71.0 Ranger R12.0 R Norkotah7.5 Shepody3.4 Silverton R2.4 Other3.7 TOTAL100.0

22 CASE STUDY Agricultural Marketing Service implementation of mandatory price reporting law illustrates basic decision factor principles. N ational A gricultural S tatistics S ervice Washington, D.C.

23 OUTLINE Define the problem Explain the approach Describe the solution Present some results N ational A gricultural S tatistics S ervice Washington, D.C.

24 Large meat packers must report details of all purchase transactions. N ational A gricultural S tatistics S ervice Washington, D.C.

25 Data must be reported by fixed times each day. AND Summaries published an hour later. N ational A gricultural S tatistics S ervice Washington, D.C.

26 AMS must protect confidentiality in issuing summaries. N ational A gricultural S tatistics S ervice Washington, D.C.

27 90% of slaughter is in 114 plants, owned by 65 companies. N ational A gricultural S tatistics S ervice Washington, D.C.

28 4 Companies: 80% of fed cattle 80% of fed lambs 55% of all hogs N ational A gricultural S tatistics S ervice Washington, D.C.

29 AMS adopted a 3/60 standard for each data cell. N ational A gricultural S tatistics S ervice Washington, D.C.

30 At least three companies and no company exceeding 60%. N ational A gricultural S tatistics S ervice Washington, D.C.

31 AMS never identified how many companies were in an aggregate. N ational A gricultural S tatistics S ervice Washington, D.C.

32 AMS did not distinguish blank cells from suppressed cells. N ational A gricultural S tatistics S ervice Washington, D.C.

33 Many (24 %) daily reports were suppressed plus many cells in released reports. N ational A gricultural S tatistics S ervice Washington, D.C.

34 Some observers calculated 90% suppression. N ational A gricultural S tatistics S ervice Washington, D.C.

35 When in Doubt – Examine the Data! N ational A gricultural S tatistics S ervice Washington, D.C.

36 Plant AXXXXXXX BX CXXXXXXXXXXX DXXXXXXX EXXX FXXXXXXXXXXX GXXXXXXX HXXXXXXXXXXX IX SEQUENCE DAY N ational A gricultural S tatistics S ervice Washington, D.C.

37 When are proprietary data not unique? N ational A gricultural S tatistics S ervice Washington, D.C.

38 N ational A gricultural S tatistics S ervice Washington, D.C. A new 3/70/20 confidentiality standard based on 60-days of data.

39 N ational A gricultural S tatistics S ervice Washington, D.C. At least 3 companies operated 50% of the time.

40 N ational A gricultural S tatistics S ervice Washington, D.C. No company had more than 70 percent of the volume.

41 N ational A gricultural S tatistics S ervice Washington, D.C. No company would be exposed more than 20% of the time.

42 Nearly all cattle daily reports and all cells now released. N ational A gricultural S tatistics S ervice Washington, D.C.

43 EXAMPLE TALLY Cattle Open Market, One Region, Daily N ational A gricultural S tatistics S ervice Washington, D.C. 3 or more68% Concen.47% Exposure2%

44 EXAMPLE TALLY Cattle Forward Contracts, National, Daily N ational A gricultural S tatistics S ervice Washington, D.C. 3 or more41% Concen.51% Exposure6%

45 Modification 1 N ational A gricultural S tatistics S ervice Washington, D.C. Cattle forward contracts are now published weekly

46 Modification 2 N ational A gricultural S tatistics S ervice Washington, D.C. All swine packer sales alternatives weighted together daily.

47 Modification 3 N ational A gricultural S tatistics S ervice Washington, D.C. Lamb purchases accumulated until the 3/70/20 is reached.

48 Author Contact Information Rich Allen Deputy Administrator, NASS Room 5905 South Building 1400 Independence Ave., S.W. Washington, D.C Fax N ational A gricultural S tatistics S ervice Washington, D.C.