An ETF Portfolio for [Very] Uncertain Times

 

Folk music icon Bob Dylan’s song “The times they are a-changin” has never been more true. We are living in a period of unprecedented geopolitical and economic upheaval-the final consequences of which we are not yet able to discern. Uncertainty is the silent killer of financial markets. But while Uncertainty by itself is bad for the financial markets, the events that unfold during the period of uncertainty can be either good or bad. More US tariffs? Bad. Fed rate cuts? Good. Powell’s term ends in May 2026 and new Trump-appointed Fed Chair leads to even more rate cuts in 2026? Good in the short-term. This leads to heightened inflation and asset bubbles in the long-term? Ultimately bad.  

So, today we design a portfolio in which the main objective is to be low risk and yet yield a return that is above the rate of inflation. And we will do an analysis and 1-year ahead forecast for the Portfolio, and use the SPY ETF as a reference benchmark.  

We   are not attempting to outperform the S&P500 but to build a resilient portfolio. This portfolio will be good for your nerves and for poor old retirees like me. The portfolio will consist of 3 established ETFS with moderate liquidity.  Two of the ETFs will be for fixed income, and one will be for high quality dividend global equities. Our heuristics for selection of the ETFs is based on the premises that: (1) For the next 1 year, interest rates are not going to rebound (2) The USD will keep trending downwards as BRICS and China find ways to bypass it in cross-border transactions (3) Blue-chip global companies will still be able to  maintain their dividend policy. (4)  Investment grade corporate bonds will still have a place in the portfolios of Institutional Funds.

Portfolio ETFs and their percentage of $ invested in the portfolio.

LQD: iShares iBoxx Investment Grade Corporate Bond ETF: 30%

EMLC: VanEck J.P. Morgan Emerging Market Local Currency Bond ETF

IDV: iShares International Select Dividend ETF

Methodology and sequence in building the model

To model returns realistically without assuming a rigid formula, we used a regularized autoregression called Ridge AR(5). Think of it like ordinary least squares, but with a gentle brake (the L2 regularization) that keeps coefficients from overreacting to noise. The model learns short-run structure from the last five days of returns, capturing whatever mild momentum or mean-reversion is present, while staying stable in the face of noisy financial data.

For forward-looking scenarios, we simulate with a residual-bootstrap Monte Carlo. Each day ahead, the model predicts a return from recent history (the “predictable” piece), and we then add a randomly resampled residual (the “unpredictable” shock the model couldn’t explain). Repeating this step-by-step for 252 trading days and thousands of paths gives us a realistic fan of outcomes: medians, percentile bands, and endpoint distributions (via Kernel Density Estimation [KDE}). This approach preserves short-term dynamics while letting shocks reflect actual market variability, aligning with our emphasis on resilience in uncertain times.

From the simulations, we produced cumulative return trajectories and extracted key distributional summaries (median, 10th–90th percentile bands, endpoint quantiles, and histograms/KDEs [Kernel Density Estimates). In line with the low-volatility objective, we emphasized risk diagnostics: historical annualized volatility, maximum drawdown, daily VaR95 (95^th quantile Variance), CVaR95 (Conditional 95^th quantile variance=expected shortfall), and rolling max drawdown series. The resulting visuals and tables show the portfolio’s tighter forecast band and milder downside compared with SPY, aligning with an “uncertain times” mandate that accepts slightly lower expected returns in exchange for a narrower range of outcomes and improved downside characteristics.

Simple language explanation of  KDE, Cvar95  and Maximum Rolling Drawdown

KDE: KDE stands for Kernel Density Estimation. It’s a smooth, non-parametric way to estimate a probability density from data. Intuition: place a small “bump” (kernel) on each data point, then sum them up to form a continuous curve. The bandwidth controls how wide each bump is: Small bandwidth: very wiggly (overfits noise). Large bandwidth: very smooth (may hide structure). We used a Gaussian kernel with automatic bandwidth rule (Scott or Silverman rule).

CVar95: We based our model on daily returns.  VaR95 is the loss threshold you don’t expect to exceed on 95% of days. Since the remaining tail is 5%, that’s about 1 day out of every 20.  CVaR95 is the average loss on those tail days. It tells you how bad losses tend to be when you are in that “1 out of 20” bucket. At 95^th quantile the residual is 0.05   Frequency=1/0.5 =20 days

Rolling Maximum Drawdown: Rolling maximum drawdown (MDD) tells you how far an investment has fallen from its most recent peak, at each point in time. Example:

Day 1: Value = 100 → Peak = 100 → Drawdown = 100/100 − 1 = 0%

Day 2: Value = 110 → Peak = 110 → Drawdown = 110/110 − 1 = 0%

Day 3: Value = 99 → Peak = 110 → Drawdown = 99/110 − 1 ≈ −10.0%

Day 4: Value = 105 → Peak = 110 → Drawdown = 105/110 − 1 ≈ −4.5%

Day 5: Value = 112 → Peak = 11

Analysis and Visuals

Table 1: 1-year ahead probabilistic forecast.

Table 1 shows that in terms of return, SPY median forecast is 1.56 percentage points higher (16.10-14.54). We use median instead of mean forecast because of the element of uncertainty which is the subject of our post.  In terms of volatility (standard deviation) our Portfolio beats SPY hands down: 9.42% vs 23.31 %

Table 2: Historical Risk Metrics: Portfolio vs SPY

Image 1: Post-Monte Carlo Probability Density Function (PDF) of Portfolio and SPY

Image 1 shows that the PDF of SPY has very long tails and is skewed to the right, indicating higher volatility but with possible significant upside. In comparison, our Portfolio PDF is narrow and symmetrical and forecast is more reliable.

Image 2: Rolling maximum drawdown

Image 2 shows the great disparity in rolling maximum drawdown between our Portfolio and SPY.

Image 3: Monte-Carlo path of 1-year ahead Portfolio and SPY forecast

Conclusion

Who knows what comes tomorrow? The current US Administration is utterly unpredictable. Who knows what they will do next? Revalue the Gold stored in Fort Knox from $42.00 to $4000.00? Create a Stable Coin underpinned by USD and force issuers to buy Treasuries and back it with Treasuries? Change their mind about China security risks if China invests US $1 trillion in the US? Invade Greenland and declare Canada the 51^st S State? And the list goes on. So, stay safe and live well.