1. Introduction

Private mortgage lending occupies a distinctive niche in credit markets. Unlike regulated banks with diversified portfolios spanning thousands of loans across geographic regions and borrower segments, private lenders—known as Mortgage Investment Corporations (MICs) in Canada and analogous structures in other jurisdictions—typically hold concentrated portfolios of 50 to 200 loans. These loans are characterized by higher interest rates (8–14%), higher loan-to-value (LTV) ratios, shorter terms (typically 12 months), and borrowers who do not qualify for conventional financing.

This concentration creates a risk profile fundamentally different from that of a diversified bank. When a bank holding 10,000 mortgages experiences a rate shock, the law of large numbers provides substantial smoothing of default outcomes. A private lender holding 100 loans has no such luxury. A cluster of five defaults can meaningfully impact portfolio returns; ten simultaneous defaults can threaten the fund's viability.

The sensitivity of these portfolios to interest rate movements operates through multiple channels simultaneously. Rising rates increase the cost of borrowing for refinancing—and since most private mortgages are short-term bridge loans, borrowers must refinance or sell within 12 months. When rates rise, borrowers who planned to refinance into conventional mortgages may no longer qualify, pushing them toward default. Simultaneously, rising rates compress property valuations, reducing the collateral backing each loan and increasing loss severity when defaults do occur. This "double hit" mechanism—where the same macroeconomic factor simultaneously increases both the probability of default and the loss given default—creates nonlinear risk amplification that is poorly captured by traditional single-factor risk models.

Despite the growing importance of private lending in housing finance—particularly in markets like Canada, Australia, and the United Kingdom where private lenders fill gaps left by tightening bank regulation—quantitative risk analysis of concentrated private mortgage portfolios remains limited. Existing literature on mortgage credit risk (Merton, 1974; Vasicek, 2002) focuses primarily on large, diversified institutional portfolios where asymptotic approximations are valid. These approximations break down precisely in the regime relevant to private lenders: small portfolios with heterogeneous loans and concentrated exposures.

This paper develops a Monte Carlo simulation framework specifically designed to quantify the interest rate sensitivity of private mortgage portfolios. We construct a synthetic portfolio model calibrated to the characteristics of typical private lending operations and subject it to 10,000 portfolio realizations across 13 rate shock scenarios ranging from −200 basis points to +400 basis points. We examine how portfolio size, LTV composition, default correlation, and combined stress factors interact to determine portfolio risk.

Our approach is deliberately computational rather than analytical. While closed-form solutions exist for stylized portfolio models (Vasicek, 2002), the interaction effects we study—LTV-dependent default sensitivity, property-value-linked recovery rates, and correlation-amplified tail risk in small portfolios—resist tractable analytical treatment. Monte Carlo simulation allows us to capture these nonlinear interactions directly.

The remainder of this paper is organized as follows. Section 2 describes the portfolio model and its parameterization. Section 3 details the simulation design. Section 4 presents results across four dimensions: rate shock sensitivity, concentration risk, LTV composition effects, and stress testing. Section 5 discusses practical implications for portfolio construction and risk management. Section 6 addresses limitations, and Section 7 concludes.

2. Model Specification

2.1 Portfolio Structure

We model a private mortgage portfolio consisting of $N$ loans, each characterized by three attributes: principal amount $P_i$, loan-to-value ratio $\text{LTV}_i$, and contractual interest rate $r_i$. All loans have a one-year term, reflecting the typical duration of private mortgage arrangements.

Loan principals are drawn from a log-normal distribution:

$$P_i \sim \text{LogNormal}(\mu = 12.2, \sigma = 0.5)$$

This parameterization produces a right-skewed distribution with a mean principal of approximately $200,000 and a median near$180,000, consistent with the mix of residential renovation loans, bridge financing, and small commercial mortgages that typify private lending portfolios. The log-normal form captures the empirical observation that most private loans cluster in the $100,000–$300,000 range with occasional larger exposures up to $1,000,000.

Loan-to-value ratios are drawn uniformly:

$$\text{LTV}_i \sim \text{Uniform}(0.55, 0.85)$$

The lower bound of 55% reflects the conservative end of private lending (well-secured first mortgages on desirable properties), while the upper bound of 85% represents the aggressive end (higher-risk second mortgages or loans on properties requiring significant renovation). In practice, most private lenders target maximum LTVs of 75–80%, but portfolio composition inevitably includes higher-LTV positions through second mortgages or appraisal adjustments.

