UCSB PGDCSM (Civil Service Management) cheat sheet

Merit system principles (5 U.S.C. §2301) prohibit appointment or advancement based on political affiliation and require selection on the basis of relative ability, knowledge, and skills.

Spoils/patronage systems allocate posts by political loyalty; merit systems allocate by competitive examination or demonstrated qualification - the Pendleton Act (1883) is the canonical U.S. transition point.

Position classification ties pay to the duties and responsibilities of a post; rank-in-person systems tie status to the individual employee regardless of assigned post.

Career vs. non-career appointments: career posts are competitive and protected from arbitrary removal; non-career/excepted posts are policy-linked and serve at the pleasure of the appointing authority.

Central personnel agency vs. delegated authority: central agencies set rules and standards (e.g., OPM-type bodies), while line departments execute recruitment, selection, and discipline under delegated authority.

Independent oversight bodies (merit boards, civil service commissions, ombuds institutions) hear appeals on adverse actions and enforce merit rules; their decisions bind agency personnel actions.

Unionization and collective bargaining in the public sector typically exclude core managerial prerogatives (hiring, discipline, mission assignment) from the bargaining scope.

Governance frameworks distinguish policy, management, and oversight functions; separating them reduces conflicts of interest between rule-setting and rule-application.

Total compensation = base salary + benefits + employer pension contribution + paid leave value; comparing base salary alone understates public-sector compensation relative to private-sector comparisons.

Job analysis produces the position description, which is the legal basis for classification, recruitment criteria, and performance standards; changing duties without updating it undermines defensibility.

Structured interviews with behaviorally anchored, job-related questions and panel scoring have higher predictive validity than unstructured interviews; scoring rubrics must be applied before discussion.

Veterans' preference and other statutory preferences alter the order of certification from eligible lists; the rule applied depends on the applicable statute and the type of appointment.

Probationary periods allow separation with reduced procedural burden; after probation, adverse actions require documented cause and due process (notice, opportunity to respond, decision).

Adverse action due process elements: written notice of the proposed action, the specific charges and supporting evidence, an opportunity to reply, and a written decision by an authorized official.

Workforce planning formula: projected separations + projected new-position growth − projected internal fill rate = external hiring need for the planning horizon.

Training return is evaluated at four levels - reaction, learning, behavior change on the job, and organizational results; only the last two indicate transfer to performance.

Data governance assigns decision rights: who may define a data element, who may change it, who may access it, and who is accountable for its quality.

Data stewardship is operational (quality, definitions, remediation); data ownership is accountable (policy, access approval, retention decisions) - the two roles should not be conflated.

Records retention schedules specify the minimum retention period and final disposition (destruction or transfer to archives); destruction before the schedule expires is a compliance violation.

Metadata standards describe content, format, provenance, and access restrictions; a dataset without metadata is not reliably reusable or defensible in audit.

Privacy controls follow purpose limitation, data minimization, and access on a need-to-know basis; re-identification risk rises with the number of quasi-identifiers combined.

Open data release requires de-identification review, license specification, and a documented publication workflow; sensitive categories (PII, law enforcement, security) require redaction or withholding.

Data quality dimensions: accuracy, completeness, timeliness, consistency, validity, and uniqueness; each dimension needs a defined measurement rule to be auditable.

Interoperability requires shared identifiers, controlled vocabularies, and versioned schemas; semantic mismatch between agencies is a governance failure, not a technical one.

Correlation does not establish causation; causal claims require a design (randomized assignment, difference-in-differences, regression discontinuity, instrumental variables) with stated identifying assumptions.

Difference-in-differences estimates the effect as (treated post − treated pre) − (comparison post − comparison pre); validity requires parallel trends absent treatment.

Regression discontinuity applies when treatment is assigned by a threshold on a running variable; the estimate is the jump at the cutoff, identified from units near it.

Statistical significance (p-value) is not effect size or policy importance; report confidence intervals and the magnitude of the estimate in decision-relevant units.

Power analysis determines the sample size needed to detect a stated effect at a given significance level and power; underpowered evaluations can miss real effects.

Attrition and selection bias threaten validity when the observed sample differs systematically from the population; compare baseline characteristics across groups to detect it.

