diff --git a/NEWS.md b/NEWS.md
index c29f165..6538ff4 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -1,3 +1,6 @@
+# (development)
+* `as_survey_rep()` now has an argument `degf`, corresponding to the same argument in the survey function `svrepdesign()`. This argument can be useful for large data sets, since specifying a value for `degf` avoids a calculation which can be slow for very large data sets.
+
 # srvyr 1.2.0
 * `survey_prop()` now uses proportions as the default, which should confidence interval improve coverage, but does mean results may slightly change (#141, #142, thanks @szimmer)
 * New function `survey_corr()` calculates the correlation between 2 variables, (#150, #151, thanks @szimmer & @bschneidr)
diff --git a/R/as_survey_rep.r b/R/as_survey_rep.r
index 19ec236..e001f05 100644
--- a/R/as_survey_rep.r
+++ b/R/as_survey_rep.r
@@ -32,6 +32,10 @@
 #' @param fpctype Finite population correction information
 #' @param mse if \code{TRUE}, compute variances based on sum of squares
 #' around the point estimate, rather than the mean of the replicates
+#' @param degf Design degrees of freedom: a single number, or \code{NULL},
+#' in which case a value will be computed automatically, which can be slow
+#' for very large data sets. See \code{\link[survey]{svrepdesign}}
+#' for more details.
 #' @param ... ignored
 #' @param compress if \code{TRUE}, store replicate weights in compressed form
 #' (if converting from design)
@@ -73,7 +77,7 @@ as_survey_rep.data.frame <-
                     "other"), combined_weights = TRUE,
            rho = NULL, bootstrap_average = NULL, scale = NULL,
            rscales = NULL, fpc = NULL, fpctype = c("fraction", "correction"),
-           mse = getOption("survey.replicates.mse"), ...) {
+           mse = getOption("survey.replicates.mse"), degf = NULL, ...) {
     variables <- srvyr_select_vars(rlang::enquo(variables), .data)
     repweights <- srvyr_select_vars(rlang::enquo(repweights), .data)
     weights <- srvyr_select_vars(rlang::enquo(weights), .data)
@@ -85,6 +89,7 @@ as_survey_rep.data.frame <-
       repweights = repweights,
       weights = weights,
       data = .data,
+      degf = degf,
       type = type,
       combined.weights = combined_weights,
       rho = rho,
@@ -112,7 +117,7 @@ as_survey_rep.tbl_lazy <-
                     "other"), combined_weights = TRUE,
            rho = NULL, bootstrap_average = NULL, scale = NULL,
            rscales = NULL, fpc = NULL, fpctype = c("fraction", "correction"),
-           mse = getOption("survey.replicates.mse"), ...) {
+           mse = getOption("survey.replicates.mse"), degf = NULL, ...) {
 
     variables <- rlang::enquo(variables)
     repweights <- rlang::enquo(repweights)
@@ -134,6 +139,7 @@ as_survey_rep.tbl_lazy <-
       repweights = repweights,
       weights = weights,
       data = survey_vars_local,
+      degf = degf,
       type = type,
       combined.weights = combined_weights,
       rho = rho,
diff --git a/man/as_survey_rep.Rd b/man/as_survey_rep.Rd
index 00682f0..369f98a 100644
--- a/man/as_survey_rep.Rd
+++ b/man/as_survey_rep.Rd
@@ -26,6 +26,7 @@ as_survey_rep(.data, ...)
   fpc = NULL,
   fpctype = c("fraction", "correction"),
   mse = getOption("survey.replicates.mse"),
+  degf = NULL,
   ...
 )
 
@@ -44,6 +45,7 @@ as_survey_rep(.data, ...)
   fpc = NULL,
   fpctype = c("fraction", "correction"),
   mse = getOption("survey.replicates.mse"),
+  degf = NULL,
   ...
 )
 
@@ -105,6 +107,11 @@ weights have been averaged, gives the number of iterations averaged over.}
 \item{mse}{if \code{TRUE}, compute variances based on sum of squares
 around the point estimate, rather than the mean of the replicates}
 
+\item{degf}{Design degrees of freedom: a single number, or \code{NULL},
+in which case a value will be computed automatically, which can be slow
+for very large data sets. See \code{\link[survey]{svrepdesign}}
+for more details.}
+
 \item{compress}{if \code{TRUE}, store replicate weights in compressed form
 (if converting from design)}
 }
diff --git a/tests/testthat/test_as_survey_rep.r b/tests/testthat/test_as_survey_rep.r
index 7662faf..6c6456c 100644
--- a/tests/testthat/test_as_survey_rep.r
+++ b/tests/testthat/test_as_survey_rep.r
@@ -228,3 +228,18 @@ test_that("as_survey_rep works when using SDR/ACS method of replicate weights",
   expect_equal(c(out_survey[[1]], sqrt(attr(out_survey, "var"))),
                c(out_srvyr_acs[[1]][[1]], out_srvyr_acs[[2]][[1]]))
 })
+
+# ------------------------------------------------------------------
+# Test user-specified degrees of freedom
+# ------------------------------------------------------------------
+
+sdr_srvyr <- cbind(sdr_sample, as.data.frame(sdr_factors)) %>%
+  as_survey_rep(repweights = starts_with("REP_"),
+                weights = "weights",
+                type = "successive-difference",
+                combined = FALSE,
+                degf = 4)
+
+test_that("as_survey_rep accepts user-specified degf", {
+  expect_equal(degf(sdr_srvyr), 4)
+})