understanding the statistical model - futurework systems...statistical model in the negotiations...
TRANSCRIPT
Understanding the
Statistical Model
Presented by the U.S. Department of Labor
Employment and Training Administration
Office of Policy Development and Research
And Its Use in Negotiations
Today’s Objectives
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 2
Understanding the Statistical Adjustment Model
The WIOA statute requires the use of a Statistical Adjustment Model in
establishing performance goals for core programs.
This workshop will help you learn about:
The design of the statistical regression model
Its use in establishing performance metrics
Its application at the national level
Its practical use at the state level.
Today’s Agenda
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 3
Overview of the requirement to use a Statistical Model
Discuss the phased-in approach to the Model and baseline indicators
Present the elements of the Model and the variables used
Explore the process of using the model to Negotiate Levels of
Performance
Explore the Model’s role in Determination of Performance Failure
Explore next steps in the development and use of the Model
Quick Poll
3/23/2017 U.S. Department of Labor, Employment and Training Administration 4
What is your level of involvement with the Statistical
Adjustment Model (SAM) and negotiations?
a. Integral
b. Minimal
c. Huh, who is SAM?
The Regulations
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 5
Workforce Innovation and Opportunity Act; Joint Rule for Unified and
Combined State Plans, Performance Accountability, and the One-
Stop System Joint Provisions
https://www.gpo.gov/fdsys/pkg/FR-2016-08-19/pdf/2016-15977.pdf
Guidance
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 6
TEGL 26-15 Negotiating Performance Goals for the Workforce
Innovation and Opportunity Act (WIOA) Title I Programs and the
Wagner-Peyser Employment Service as amended by Title III, for
Program Year (PY) 2016 and PY 2017
https://wdr.doleta.gov/directives/corr_doc.cfm?DOCN=9363
TEGL 10-16 Performance Accountability Guidance for Workforce
Innovation and Opportunity Act (WIOA) Title I, Title II, Title III and Title IV
Core Programs
https://wdr.doleta.gov/directives/corr_doc.cfm?DOCN=8226
Today’s Agenda
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 7
Overview of the requirement to use a Statistical Model
Discuss the phased-in approach to the Model and baseline indicators
Present the elements of the Model and the variables used
Explore the process of using the model to Negotiate Levels of
Performance
Explore the Model’s role in Determination of Performance Failure
Explore next steps in the development and use of the Model
Overview of
Requirements
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 8
WIOA 116(b)(3)(A) Performance Accountability requires the use of the
statistical model in the negotiations process: (v) Consideration of four factors when negotiating levels of
performance for each indicator;
(viii) The model is used to make adjustments to the negotiated State
levels of performance; and
(iv) The model is to be used in negotiations for the first two program
years (PY2016 and PY2017).
Negotiations
Requirements
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 9
The negotiated levels for each indicator are based upon:
Comparison with performance levels of other States;
The results of an objective statistical model;
Continuous improvement; and
The extent to which performance levels assist in reaching the
GPRA goals
WIOA specifically requires use of the statistical model in negotiations for
the first two program years (PY2016 and PY2017).
Statistical Model
Requirements
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 10
The Model must consider States’ actual economic conditions, including;
Differences in Unemployment Rate
Differences in job losses or gains in particular industries
The Model must also consider the characteristics of participants at the
time of enrollment;
Indicators of poor work history
Lack of work experience
Lack of educational or
occupational skills attainment
Dislocation from high-wage or
high-benefit employment
Low levels of literacy or English
proficiency
Disability status
Homelessness
Ex-offender status
Welfare dependency
Adjustment
Requirements
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 11
Is this my final answer?
The model must be used on the back end to
adjust the negotiated State levels of
performance for:
Actual participant characteristics, and
Actual economic conditions
Today’s Agenda
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 12
Overview of the requirement to use a Statistical Model
Discuss the phased-in approach to the Model and baseline indicators
Present the elements of the Model and the variables used
Explore the process of using the model to Negotiate Levels of
Performance
Explore the Model’s role in Determination of Performance Failure
Explore next steps in the development and use of the Model
Phase-In of the
Statistical Adjustment Model
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 13
As past outcomes and other labor market data becomes available, the
model will be updated over time with new data.