Dashboards should distinguish leading indicators (early signals) from lagging indicators (outcomes) and must display denominators, definitions, and data vintage.

Predictive models used in government decisions require performance reporting on subgroups, documented error costs, and human review before adverse action.

Budget cycle stages: executive preparation, legislative authorization/appropriation, execution, audit and evaluation; each stage has distinct legal controls.

Appropriation = legal authority to incur obligations and make payments for specified purposes up to a stated amount and period; obligation occurs when a binding commitment is made.

Fund accounting separates resources by legal restriction; commingling restricted and unrestricted funds is a control failure.

Line-item budgeting controls inputs; performance budgeting links funding to outputs/outcomes; program budgeting groups costs by program; zero-based budgeting re-justifies each base from zero.

Budget execution controls include allotments/apportionments, obligation reviews, and expenditure monitoring against plan; variance analysis compares actual to allotment.

Internal controls follow the control environment, risk assessment, control activities, information/communication, and monitoring framework; segregation of duties is a core control activity.

Audit types: financial audit (fair presentation), compliance audit (adherence to law/regulation), performance audit (economy, efficiency, effectiveness).

Cost-benefit analysis decision rule: proceed when discounted benefits exceed discounted costs (NPV > 0); the discount rate and time horizon materially change the result.

Conflict-of-interest rules target financial interests and outside relationships that could impair impartial judgment; recusal or divestiture is the standard remedy.

Gift and hospitality rules typically set a de minimis threshold and require disclosure above it; the appearance standard can be stricter than the actual-bias standard.

Whistleblower protections generally require that the disclosure be made to a designated channel and concern a covered wrongdoing; retaliation against protected disclosure is prohibited.

Transparency mechanisms include proactive publication, freedom-of-information request processes, and open meetings; each carries defined exemptions (privacy, security, deliberation).

Accountability has vertical (elections, oversight) and horizontal (legislatures, courts, auditors, inspectors general) dimensions; internal controls support both.

Ethical decision frameworks require stating the public interest, identifying affected parties, checking legal authority, and documenting the reasoning for review.

Post-employment restrictions limit lobbying or representation before the former agency for a defined period; scope depends on the employee's duties and rank.

Records of decisions - rationale, alternatives considered, and evidence relied on - are the primary evidence in accountability review; undocumented decisions are hard to defend.

Service journey mapping identifies touchpoints, handoffs, and failure points from the user's perspective; back-office integration determines whether the front end can be simple.

Once-only principle: collect citizen data once and reuse it across services under legal authority, rather than requesting the same information repeatedly.

Digital identity assurance levels combine identity proofing, authenticator strength, and federation; access level must match the risk of the transaction.

Accessibility conformance (e.g., WCAG-type criteria) is a legal and functional requirement; inaccessible services exclude eligible users and create compliance risk.

Procurement of digital services should specify outcomes, data portability, and exit/transition terms; vendor lock-in is a governance risk, not just a cost risk.

Service performance measures include uptake, completion rate, time to complete, error/rework rate, cost per transaction, and user-reported satisfaction.

Multi-channel delivery must retain an assisted or non-digital path for users who cannot or will not use digital channels; channel choice is an equity issue.

System development lifecycle controls include requirements baselines, change control, testing evidence, and post-implementation review; skipping them raises delivery failure risk.

Logic model chain: inputs → activities → outputs → outcomes (short/intermediate) → impact; indicators must be attached to a specific link in the chain.

SMART criteria: specific, measurable, achievable, relevant, time-bound; a target without a baseline and a measurement source is not verifiable.

Outcome vs. output: outputs count what was produced (cases processed); outcomes describe change in the condition of the target population.

Performance measure quality tests: validity (measures the intended construct), reliability (consistent over repeated measurement), sensitivity (detects real change), and cost of collection.

Evaluation designs: pre-post (weak, no counterfactual), comparison group, randomized controlled trial, and quasi-experimental designs; strength depends on the counterfactual.

Process evaluation asks whether the program was implemented as designed (fidelity, reach, dosage); impact evaluation asks whether it caused the observed change.

Goodhart's law: when a measure becomes a target, it ceases to be a good measure - gaming risk rises when measures drive high-stakes consequences.

Performance reporting should include baselines, targets, actuals, variance explanations, and data limitations; a bare percentage without context is not decision-useful.