diff --git a/tabulate/__init__.py b/tabulate/__init__.py
index 12a2950..e1abad2 100644
--- a/tabulate/__init__.py
+++ b/tabulate/__init__.py
@@ -2385,11 +2385,38 @@ def tabulate(
                 break
             elif align != "global":
                 aligns[idx] = align
+    # Formats that escape special characters (latex, github) used to escape at
+    # render time, i.e. after padding, so escaped cells were wider than the
+    # measured column widths and rows misaligned (issue #244). Escape before
+    # widths are measured, then drop the escape rules from the row formatters
+    # to avoid escaping twice.
+    _is_colon_grid = tablefmt == "colon_grid"
+    if not isinstance(tablefmt, TableFormat):
+        tablefmt = _table_formats.get(tablefmt, _table_formats["simple"])
+    escape_map = None
+    for _rowfmt in (tablefmt.headerrow, tablefmt.datarow):
+        if isinstance(_rowfmt, DataRow) and _rowfmt.escape_map:
+            escape_map = _rowfmt.escape_map
+            break
+    if escape_map:
+
+        def _escape_text(s):
+            return "".join(escape_map.get(c, c) for c in s) if isinstance(s, str) else s
+
+        cols = [[_escape_text(v) for v in c] for c in cols]
+        headers = [_escape_text(h) for h in headers]
+        _repl = {}
+        for _name in ("headerrow", "datarow"):
+            _rowfmt = getattr(tablefmt, _name)
+            if isinstance(_rowfmt, DataRow) and _rowfmt.escape_map:
+                _repl[_name] = DataRow(_rowfmt.begin, _rowfmt.sep, _rowfmt.end)
+        if _repl:
+            tablefmt = tablefmt._replace(**_repl)
     minwidths = [width_fn(h) + min_padding for h in headers] if headers else [0] * len(cols)
     aligns_copy = aligns.copy()
     # Reset alignments in copy of alignments list to "left" for 'colon_grid' format,
     # which enforces left alignment in the text output of the data.
-    if tablefmt == "colon_grid":
+    if _is_colon_grid:
         aligns_copy = ["left"] * len(cols)
     cols = [
         _align_column(
diff --git a/test/test_output.py b/test/test_output.py
index ea3da87..1277041 100644
--- a/test/test_output.py
+++ b/test/test_output.py
@@ -2684,8 +2684,8 @@ def test_latex():
             r"\hline",
             r" strings   &   numbers (\$N\_0\$) \\",
             r"\hline",
-            r" spam      &           41.9999 \\",
-            r" eggs      &          451      \\",
+            r" spam      &              41.9999 \\",
+            r" eggs      &             451      \\",
             r"\hline",
             r"\end{tabular}",
         ]
diff --git a/test/test_regression.py b/test/test_regression.py
index 9555676..ca5e64f 100644
--- a/test/test_regression.py
+++ b/test/test_regression.py
@@ -211,13 +211,36 @@ def test_column_with_mixed_value_types():
     assert_equal(table, expected)
 
 
+def test_latex_escaped_cell_width():
+    "Regression: escaped chars widen cells; column widths must account for it (issue #244)"
+    data = {"col1": ["first", "a_long_second_row", "third_row"], "col2": [1, 2, 3]}
+    result = tabulate(data, tablefmt="latex", headers="keys")
+    expected = "\n".join(
+        [
+            r"\begin{tabular}{lr}",
+            r"\hline",
+            r" col1                 &   col2 \\",
+            r"\hline",
+            r" first                &      1 \\",
+            r" a\_long\_second\_row &      2 \\",
+            r" third\_row           &      3 \\",
+            r"\hline",
+            r"\end{tabular}",
+        ]
+    )
+    assert_equal(expected, result)
+    # the & separator must sit at the same column on every row
+    amp_cols = {line.index("&") for line in result.splitlines() if "&" in line}
+    assert len(amp_cols) == 1
+
+
 def test_latex_escape_special_chars():
     "Regression: escape special characters in LaTeX output (issue #32)"
     expected = "\n".join(
         [
             r"\begin{tabular}{l}",
             r"\hline",
-            r" foo\^{}bar     \\",
+            r" foo\^{}bar                                                      \\",
             r"\hline",
             r" \&\%\^{}\_\$\#\{\}\ensuremath{<}\ensuremath{>}\textasciitilde{} \\",
             r"\hline",
@@ -596,5 +619,5 @@ def test_asciidoc_without_trailing_whitespace():
 def test_github_escape_pipe_character():
     "Regression: github format must escape pipe character with a backslash (issue #241)"
     result = tabulate([["foo|bar"]], headers=("spam|eggs",), tablefmt="github")
-    expected = "| spam\\|eggs   |\n|:------------|\n| foo\\|bar     |"
+    expected = "| spam\\|eggs   |\n|:-------------|\n| foo\\|bar     |"
     assert_equal(expected, result)