Contractual interest rates are drawn uniformly:

$$r_i \sim \text{Uniform}(0.08, 0.14)$$

These rates reflect the premium charged for private mortgage financing. Rates at the lower end (8%) correspond to well-secured first mortgages with creditworthy borrowers who simply need non-traditional financing (self-employed borrowers, for instance), while rates at the upper end (14%) reflect higher-risk positions including second mortgages, construction loans, or borrowers with credit impairments.

2.2 Default Model

Default probability is modeled as a logistic function of loan characteristics and the macroeconomic rate environment:

$$P(\text{default}_i) = \sigma(\beta_0 + \beta_1 \cdot \text{LTV}_i + \beta_2 \cdot \Delta r + \beta_3 \cdot \text{LTV}_i \cdot \Delta r)$$

where $\sigma(\cdot)$ denotes the sigmoid function $\sigma(x) = 1/(1 + e^{-x})$ and $\Delta r$ represents the change in interest rates relative to the baseline environment (in decimal form, so +200 basis points = 0.02).

The model parameters are set as follows:

These parameters produce baseline default rates (at $\Delta r = 0$) ranging from approximately 17% at LTV = 55% to 37% at LTV = 85%. While these rates appear high by conventional mortgage standards, they reflect the reality of private lending where borrowers are, by definition, those who cannot access conventional financing. Annual default rates of 15–30% are consistent with industry reports for subprime and non-qualifying mortgage segments, where "default" encompasses not only foreclosure but also missed payments, term extensions under duress, and forced refinancing at unfavorable terms.

The inclusion of the interaction term $\beta_3 \cdot \text{LTV} \cdot \Delta r$ is critical to our analysis. It captures the empirical observation that rate shocks disproportionately affect high-LTV borrowers. A borrower at 55% LTV who faces a 200-basis-point rate increase can typically sell the property and repay the loan even if refinancing proves difficult. A borrower at 85% LTV in the same scenario has minimal equity cushion and is far more likely to find themselves underwater if property values simultaneously decline.

Under our parameterization, a +200 basis point rate shock increases default probability from 17.2% to 25.8% for a loan at 55% LTV (a 50% relative increase), but from 37.2% to 52.9% for a loan at 85% LTV (a 42% relative increase in probability but a much larger absolute increase of 15.7 percentage points). At +400 basis points, the high-LTV default rate reaches 68.0%.

2.3 Loss Given Default

When a loan defaults, the lender recovers value through property disposition—typically a power-of-sale or foreclosure process. The recovery amount depends on the relationship between the outstanding loan balance and the realizable property value:

$$\text{LGD}_i = \max\left(0, 1 - \frac{1}{\text{LTV}_i} \cdot (1 - d_i) \cdot (1 + s)\right)$$

where $d_i$ is the forced-sale discount and $s$ is a property value shock (negative values indicate property depreciation).

The forced-sale discount $d_i$ captures the empirical observation that distressed property sales realize prices below fair market value due to time pressure, poor property condition (defaulting borrowers often defer maintenance), legal costs, and market stigma:

$$d_i \sim \text{Beta}(2, 5)$$

This Beta(2,5) distribution has a mean of 0.286 (approximately 29% discount) and a mode of 0.20 (20% discount), with the majority of realizations falling between 10% and 50%. The right skew of this distribution reflects that while most forced sales involve moderate discounts, occasional fire sales—particularly of specialized or damaged properties—can result in severe value destruction.

For a loan at 75% LTV with no property shock and a typical 29% forced-sale discount, the expected LGD is approximately:

$$\text{LGD} = \max\left(0, 1 - \frac{1}{0.75} \cdot (1 - 0.29)\right) = \max(0, 1 - 0.947) = 0.053$$

This modest 5.3% loss given default reflects the equity cushion at 75% LTV. However, for a loan at 85% LTV with a 20% property value decline:

$$\text{LGD} = \max\left(0, 1 - \frac{1}{0.85} \cdot (1 - 0.29) \cdot (1 - 0.20)\right) = \max(0, 1 - 0.669) = 0.331$$

The LGD jumps to 33.1%—a sixfold increase demonstrating the nonlinear amplification that occurs when multiple adverse factors coincide.

