本文へ移動
cccskills
無料GitHub で公開

modeling-payment-in-kind-structures

Builds PIK and PIK toggle models with compound interest analysis, cash vs PIK election scenarios, and leverage impact assessment. Use when modeling PIK instruments, analyzing toggle mechanics, or evaluating accrued interest impact.

インストール方法を見る

含まれるファイル(1)

  • SKILL.md6.9 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Modeling Payment In Kind Structures

Builds PIK and PIK toggle models with compound interest analysis, cash vs PIK election scenarios, and leverage impact assessment.

When To Use

  • Modeling a new PIK or PIK toggle note issuance to project accrued principal growth and terminal payoff
  • Comparing cash-pay vs. PIK election economics for a borrower considering toggle exercise
  • Evaluating leverage creep on a credit with PIK features — assessing covenant headroom erosion over time
  • Stress-testing a PIK instrument under different rate, prepayment, and toggle-exercise scenarios
  • Preparing lender-side or sponsor-side analysis of PIK impact on returns (IRR/MOIC) and exit proceeds waterfall

Inputs To Gather

  • Instrument terms: Original principal, stated coupon rate, PIK spread (if separate from cash-pay spread), maturity date, compounding frequency (quarterly, semi-annual)
  • Toggle mechanics (if applicable): Toggle period start/end, exercise conditions (e.g., leverage ratio threshold, borrower election, mandatory triggers), partial PIK allowance (50/50 split provisions)
  • Cash-pay component: Cash interest rate when toggle is not exercised; any step-up or step-down schedule
  • Fees and OID: Original issue discount, upfront fees, call protection schedule (NC periods, make-whole, par call dates)
  • Capital structure context: Senior secured debt balance, total leverage at close, EBITDA projections, covenant levels (Total Leverage, Senior Secured Leverage, Fixed Charge Coverage)
  • Projection assumptions: Revenue/EBITDA growth rates, capex, mandatory amortization on other tranches, tax rate for interest deductibility analysis
  • Exit / refinancing assumptions: Target exit date, exit multiple, prepayment penalties at various horizons

Workflow

  1. Set up the PIK accrual schedule

    • Build a period-by-period schedule (quarterly or semi-annual) from closing to maturity
    • For each period: Beginning Balance + PIK Accrual = Ending Balance
    • PIK Accrual = Beginning Balance × (PIK Rate / Periods per Year)
    • Track cumulative accrued principal separately from original principal for reporting
  2. Model toggle election logic

    • Define toggle decision rules: borrower-elected, leverage-triggered, or hybrid
    • For leverage-triggered toggles: link toggle activation to projected Total Debt / EBITDA exceeding the threshold [VERIFY toggle threshold in credit agreement]
    • Build a binary flag (Cash-Pay = 1, PIK = 0) for each period; allow partial toggles if the instrument permits a blended cash/PIK split
    • Compute cash interest expense and PIK accrual separately based on the flag
  3. Build scenario matrix

    • Base case: Management EBITDA projections, toggle exercised per expected cadence
    • Full cash-pay: No PIK election — establishes maximum cash burden / minimum principal at maturity
    • Full PIK: PIK every period — establishes maximum principal growth / minimum near-term cash drain
    • Stress case: Revenue decline (e.g., −10–20%), forced toggle activation, extended PIK period
    • For each scenario, flow through: cash interest, PIK accrual, ending principal, total debt, leverage ratios
  4. Assess leverage and covenant impact

    • Recompute Total Leverage (Total Debt / EBITDA) each period, including accrued PIK principal in total debt
    • Check covenant compliance: flag periods where PIK accrual pushes leverage above maintenance or incurrence thresholds
    • Calculate incremental leverage attributable solely to PIK accrual (isolate PIK principal growth from operational deleveraging)
    • Model Fixed Charge Coverage impact — PIK reduces near-term cash interest but increases eventual principal repayment
  5. Compute lender return and borrower cost metrics

