Table 2 Portfolio-level results, conditioned to Shared Socioeconomic Pathway 3-Representative Concentration Pathway 4.5 scenario (SSP3-RCP4.5), year 2040

From: Asset-level assessment of climate physical risk matters for adaptation finance

Panel A: portfolio-level results

Row

Case

Estimate (%)

Confidence interval (%)

1

EAI, asset-level (mean)

−0.085

(−0.16, −0.049)

2

RP250, asset-level (VaR)

−3.3

(−4.9, −2.2)

3

EAI, proxy (mean)

−0.013

(−0.056, −0.0028)

4

RP250, proxy (VaR)

−0.59

(−1.62, −0.17)

5

Chronic + EAI, asset-level (mean)

−0.84

(−1.09, −0.66)

6

Chronic + RP250, asset-level (VaR)

−3.9

(−5.5, −2.8)

7

Chronic + EAI, proxy (mean)

−0.77

(−1.01, −0.61)

8

Chronic + RP250, proxy (VaR)

−1.3

(−2.3, −0.8)

Panel B: underestimation of portfolio losses

Row

Compared cases

Underestimation range (%)

1

Tail acute (proxy, RP250) vs tail acute (asset-level, RP250)

67.4–92.3

2

Chronic and tail acute (proxy, RP250) vs Chronic and tail acute (asset-level, RP250)

58.0–70.8

  1. Panel A: portfolio-level results showing the mean and Value at Risk (VaR) computed for different cases of physical risk. The second column (Case) shows the selected case. The third column (Estimate (%)) shows the point estimate for the given metric and case. The fourth column (Confidence interval (%)) shows the 95% confidence intervals for the statistics, computed using the bias-corrected and accelerated percentile method over 15,000 samples. Cases labelled as asset-level are computed considering all data on assets for firms in the sample. Cases labelled as proxy are computed considering only proxy data for firms in the sample. Panel B: underestimation of portfolio losses, comparing cases pairwise. In each row, column 2 (Compared cases) lists the compared cases as case 1 vs case 2. Column 3 (Underestimation range (%)) is computed as the range of relative underestimation of the lower and upper bounds of the confidence intervals. The relative underestimation is computed as the relative difference between the boundaries of the confidence intervals for the first and second case. For example, on row 3 the underestimation range is computed as “[(confidence intervals, tail acute) - (confidence intervals, average acute)]./(confidence intervals, tail acute)", where “./" indicates element-wise division. Thus, the last column represents how large the underestimation of losses is when using case 1 with respect to case 2: a value of 50% implies that using case 1 we fail to capture 50% of the risk as quantified using case 2.
Back to article page