2.4 Default Correlation

In the baseline model, defaults are independent conditional on loan characteristics and the rate environment. However, private mortgage portfolios often exhibit geographic concentration (loans within a single metropolitan area) and borrower-type concentration (similar credit profiles), creating systematic default correlation.

We model correlated defaults using the Vasicek (2002) single-factor Gaussian copula framework. Each loan $i$ has a latent variable:

$$Z_i = \sqrt{\rho} \cdot M + \sqrt{1 - \rho} \cdot \epsilon_i$$

where $M \sim N(0,1)$ is a common systematic factor, $\epsilon_i \sim N(0,1)$ are idiosyncratic factors, and $\rho$ is the asset correlation parameter. Loan $i$ defaults when:

$$\Phi(Z_i) < P(\text{default}_i)$$

where $\Phi(\cdot)$ is the standard normal CDF. This framework preserves the marginal default probabilities while introducing dependence through the shared factor $M$. We examine $\rho = 0$ (independent defaults), $\rho = 0.3$ (moderate geographic concentration), and $\rho = 0.5$ (high concentration).

3. Simulation Design

3.1 Monte Carlo Framework

For each scenario, we generate $S = 10{,}000$ independent portfolio realizations. Each realization proceeds as follows:

1. Portfolio generation: Draw $N$ loans with principals $P_i$, LTV ratios $\text{LTV}_i$, and contractual rates $r_i$ from the distributions specified in Section 2.1.

2. Interest income calculation: Compute total expected interest income $I = \sum_{i=1}^{N} P_i \cdot r_i$.

3. Default determination: Compute default probabilities $p_i$ using the logistic model. Generate defaults either independently (Bernoulli draws) or using the Gaussian copula for correlated scenarios.

4. Loss calculation: For each defaulted loan, draw a forced-sale discount $d_i$ from Beta(2,5) and compute the loss $L_i = P_i \cdot \text{LGD}_i$.

5. Portfolio return: Compute the portfolio return as $R = (I - \sum L_i) / \sum P_i$.

3.2 Scenario Design

We organize our simulations along four dimensions:

Rate shock sensitivity: We vary $\Delta r$ from −200 bps to +400 bps in 50 bps increments (13 scenarios), with $N = 100$, no property shock, independent defaults.

Concentration risk: We vary portfolio size $N \in {25, 50, 100, 150, 200}$ at three rate shock levels ($\Delta r \in {0, +200, +400}$ bps).

LTV composition: We examine four portfolio types defined by LTV range—low (55–65%), medium (65–75%), high (75–85%), and full range (55–85%)—across six rate shock levels.

Stress testing: We examine extreme scenarios including a 2008-style crisis ($\Delta r = +300$ bps, property values −20%, doubled default rates), gradual tightening (cumulative 50 bps per quarter), correlation shocks ($\rho = 0.3$ and $\rho = 0.5$), and combined extreme scenarios.

3.3 Risk Metrics

For each scenario, we report:

• Mean return: Expected portfolio return across simulations.
• Value-at-Risk (VaR) at 95%: The maximum loss at the 5th percentile of the return distribution. Formally, $\text{VaR}_{95} = -F^{-1}(0.05)$ where $F$ is the portfolio return CDF. We report this as a positive number when representing a loss and negative when the 5th percentile return is positive (i.e., even the worst 5% of outcomes are profitable).
• Conditional VaR (CVaR) at 95%: The expected loss in the worst 5% of outcomes, $\text{CVaR}_{95} = -E[R | R \leq F^{-1}(0.05)]$.
• Probability of negative return: The fraction of simulations where the portfolio loses money.
• Maximum loss: The worst single outcome across 10,000 simulations.

All simulations were implemented in Python using NumPy and SciPy with vectorized operations across the simulation dimension for computational efficiency. The complete simulation suite (approximately 600,000 portfolio realizations across all scenarios) executed in 25.9 seconds.