The original model was used in the negotiation of PY2016 and PY2017
goals for indicators where data was available.
The revised model (once updated with recent data) will be used to
re-negotiate PY2017 levels of performance, if a state opts to re-open
negotiations.
Generally, indicators for Titles I and III will be estimated in the model,
where data through PY 2015 is available, by the spring of 2017.
Data Availability and
Indicator Estimates
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 14
Indicators are categorized in the model based upon when data are
available.
Those indicators where there are insufficient data yet to support the
statistical model have been designated as Baseline Indicators.
PY2018 is the first year that the Baseline Indicators will be estimated
using the statistical adjustment model.
For PY2017 re-negotiations, these indicators will remain incompletely
populated with data, because some data is not yet available.
Baseline Indicators
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 15
Baseline Indicators, by core program title, include:
For Title I:
Measurable Skill Gains,
Effectiveness in Serving Employers,
Median Earnings (Youth program only)
For Title II:
All primary indicators (EXCEPT Measurable Skill Gains)
For Title III:
Effectiveness in Serving Employers
For Title IV:
All primary indicators
Today’s Agenda
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 16
Overview of the requirement to use a Statistical Model
Discuss the phased-in approach to the Model and baseline indicators
Present the elements of the Model and the variables used
Explore the process of using the model to Negotiate Levels of
Performance
Explore the Model’s role in Determination of Performance Failure
Explore next steps in the development and use of the Model
Elements of the Model
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 17
The model is required to include two elements:
Participant Characteristics – variables are presented
as percentages of the total exiters for the particular
cohort.
Economic Conditions – variables are presented as
percentages of total employment level by industrial
sector and the State unemployment rate.
Elements of the Model (cont’d)
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 18
Variables representing Participant Characteristics rely on existing
WIA data
Workforce Investment Act Standardized Record Data
(WIASRD) from PY 2005 through 2014 was used in the
development of the model and was the basis for PY 2016
and 2017 negotiations
The model is being updated with the PY 2015 WIASRD and
will be available for those States that wish to renegotiate
PY 2017 levels of performance
Elements of the Model (cont’d)
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 19
Gender
Age
Race/Ethnicity
Educational Level
Employment Status at
Participation
For reference, details of the variables can be found here: be found here:
https://www.doleta.gov/performance/guidance/docs/WIOA_St
atistical_Model_Methodology_Report-6-24-2016.pdf
https://wdr.doleta.gov/directives/attach/TEGL/TEGL_26-15-
Attachment-II_Acc.pdf
In addition to required elements, other relevant characteristics of
participants served are captured, including:
Elements of the Model (cont’d)
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 20
For reference, details of these sources can be found here: be found here:
https://www.bls.gov/cew/
https://www.bls.gov/lau/
Variables representing Economic Conditions rely on two datasets
provided by the Bureau of Labor Statistics (BLS):
Quarterly Census of Employment and Wages (QCEW) – used
to calculate the percentage of total employment for each
industrial sector within a State
Local Area Unemployment Statistics (LAUS) – used to
calculate the State Unemployment Rate
Today’s Agenda
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 21
Overview of the requirement to use a Statistical Model
Discuss the phased-in approach to the Model and baseline indicators
Present the elements of the Model and the variables used
Explore the process of using the model to Negotiate Levels of
Performance
Explore the Model’s role in Determination of Performance Failure
Explore next steps in the development and use of the Model
Negotiated Levels of
Performance
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 22
One part of the negotiations process
WIOA Sec. 116(b)(3)(A)(v) requires consideration of 4 factors in
the negotiation process.
comparison of levels of performance for other States,
the results of an objective statistical model,
continuous improvement in performance accountability by the
State and
the extent to which State performance levels will assist in
reaching GPRA goals.
