diff --git a/changelog_entry.yaml b/changelog_entry.yaml
index e69de29..ab1cb4c 100644
--- a/changelog_entry.yaml
+++ b/changelog_entry.yaml
@@ -0,0 +1,4 @@
+- bump: minor
+  changes:
+    added:
+    - Normalization parameter to handle multi-level geography calibration.
diff --git a/src/microcalibrate/reweight.py b/src/microcalibrate/reweight.py
index 3145a70..e56ae71 100644
--- a/src/microcalibrate/reweight.py
+++ b/src/microcalibrate/reweight.py
@@ -24,6 +24,7 @@ def reweight(
     epochs: Optional[int] = 2_000,
     noise_level: Optional[float] = 10.0,
     learning_rate: Optional[float] = 1e-3,
+    normalization_factor: Optional[torch.Tensor] = None,
     csv_path: Optional[str] = None,
     device: Optional[str] = None,
 ) -> tuple[np.ndarray, np.ndarray]:
@@ -38,6 +39,9 @@ def reweight(
         epochs (int): Optional number of epochs for training.
         noise_level (float): Optional level of noise to add to the original weights.
         learning_rate (float): Optional learning rate for the optimizer.
+        normalization_factor (Optional[torch.Tensor]): Optional normalization factor for the loss (handles multi-level geographical calibration).
+        csv_path (Optional[str]): Optional path to save the performance metrics as a CSV file.
+        device (Optional[str]): Device to run the calibration on (e.g., 'cpu' or 'cuda'). If None, uses the default device.
 
     Returns:
         np.ndarray: Reweighted weights.
@@ -103,7 +107,7 @@ def dropout_weights(weights: torch.Tensor, p: float) -> torch.Tensor:
         optimizer.zero_grad()
         weights_ = dropout_weights(weights, dropout_rate)
         estimate = estimate_function(torch.exp(weights_))
-        l = loss(estimate, targets)
+        l = loss(estimate, targets, normalization_factor)
         close = pct_close(estimate, targets)
 
         if i % progress_update_interval == 0:
diff --git a/src/microcalibrate/utils/metrics.py b/src/microcalibrate/utils/metrics.py
index db4d975..da55edf 100644
--- a/src/microcalibrate/utils/metrics.py
+++ b/src/microcalibrate/utils/metrics.py
@@ -8,17 +8,23 @@
 def loss(
     estimate: torch.Tensor,
     targets_array: torch.Tensor,
+    normalization_factor: Optional[torch.Tensor] = None,
 ) -> torch.Tensor:
     """Calculate the loss based on the current weights and targets.
 
     Args:
         estimate (torch.Tensor): Current estimates in log space.
         targets_array (torch.Tensor): Array of target values.
+        normalization_factor (Optional[torch.Tensor]): Optional normalization factor for the loss (handles multi-level geographical calibration).
 
     Returns:
         torch.Tensor: Mean squared relative error between estimated and target values.
     """
     rel_error = (((estimate - targets_array) + 1) / (targets_array + 1)) ** 2
+    if normalization_factor is not None:
+        rel_error *= normalization_factor
+    if torch.isnan(rel_error).any():
+        raise ValueError("Relative error contains NaNs")
     return rel_error.mean()