4. Results

4.1 Baseline Portfolio Characteristics

Under the baseline scenario ($\Delta r = 0$, $N = 100$, full LTV range, independent defaults), the private mortgage portfolio generates a mean annual return of 8.39% with a standard deviation of 1.01%. The return distribution is moderately left-skewed, with a median of 8.49% slightly exceeding the mean.

The 5th percentile return is 6.58%, meaning that even in relatively unfavorable outcomes, the portfolio remains solidly profitable. The 1st percentile return is 5.52%, and the worst outcome across 10,000 simulations is 3.25%. At baseline, the probability of a negative portfolio return is effectively zero.

These statistics reflect the fundamental economics of private lending: the high interest rates (averaging 11% across the portfolio) provide a substantial income buffer against credit losses. Under baseline conditions, an average of 26.3 loans out of 100 default, but the mean loss rate is only 1.66% of portfolio principal, well within the interest income margin.

The full distribution of baseline returns illustrates this point:

4.2 Rate Shock Sensitivity

Table 1 presents portfolio risk metrics across the full range of rate shocks. The results reveal a monotonic but nonlinear degradation in portfolio performance as rates increase.

Table 1: Portfolio Risk Metrics by Rate Shock (N = 100, Full LTV Range)

Negative VaR indicates that even the 5th percentile outcome produces a positive return.

Several patterns emerge from Table 1. First, the mean return declines approximately 58 basis points per 100 basis points of rate shock—a sensitivity ratio of 0.58. This is substantial: a +400 bps rate shock reduces the expected return from 9.34% (at −200 bps) to 5.89% (at +400 bps), a decline of 345 basis points representing 37% of the favorable-environment return.

Second, the number of defaults rises from 16.7 (at −200 bps) to 52.6 (at +400 bps)—more than tripling. This dramatic increase in default frequency is absorbed by the portfolio's interest income buffer under normal conditions, but it significantly narrows the margin of safety.

Third, the VaR compression tells an important story about tail risk resilience. The gap between mean return and 5th percentile return narrows from 152 bps (at −200 bps) to 236 bps (at +400 bps), reflecting increased dispersion in outcomes as defaults become more frequent and variable. While the portfolio remains profitable in all 10,000 simulations even at +400 bps, the safety margin has thinned substantially—the worst-case return at +400 bps is likely near breakeven, leaving no room for additional adverse factors.

4.3 Concentration Risk

Table 2 examines how portfolio size affects risk metrics, a critical question for private lenders choosing between fewer large loans and many smaller ones.

Table 2: Risk Metrics by Portfolio Size

The concentration risk results are striking. A 25-loan portfolio—which is not unusual for a small private lender or a single-property-type fund—has a 2.9% probability of negative returns under a +400 bps rate shock, even with independent defaults. This means that in roughly one out of every 34 years (assuming annual rate shocks of this magnitude are rare), such a portfolio would lose money. The VaR at the 5th percentile is only 0.95%, meaning the portfolio is barely profitable in the worst 5% of outcomes.

The diversification benefit of increasing portfolio size is substantial but exhibits diminishing returns. Moving from 25 to 50 loans improves the baseline VaR by 117 basis points; moving from 50 to 100 adds another 85 basis points; but moving from 100 to 200 adds only 52 basis points. This pattern reflects the $1/\sqrt{N}$ scaling of idiosyncratic risk reduction.

A practical implication emerges: the minimum viable portfolio size for a private lender depends on the risk tolerance and the expected rate environment. Under baseline conditions, even 25 loans suffice (only 0.2% probability of loss). Under stress, the minimum increases sharply—a lender targeting zero probability of loss at the 95th percentile under a +200 bps shock needs at least 50 loans.

4.4 LTV Composition Effects

Table 3 reveals the most policy-relevant finding of our analysis: the dramatic difference in risk profiles between low-LTV and high-LTV portfolios.

Table 3: Risk Metrics by LTV Composition and Rate Shock

The divergence between LTV segments is extraordinary. At +400 bps:

• A low-LTV portfolio (55–65%) retains a mean return of 9.11% with a 5th percentile return of 7.67%. This portfolio is essentially impervious to rate shocks—the combination of low default probability and strong collateral coverage provides a robust buffer.

• A high-LTV portfolio (75–85%) has its mean return collapse to 1.71%, and critically, the VaR becomes positive at +1.26%, meaning the 5th percentile outcome is a loss of 1.26%. This portfolio is losing money in more than 5% of scenarios under a +400 bps shock.