Negotiated Levels of
Performance
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 23
Using the Performance Negotiations Tool
(Titles I and III)
The tool is intended for States and their Local Areas.
The tool provides users with benchmarks stemming from the
statistical adjustment model mandated in statute along with
a review of historical performance.
These benchmarks are intended to inform the negotiations
process between States and their Local Areas.
The tool is a starting point in the negotiations process.
Negotiated Levels of
Performance
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 24
Components of the Statistical Model Tool:
Target Outcome
the level at which states are expected to be as indicated by the statistical adjusted model (i.e., the 'national' model).
Predicted Outcome
The level at which states will most likely be as indicated by the statistical model (i.e., the fixed effects model).
Simulated Outcome
The simulated outcomes for the States or Local Area’s prior program year use the individual records for the last program year of each indicator's exiter cohort found in the Wagner-Peyser Employment Service data.
Negotiated Levels of
Performance
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 25
Target Outcome Calculation: 1. Multiply each coefficient by the associated input value.
• For example, for Alabama’s Adult Employment Rate in the 2nd
Quarter After Exit, the coefficient for female exiters is 0.05271.
• The result is multiplied by the percent of female exiters for
Alabama which is 0.58237. The product of multiplying these two
numbers is 0.03070.
• This step is repeated for every coefficient and associated input
value.
2. Add up all of the products from Step 1.
• For the Adult Employment Rate in the 2nd Quarter After Exit, the
sum of these products for Alabama is 0.18964.
Negotiated Levels of
Performance
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 26
Target Outcome Calculation: (continued)
3. Calculate the average State fixed effect by adding all of the State
fixed effects for the indicator and dividing that sum by the number of
States (52).
• For Alabama the Adult Employment Rate in the 2nd Quarter
After Exit, the average State fixed effect is 0.56105.
4. Add the average State fixed effect from Step 3 (0.56105) to the sum
in Step 2 (0.18964) for a target Adult Employment Rate in the 2nd
Quarter After Exit of 0.75069r 75.1%.
Negotiated Levels of
Performance
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 27
Predicted Outcome Calculation: 1. The predicted outcomes are calculated using Steps 1, 2 and 3 for the
Targeted Outcome.
2. In Step 4, instead of adding the average State effect, the specific State effect
is added to get a prediction which accounts for the effect of the State
program.
Simulated Outcome: 1. The simulated outcomes for the States or Local Area’s prior program year are
calculated using the individual records for the last program year of each
indicator's exiter cohort found in the Wagner-Peyser Employment Service
data.
1. effects of the specific State programs as well as other factors
Negotiated Levels of
Performance
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 28
Key Elements to be considered in the negotiation process:
Trend analysis/Past Performance
Interpret the range between projected outcome and projected
target
Adjusting some of the model inputs
Comparison across States
Continuous improvement
GPRA Goals
Reaching consensus on a negotiated target
Scenario: Additional
Environmental Factors & Low
Baselines
In one situation, the state wanted to increase its focus on serving a
particular population, in this case TANF.
DOL utilized the statistical model to show what the impact would be
on focusing on TANF by inputting a 200% increase in number of TANF
recipients served, which ultimately indicated minimal impact on
overall performance.
The state asked DOL to consider setting a lower EER goal. They cited
both the uncertainty of working with multiple partners with whom they
had no history and the unknown with regards to how this change
would impact their ability to meet enrollment goals and achieving
positive outcomes.
