Transcript Dr Cynthia F. Clark

Discussion, Q2010
Cynthia Clark
National Agricultural Statistics
Service
Alternative to Probability
Sampling






Comes from the discipline of applied
economics
Applies to real world problems and is
based on the intersection of:
- economic theory
- methods of analysis (econometrics)
- history/institutions/judgment
Describe the procedure used by the
National Agricultural Statistical Service
(NASS), a U.S. National Statistical Office
2
Real World
Problems
Methods of
Analysis
Applied
Economics
Economic
Theory
History
Institutions
Judgment
Definition of Applied Economics
3
NASS Statistical
Program




Produces forecasts and estimates of
agricultural production for crops and
livestock
Issues reports on agricultural
commodity prices & expenditures
Reports on farm numbers, employment,
income and land use
Provides data on costs of production
4
NASS Crop Forecasts
and Estimates







Surveys of farm operators
Weekly report on crop progress
Surveys of objective yield for crop using
measurements of the commodity
Administrative data from farm program
payments or other “final data”
Time series for yield of commodity
Weather and precipitation data
Days to frost
5
Estimation Process




Data collected and analyzed at state
level
Statisticians provide best estimate of
commodity for their state
Data sent encrypted to HQ office where
it is summarized
Board of internal experts review the
data in secure setting
6
Agricultural Statistics
Board





Review summarized data for survey –
list, area, & multi-frame estimates
Review objective yield models for
number, size, weight
Examine weather conditions
Review survey procedures
“Set” national & then state estimates
7
Release Procedures for
Market Sensitive Data





ASB “locked up” during review process
No internet, phone, or other
communication capabilities
Press come into locked site one hour
before announced release time
No political summary until one hour
after release
Data received credibly by agricultural
community
8
Concerns with Process




Not reproducible
Not transparent – not clear how
estimates were derived
No estimates of variability or bias
released with the estimates
Does not adhere to OMB Statistical
Standards Directive
9
Difficulties in NASS
Commodity Estimation



NASS survey program has designed
multiple estimates where some work for
some commodities; some for other
No criteria developed for estimate
selection from program
Non-surveyed variables are important in
estimation of production
10
Where do we go from here?es?



Continue as is
Release survey estimates with
coefficients of variation where possible
(or other measures of error)
Develop model based estimates
bringing in covariates
11
Planned Approach



Review program to determine where
survey estimates can be released
Examine measured sampling and
nonsampling errors associated with
those surveys & issue those relevant
Include response and coverage rates
and biases
12
Planned Approach
(continued)



Develop model based estimates where
covariates are needed
Provide advance notice of methodology
changes
Bridge new estimates with past board
estimates
13
Real World
Problems
Methods of
Analysis
Applied
Economics
Economic
Theory
History
Institutions
Judgment
Definition of Applied Economics
14
New Process





Move to estimation that acknowledges
use of statistical process
Based on surveys using random
representative samples
Manage survey processes to reduce
nonsampling errors
Measure relevant errors
Provide clear documentation of process
15
16