The rate sensitivity differential is equally revealing. Each 100 bps of rate shock reduces the mean return of the low-LTV portfolio by approximately 24 bps (a sensitivity of 0.24), while the high-LTV portfolio loses approximately 115 bps per 100 bps of rate shock (a sensitivity of 1.15). The high-LTV portfolio has nearly five times the rate sensitivity of the low-LTV portfolio.

This amplification arises from the interaction term in our default model. High-LTV borrowers are not only more likely to default at baseline (the $\beta_1$ effect) but become disproportionately more likely to default as rates rise (the $\beta_3$ interaction effect). Moreover, when they do default, the higher LTV means less collateral protection, resulting in larger losses per default.

The practical implication is unambiguous: LTV is the dominant risk factor in private mortgage portfolio construction, more important than diversification (portfolio size), interest rates charged, or geographic distribution.

4.5 Stress Testing

We now examine extreme scenarios that combine multiple adverse factors simultaneously.

4.5.1 The 2008-Style Crisis

Our most severe standard stress test models conditions analogous to the 2008 financial crisis: a +300 bps rate shock, a 20% decline in property values (reducing collateral and increasing LGD), and a doubling of default rates (capturing the cascading effects of credit market dysfunction, appraisal downgrades, and lender withdrawal).

Under this scenario, the 100-loan portfolio with full LTV range produces:

This is a qualitatively different outcome. The portfolio loses money in every single simulation out of 10,000. The mean loss of 7.06% means that investors lose approximately seven cents on every dollar invested. In the worst 5% of outcomes, losses exceed 10.48%, and the single worst outcome shows a loss of 15.22%.

The transition from the baseline scenario (where the probability of loss is 0%) to the crisis scenario (where it is 100%) illustrates the nonlinear nature of private lending risk. The portfolio does not degrade gracefully—it falls off a cliff when multiple adverse factors coincide.

4.5.2 Gradual Tightening

A more realistic stress scenario involves gradual rate increases over multiple quarters:

Table 4: Gradual Tightening Scenario (50 bps per Quarter)

Under gradual tightening, the portfolio degrades steadily but remains profitable even after a cumulative 200 bps increase. The mean return declines from 8.39% (pre-tightening) to 7.19% (after four quarters), a reduction of 120 bps or approximately 14% of the initial return. This scenario, while painful for investors expecting stable high returns, does not threaten capital.

The gradual nature of the shock, however, masks an important dynamic not captured in our single-period model: in reality, borrowers who default in Q1 may not yet have been replaced, reducing portfolio size and increasing concentration risk for subsequent quarters. This cascading effect would amplify the losses beyond what our static analysis suggests.

4.5.3 Correlation Shock

Table 5 examines the impact of default correlation—the critical factor that transforms idiosyncratic loan-level risk into systemic portfolio-level risk.

Table 5: Effect of Default Correlation on Portfolio VaR

Default correlation has a dramatic effect on tail risk. At baseline ($\Delta r = 0$), introducing moderate correlation ($\rho = 0.3$) worsens the VaR by 200 basis points—from −6.58% (5th percentile return of 6.58%) to −4.58% (5th percentile return of 4.58%). The VaR ratio (correlated/independent) is 0.70, meaning correlation destroys 30% of the tail-risk protection provided by diversification.

At +400 bps with $\rho = 0.5$, the VaR narrows to just −1.21%, meaning the 5th percentile return is barely above 1%. The portfolio is approaching the danger zone where negative returns become probable.

The interaction between correlation and rate shocks is approximately additive in our model: correlation reduces VaR by approximately 200 bps (for $\rho = 0.3$) regardless of the rate shock level. This additivity means that the marginal impact of correlation is more dangerous in stressed environments, where the remaining buffer is smaller.

4.5.4 Combined Extreme Scenarios

We examine two combined stress scenarios to understand the portfolio's behavior under compounded adversity:

Table 6: Combined Stress Scenarios

The combined scenario (moderate correlation + rate shock + property decline, without the default multiplier) produces a mean return that is still positive at 3.06%, but 23.2% of simulations result in losses. The VaR flips positive to 3.52%, indicating meaningful tail risk.