Today’s Agenda
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 30
Overview of the requirement to use a Statistical Model
Discuss the phased-in approach to the Model and baseline indicators
Present the elements of the Model and the variables used
Explore the process of using the model to Negotiate Levels of
Performance
Explore the Model’s role in Determination of Performance Failure
Explore next steps in the development and use of the Model
Definition of Terms
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 31
Expected Levels of Performance – provided by States in State Plans
Negotiated Levels of Performance – agreed upon between States
and DOL or ED
Actual Results – Reported in annual Statewide Performance Report
ETA-9169
Adjusted Levels of Performance – calculated after the close of the
program year
Adjusted Levels of
Performance
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 32
Initial Target Outcomes
Calculated using the statistical adjustment model
Provided by DOL prior to the start of the program year
*image from TEGL 26-15 Attachment III-B
WIOA Performance Negotiations Tool
Adjusted Levels of
Performance (cont’d)
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 33
Post-Program Year Target Outcomes
Uses the same model as the initial target outcomes
Calculated using actual participant characteristics and
economic conditions
Provided by DOL after the close of the program year
Adjusted Levels of
Performance (cont’d)
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 34
Adjusted Levels of Performance – for each indicator of performance:
Subtract the Initial Target Outcome from the Post-Program Year
Target Outcome
Add the difference to the Negotiated Level of Performance
Adjusted Level of Performance
(Set after the close of a Program Year)
Negotiated
Level of
Performance +
Difference
between first
and second
SAM outputs =
Adjusted
Level of
Performance
76.0% -1.3% 74.7%
Performance Scores
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 35
For each primary indicator of performance:
Calculation includes the numerator and denominator
(except in the case of median earnings)
Actual performance is divided by the adjusted level of
performance
Actual
Performance
Adjusted Level of
Performance
Performance
Score
20,00030,000
74.7%= 89.2%
For each WIOA title program and Primary Indicator of Performance:
The average performance score is calculated
This calculation has been outlined in Attachment 9 of TEGL 10-16
Average performance scores will be truncated to one decimal place
Average Performance
Scores
3/23/2017
15,90024,00075.2%
+
13,00018,00072.5%
+
20,00030,00074.7%
+
9,00015,00079.7%
+
6,00012,00068.7%
+
31,55542,00075.1%
6
= .89984 = 89.9%
Performance Scores
Average
Indicator Score
U.S. Department of Labor, Employment and Training Administration 36
Determination of Failure
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 37
Performance Failure occurs when:
Any single Performance Score is less than 50%, or
Any Average Performance Score is less than 90%
Sanctions are applied when:
The State fails to submit a complete and timely annual report, or
Any single Performance Score is less than 50% for 2 consecutive
years, or
Any Average Performance score is less than 90% for 2
consecutive years.
Today’s Agenda
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 38
Overview of the requirement to use a Statistical Model
Discuss the phased-in approach to the Model and baseline indicators
Present the elements of the Model and the variables used
Explore the process of using the model to Negotiate Levels of
Performance
Explore the Model’s role in Determination of Performance Failure
Explore next steps in the development and use of the Model
Next Steps
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 39
Brief Overview of Feedback
Revisions to the Model
Challenges to the Local Level Model
Next Steps – Feedback
and Possible Revisions
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 40
DOL has completed PY16 and PY17 negotiations for some indicators
where the data was available.
We have received feedback on a range of topics from variable
definitions, clustering, dissemination of the model, and its application
to the local board negotiations tool.
DOL will evaluate the accuracy of the model and make adjustments
Likely revisions include:
Industrial sector groupings
Adding in variable for pre-program wages
Next Steps – Challenges
of the Local Level Model
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 41
WIOA requires States use the statistical adjustment model when
negotiating levels of performance with local boards.
WIOA requires States use the statistical adjustment model to adjust
local levels of performance to reflect the actual economic conditions
experienced and the characteristics of the participants served.
There are two challenges to overcome:
Local area codes change often, which makes curating a dataset
comprised of historical data difficult; and
QCEW and LAUS data is not available at the local level. County-
level LMI data must be aggregated up to the local level, which is
a resource-intensive process.
Any Questions?
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 42
Resources
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 43
Body - Insert text here
Bullet list one
Bullet list two
Bullet list three
Insert more text here
Contact Information
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 44
For questions regarding the information in presentation, send an email to*:
*Please use [INSERT TRACK TITLE HERE] as the subject line.
3/23/2017
U.S. Department of Labor, Employment and Training Administration
U.S. Department of Labor, Employment and Training Administration 45