    • Lender IRR/MOIC: model cash inflows (cash interest + principal repayment at maturity/exit) against initial funded amount net of OID
    • Borrower all-in cost: effective yield to maturity including PIK compounding; compare against a hypothetical all-cash-pay instrument
    • Sensitivity table: IRR across exit years (Year 3–7) × PIK election scenarios × exit multiples
    • Note the asymmetry: PIK benefits borrower near-term cash flow but increases lender terminal return if held to maturity
  6. Run the exit / refinancing waterfall

    • At each modeled exit date: Enterprise Value = Exit EBITDA × Exit Multiple
    • Deduct senior debt payoff, then PIK note payoff (original principal + all accrued PIK)
    • Compute residual equity value and sponsor return
    • Highlight the "PIK drag" — reduction in equity proceeds attributable to accrued PIK vs. a cash-pay alternative

Output

  • PIK accrual schedule: Period-by-period table showing beginning balance, cash interest, PIK accrual, ending balance, and cumulative accrued amount
  • Toggle scenario comparison: Side-by-side summary of full cash-pay, full PIK, base case toggle, and stress case — with ending principal, total interest cost, and leverage at maturity
  • Leverage trajectory chart data: Period-by-period leverage ratios under each scenario with covenant threshold lines marked
  • Return summary: Lender IRR/MOIC table; borrower effective cost of capital; sensitivity grids
  • Exit waterfall: Proceeds distribution at modeled exit dates showing PIK drag on equity

Quality Checks

  • Confirm compounding math: Ending Balance in period N must equal Beginning Balance in period N+1; terminal balance under full-PIK should match manual compound interest calculation (P × (1 + r/n)^(n×t))
  • Verify toggle logic fires correctly — test boundary conditions where leverage is exactly at the toggle threshold
  • Ensure PIK accrued principal is included in Total Debt for all leverage calculations (some models mistakenly exclude it) [VERIFY credit agreement definition of "Indebtedness" to confirm PIK inclusion]
  • Cross-check that cash interest + PIK accrual in any period equals the full stated coupon applied to beginning balance
  • Validate that lender IRR under full cash-pay exceeds IRR under full PIK for early exits (PIK benefits lenders only when held long enough for compounding to overcome time-value discount)
  • Stress-test for circularity if toggle is leverage-triggered and leverage depends on PIK accrual (may require iterative calculation or macro)
  • Confirm OID amortization and call protection are correctly reflected in return calculations [VERIFY call schedule and make-whole provisions]

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Evaluates activist investor campaigns with thesis assessment, proposed changes, and likely outcome analysis. Use when analyzing activist situations, evaluating campaign theses, or assessing activist track records.

日本語の概要は準備中です。原文の説明を表示しています。

CaseMark/skills442026年9月9日 更新

Evaluates corporate vulnerability to shareholder activism with governance assessment, valuation gaps, and operational improvement opportunities. Use when assessing activist risk, identifying vulnerabilities, or preparing defensive analyses.

日本語の概要は準備中です。原文の説明を表示しています。

CaseMark/skills442026年9月9日 更新

Evaluates bolt-on acquisitions for existing platforms with strategic fit, synergy quantification, and return contribution analysis. Use when screening add-on targets, building platform acquisition strategies, or modeling tuck-in economics.

日本語の概要は準備中です。原文の説明を表示しています。

CaseMark/skills442026年9月9日 更新

Evaluates signal half-life, turnover implications, and capacity constraints for systematic alpha factors. Use when analyzing signal persistence, evaluating factor decay, or estimating strategy capacity.

日本語の概要は準備中です。原文の説明を表示しています。

CaseMark/skills442026年9月9日 更新

Evaluates alternative data sources including satellite, NLP sentiment, web scraping, and geolocation for alpha signal generation. Use when analyzing alt data, evaluating new data sources, or integrating non-traditional signals.

日本語の概要は準備中です。原文の説明を表示しています。

CaseMark/skills442026年9月9日 更新

Evaluates competition law exposure with market definition, HHI analysis, and remedy estimation for proposed transactions. Use when assessing merger clearance risk, analyzing market concentration, or preparing HSR filings.

日本語の概要は準備中です。原文の説明を表示しています。

CaseMark/skills442026年9月9日 更新

CaseMark のスキルをすべて見る

このスキルの問題を報告する