The ultra-extreme scenario pushes the portfolio into comprehensive failure: mean loss of 11.79%, with the worst outcome showing a 20.10% loss. Investors would lose approximately one-fifth of their capital in the worst case.

Most alarming is the high-LTV crisis scenario. A portfolio concentrated in 75–85% LTV loans, subjected to the same extreme stress, produces a mean loss of 17.46% with a maximum loss of 24.51%. In the worst case, investors lose nearly a quarter of their capital in a single year. This scenario, while extreme, is not unprecedented—it roughly corresponds to what a high-LTV-focused private lender in a peripheral housing market would have experienced during the 2008–2009 financial crisis.

5. Practical Implications

5.1 Minimum Diversification Thresholds

Our concentration risk analysis suggests the following minimum portfolio sizes for private lenders, based on target risk tolerance:

Table 7: Recommended Minimum Portfolio Size

These thresholds assume full LTV range (55–85%) and independent defaults. With correlated defaults, all thresholds increase by approximately 50%.

5.2 LTV Management Rules

The LTV composition analysis yields the clearest practical guidance. We recommend:

1. Portfolio-weighted average LTV below 70%: At this threshold, the portfolio retains positive expected returns even under +400 bps shocks with independent defaults.

2. Maximum individual loan LTV of 80% under normal conditions, 75% under elevated rate environments: Our simulation shows that loans above 80% LTV contribute disproportionately to portfolio risk.

3. High-LTV allocation cap of 25%: Even within a well-diversified portfolio, limiting the proportion of loans above 75% LTV to 25% of portfolio principal significantly reduces tail risk.

5.3 Reserve Requirements

Based on our simulation results, we suggest the following reserve (capital buffer) structure:

Table 8: Suggested Capital Reserves by Portfolio Composition

These reserves are sized to cover the VaR at the 95th percentile under each environment. The crisis reserve is designed to absorb the combined extreme scenario (correlation + rate shock + property decline) without breaching the zero-return threshold.

5.4 The Double-Hit Mechanism

Our results confirm and quantify the "double hit" phenomenon in private lending. When rates rise:

1. Default probability increases: The logistic default model's rate sensitivity ($\beta_2 = 15.0$) means that defaults approximately triple between −200 bps and +400 bps.

2. Recovery rates decline: Because rising rates typically accompany property market cooling, the collateral backing each loan may simultaneously depreciate.

3. The interaction amplifies: High-LTV loans experience both effects more severely (captured by the interaction term $\beta_3$).

The magnitude of this double hit is substantial. Moving from the most favorable scenario (−200 bps, no property shock) to the crisis scenario (+300 bps, −20% property values, doubled defaults) transforms the portfolio from one that earns 9.34% with zero probability of loss to one that loses 7.06% with certainty. This 16.4 percentage point swing in mean return—from a healthy profit to a guaranteed loss—illustrates why private lending risk is often underestimated during benign periods and why sudden crises can be devastating.

5.5 Correlation as a Hidden Risk Factor

Perhaps the most underappreciated risk factor in private lending is default correlation. Our results show that introducing moderate correlation ($\rho = 0.3$)—a reasonable estimate for a geographically concentrated portfolio—worsens the VaR by approximately 200 basis points across all rate scenarios. This is equivalent to approximately 200 bps of additional rate shock in terms of its impact on tail risk.

Private lenders can mitigate correlation risk through:

• Geographic diversification: Lending across multiple metropolitan areas reduces the common factor loading.
• Borrower-type diversification: Mixing renovation loans, bridge financing, and small commercial reduces sector-specific correlation.
• Property-type diversification: Combining residential, small commercial, and mixed-use collateral reduces property market correlation.

However, achieving this diversification requires scale (more loans across more markets), which creates tension with the typical private lender's local expertise and relationship-based origination model.

6. Limitations

Several limitations should be noted when interpreting these results.

Synthetic data: All portfolio characteristics are drawn from assumed distributions rather than calibrated to empirical data from actual private lending portfolios. While we believe our parameterization is reasonable based on industry reports and market observation, the specific numerical results should be treated as illustrative rather than predictive. Future work should calibrate the model to proprietary lending data where available.

Simplified default model: Our logistic default model captures the key economic mechanisms (LTV sensitivity, rate sensitivity, interaction) but omits many factors that influence default in practice, including borrower creditworthiness, property location quality, loan purpose, equity investment by the borrower, and macroeconomic conditions beyond interest rates. The model also assumes a fixed functional form that may not capture the true nonlinearity of default behavior at extreme LTV or rate levels.

Static single-period framework: We model a one-year holding period without dynamic portfolio effects. In practice, defaults in early months reduce portfolio size and increase concentration for later months; prepayments and renewals alter portfolio composition; and lenders adjust origination standards in response to market conditions. A multi-period model with dynamic portfolio rebalancing would be more realistic but substantially more complex.

No real calibration of default parameters: The logistic model coefficients ($\beta_0$ through $\beta_3$) are set to produce default rates that we believe are reasonable for private lending, but they are not estimated from empirical default data. This is perhaps the most significant limitation: the absolute levels of risk we report depend directly on these parameters, and different parameterizations could yield substantially different results.

Loss given default simplification: Our LGD model captures the key mechanism (LTV-dependent recovery with forced-sale discounts) but omits several important factors including foreclosure timelines (which affect carrying costs), accrued interest during the default process, legal fees, and geographic variation in the efficiency of foreclosure proceedings.

Correlation model limitations: The Gaussian copula framework, while standard in credit risk modeling, is known to underestimate tail dependence—the tendency for defaults to become more correlated precisely during crisis periods (Li, 2000). Our fixed correlation parameter $\rho$ does not capture this time-varying correlation, which means our stress test results may underestimate the severity of true crisis outcomes.

7. Conclusion

This paper presents a Monte Carlo framework for analyzing interest rate sensitivity in private mortgage portfolios. Through 10,000 simulations across 13 rate shock scenarios, three portfolio sizes, four LTV compositions, and multiple stress scenarios, we identify several key findings.

First, private mortgage portfolios are resilient under normal conditions. A well-diversified 100-loan portfolio with mixed LTV ratios generates mean returns of approximately 8.4% with effectively zero probability of loss, even with relatively high individual default rates. The high interest rates charged in private lending (8–14%) provide a substantial income buffer against credit losses.

Second, this resilience degrades nonlinearly under stress. The transition from normal conditions to crisis conditions is not gradual but abrupt. A +300 bps rate shock combined with 20% property depreciation and doubled default rates transforms the portfolio from one that never loses money to one that always loses money—a binary shift from safety to crisis.

Third, LTV is the dominant risk factor. Portfolio-level rate sensitivity varies by a factor of five between low-LTV (55–65%) and high-LTV (75–85%) compositions. Under +400 bps stress, low-LTV portfolios maintain 7.67% returns at the 5th percentile while high-LTV portfolios incur losses. This finding has direct implications for portfolio construction: managing average and maximum LTV is more important than achieving geographic diversification or increasing portfolio size.

Fourth, default correlation is a hidden amplifier. Moderate correlation ($\rho = 0.3$) worsens the portfolio VaR by approximately 200 basis points, equivalent to approximately 200 bps of additional rate shock. Private lenders with geographically concentrated portfolios face meaningfully higher tail risk than their individual loan metrics would suggest.

Fifth, minimum portfolio size depends on the risk environment. Under baseline conditions, 25–50 loans provide adequate diversification. Under +400 bps stress, 100+ loans are needed to keep the probability of loss negligible.

These findings provide a quantitative foundation for portfolio construction rules, reserve requirements, and stress testing protocols for private mortgage lenders. The framework itself—rather than the specific numerical results, which depend on model calibration—is the primary contribution. Private lenders can adapt the model to their own portfolio characteristics, implement it with standard numerical computing tools, and use it for scenario analysis and risk monitoring.

Future work should focus on calibrating the default model to empirical data from actual private lending portfolios, extending the framework to a multi-period setting with dynamic portfolio rebalancing, incorporating time-varying correlation structures, and validating the model's predictions against historical loss experiences during past rate tightening cycles.

References

Merton, R. C. (1974). On the pricing of corporate debt: The risk structure of interest rates. Journal of Finance, 29(2), 449–470.

Vasicek, O. A. (2002). The distribution of loan portfolio value. Risk, 15(12), 160–162.